Evaluation of the potential of Riverbank Filtration in ... · Evaluation of the Potential for...

201
Evaluation of the Potential for Riverbank Filtration in Colombia Marcela Jaramillo Uribe Universidad Nacional de Colombia Facultad de Minas, Escuela de Geociencias y Medio Ambiente Medellín, Colombia 2015

Transcript of Evaluation of the potential of Riverbank Filtration in ... · Evaluation of the Potential for...

Evaluation of the Potential for Riverbank Filtration in Colombia

Marcela Jaramillo Uribe

Universidad Nacional de Colombia Facultad de Minas, Escuela de Geociencias y Medio Ambiente

Medellín, Colombia

2015

Evaluation of the Potential for Riverbank Filtration in Colombia

Marcela Jaramillo Uribe

Tesis presentada como requisito parcial para optar al título de: Doctor en Ingeniería – Recursos Hidráulicos

Director: Ph.D. Jaime Ignacio Vélez Upegui

Universidad Nacional de Colombia Facultad de Minas, Departamento de Geociencias y Medio

Ambiente Medellín, Colombia

2015

This is for my dad, my mom and my brother, the three apices of a triangle so uneven that it was just perfect.

Dad, you could not be here but I know you would be proud to see your little girl crossing the finish line. Thanks for everything you

thought me.

Mom, thanks for your support and for somehow finding the way to deal with my recurrent mood swings and never given up on me.

Sergio, you will always be the one who gave me the best present of all, my nephew Tomás, and I will always love you for that.

ACKNOWLEDGMENTS

First of all I want to thank my professor, Jaime Ignacio Velez, for his constant support and for never given up on me despite everything. Without his guidance, both professional and personal, this work would not have seen the light of day.

I also want to thank Professor Thomas Grischek from the University of Applied Sciences in Dresden because he welcomed me during my internship and was an inspiration of work and dedication.

My thanks and love to Oscar for being my rock during this past year. Without you I would have been lost.

Special thanks to the guys at Gotta Ingenieria in Medellin (Carlos, Mario, Oscar Estrada, Oscar Rueda, Cristian, Joa), for letting me use their facilities where I finally found the right place to write this document.

Finally, I want to say thanks to all the people at Gotta for their company, jokes, encouragement and pressure to finish this thesis. To Maria Luisa because, even from the distance, has always been there for me. To Mónica, for helping me to get out of the deepest hole. To Juancho, who was there from the very beginning and remains there unconditionally despite the circumstances. To all those people whose names I’m not mentioning but somehow helped me at some point during all these years.

To all of you, GRACIAS....TOTALES.

i

TABLE OF CONTENTS

RESUMEN ............................................................................................................................. xiii

ABSTRACT .............................................................................................................................. xv

1 INTRODUCTION ............................................................................................................... 1

1.1 MOTIVATION AND OBJECTIVE ................................................................................. 1

1.2 DOCUMENT OVERVIEW ........................................................................................... 5

2 THEORETICAL FRAMEWORK ........................................................................................... 7

2.1 INTRODUCTION ........................................................................................................ 7

2.2 SITING AND DESIGN ................................................................................................. 8

2.3 ATTENUATION PROCESSES DURING RBF ................................................................. 9

2.4 CONTAMINANT REMOVAL ..................................................................................... 10

2.4.1 Organic Contaminants .................................................................................... 10

2.4.2 Inorganic Compounds ..................................................................................... 10

2.4.3 Microbial Pathogens ....................................................................................... 11

2.4.3.1 Inactivation .............................................................................................. 11

2.4.3.2 Adsorption ............................................................................................... 11

2.4.3.3 Straining ................................................................................................... 11

2.4.3.4 Settling ..................................................................................................... 12

2.5 THE HYPORHEIC ZONE ............................................................................................ 12

2.6 RBF VS. CONVENTIONAL WATER TREATMENT ...................................................... 13

2.7 LIMITATIONS .......................................................................................................... 15

2.8 CONCLUSIONS ........................................................................................................ 17

2.9 ACKNOWLEDGMENTS ............................................................................................ 18

3 POTENTIAL RIVERBANK FILTRATION SITES IN COLOMBIA – A GEOMORPHOLOGICAL ANALYSIS ............................................................................................................................... 19

3.1 INTRODUCTION ...................................................................................................... 19

3.2 METHODS ............................................................................................................... 22

3.2.1 Site Selection ................................................................................................... 22

Evaluation of the Potential for Riverbank Filtration in Colombia

ii

3.2.2 Google Earth Digitalization ............................................................................. 23

3.2.3 Reach Analysis ................................................................................................ 24

3.2.4 GIS Processing ................................................................................................ 25

3.2.5 Reach-Morphology Type Definition ............................................................... 26

3.2.6 Hydraulic Geometry Analysis ......................................................................... 26

3.2.7 Streams Self-Cleaning Potential ..................................................................... 26

3.3 RESULTS AND DISCUSSION .................................................................................... 27

3.3.1 Site Selection in Colombia .............................................................................. 27

3.3.2 River Profiles................................................................................................... 29

3.3.3 Stream Reaches Digitalization ........................................................................ 31

3.3.4 Morphologic Classification ............................................................................. 35

3.3.5 Stream-Reach Features .................................................................................. 37

3.3.6 Stream Power as an Indicator of the River Self-Cleaning Capacity................ 42

3.4 CONCLUSIONS AND RECOMMENDATIONS ........................................................... 45

3.5 ACKNOWLEDGMENTS............................................................................................ 47

4 RIVERBANK FILTRATION AND WATER QUALITY AT TWO SITES IN ANTIOQUIA ........... 49

4.1 INTRODUCTION ..................................................................................................... 49

4.2 METHODS .............................................................................................................. 51

4.2.1 Site Selection .................................................................................................. 51

4.2.2 Water Sampling and Analysis ......................................................................... 52

4.2.3 Quality of the Bank Filtrate vs. the River Water ............................................ 52

4.2.4 Hydroclimatological Information ................................................................... 54

4.2.5 Hydrochemistry .............................................................................................. 54

4.2.6 Portion of River-Borne Water in the Bank Filtrate ......................................... 54

4.3 SANTA FE DE ANTIOQUIA ...................................................................................... 54

4.3.1 Site Description .............................................................................................. 54

4.3.1.1 Generalities ............................................................................................. 54

4.3.1.2 Hydroclimatology of the area ................................................................. 57

4.3.2 Results and Discussion ................................................................................... 58

4.3.2.1 Quality of the bank filtrate vs. the river water ....................................... 58

4.3.2.2 Effects of climate and geology in water chemistry ................................. 64

Table of Contents

iii

4.3.2.3 Hydrochemistry ....................................................................................... 67

4.3.2.4 Portion of river water in the bank filtrate ............................................... 69

4.3.2.5 Removal rates .......................................................................................... 71

4.4 GUARNE .................................................................................................................. 72

4.4.1 Site Description ............................................................................................... 72

4.4.1.1 Generalities .............................................................................................. 72

4.4.1.2 Hydroclimatology of the area .................................................................. 74

4.4.2 Results and Discussion .................................................................................... 75

4.4.2.1 Quality of the bank filtrate vs. the river water ........................................ 75

4.4.2.2 Effects of climate and geology in water chemistry ................................. 80

4.4.2.3 Hydrochemistry ....................................................................................... 81

4.4.2.4 Portion of river water in the bank filtrate ............................................... 82

4.4.2.5 Removal rates .......................................................................................... 83

4.5 CONCLUSIONS AND RECOMMENDATIONS ............................................................ 83

4.6 ACKNOWLEDGMENTS ............................................................................................ 86

5 PESTICIDES REMOVAL DURING RBF .............................................................................. 87

5.1 INTRODUCTION ...................................................................................................... 87

5.2 COLUMN EXPERIMENTS ......................................................................................... 90

5.2.1 Laboratory Work ............................................................................................. 90

5.2.1.1 Experiment setup .................................................................................... 90

5.2.1.2 Soil characterization ................................................................................ 91

5.2.1.3 Tracer tests .............................................................................................. 92

5.2.1.4 Sampling and analysis .............................................................................. 95

5.2.2 Inverse Modeling ............................................................................................ 99

5.2.2.1 Conceptual framework ............................................................................ 99

5.2.2.2 Results and discussion ........................................................................... 101

5.3 CONTAMINANT TRANSPORT MODEL ................................................................... 105

5.3.1 Solute Transport Model ................................................................................ 105

5.3.1.1 Methods ................................................................................................ 105

5.3.1.2 Results and discussion ........................................................................... 108

5.3.2 Contaminant Transport Model ..................................................................... 110

Evaluation of the Potential for Riverbank Filtration in Colombia

iv

5.3.2.1 Advection package ................................................................................ 111

5.3.2.2 Dispersion package ............................................................................... 112

5.3.2.3 Chemical reaction package ................................................................... 114

5.3.2.4 Other parameters required by MT3DMS .............................................. 116

5.3.2.5 Results and analysis .............................................................................. 117

5.4 CONCLUSIONS AND RECOMMENDATIONS ......................................................... 129

5.4.1 COLUMN EXPERIMENTS AND INVERSE MODELING ..................................... 129

5.4.2 GROUNDWATER FLOW AND CONTAMINANT TRANSPORT MODELING ...... 130

5.5 ACKNOWLEDGMENTS.......................................................................................... 132

6 FINAL CONCLUSIONS AND RECOMMENDATIONS ...................................................... 133

6.1 CONCLUSIONS...................................................................................................... 133

6.1.1 Potential RBF Sites in Colombia – A Geomorphological Analysis ................ 133

6.1.2 RBF and Water Quality at Two Sites in Antioquia ........................................ 134

6.1.3 Pesticides Removal ....................................................................................... 135

6.2 RECOMMENDATIONS FOR FURTHER RESEARCH ................................................. 135

7 BIBLIOGRAPHIC REFERENCES ..................................................................................... 141

APPENDICES........................................................................................................................ 151

v

FIGURES

Figure 2-1 Basic scheme of Riverbank Filtration .................................................................... 7

Figure 2-2 Attenuation processes during RBF ........................................................................ 9

Figure 2-3 Water chemistry changes in the hyporheic zone of some RBF sites .................. 13

Figure 2-4 Scheme of conventional steps during surface water treatment processes vs. RBF .............................................................................................................................................. 14

Figure 2-5 Processes for the treatment of raw river water and bank filtrate in Germany .. 14

Figure 3-1 Typical longitudinal river profile and associated processes ................................ 19

Figure 3-2 General distribution of alluvial channel types along a river longitudinal profile .............................................................................................................................................. 20

Figure 3-3 Relative trends in sediment supply (Qs) and transport capacity (Qc) in alluvial mountain channels ............................................................................................................... 21

Figure 3-4 Stream morphology classification framework .................................................... 21

Figure 3-5 Example of a digitalized reach and some of the components in width estimation .............................................................................................................................................. 24

Figure 3-6 Example of the diagram resulting from the estimation of the mean curvature radius, Rm, based on maximum and minimum local variances ............................................ 25

Figure 3-7 Example of the correction performed by the macros to a river profile extracted from the digital elevation model. ......................................................................................... 26

Figure 3-8 Localization of the selected gauging stations in Colombia ................................. 27

Figure 3-9 Examples of multiannual monthly values of discharge in m3/s and total suspended solids in kg/m3 for two of the selected stations in Colombia. ........................... 28

Figure 3-10 River profiles, after filtering, divided by river length ........................................ 30

Figure 3-11 Digitalized stream reaches in Colombia: Apartado, El Alambrado, El Limon and El Limonar ............................................................................................................................. 31

Figure 3-12 Digitalized stream reaches in Colombia: La Coquera, La Virginia, Peñas Blancas, Pte Lleras, Pte Real and Pte Texas ........................................................................................ 32

Figure 3-13 Digitalized stream reaches in Colombia: San Juan, Sta Rita, Pte Yopal and Pto Leon ....................................................................................................................................... 33

Figure 3-14 Digitalized stream reaches in Europe: Enns River and Rhine River ................... 33

Evaluation of the Potential for Riverbank Filtration in Colombia

vi

Figure 3-15 Digitalized stream reaches in the US: Great Miami River, Missouri River, Ohio River, Platte River, Raccoon River and Russian River ........................................................... 34

Figure 3-16 Box plots of the slopes for the pool-riffle reaches in Colombia, and Europe and the United States .................................................................................................................. 36

Figure 3-17 Average width, median curvature radius, bend amplitude, and bend wavelength, related to watershed area ............................................................................... 39

Figure 3-18 Curvature radius, amplitude and bend length, related to bankfull width, and the box plots correspondent to the ratios Rm/W, Amp/W and Lb/W .................................. 40

Figure 3-19 Scatter graphs of Rm to Lb, Rm to Amp and Amp to Lb, and the correspondent box plots ............................................................................................................................... 41

Figure 3-20 Relation between curvature radius, Rm, and bankfull width, W, and degree of confinement expressed in terms of the ratio between bend wavelength and bankfull width, Lb/W .......................................................................................................................... 41

Figure 3-21 Bankfull width and median curvature radius compared with bankfull discharge, and their correspondent box plots ..................................................................... 42

Figure 3-22 Box plots of the total suspended solids, discharge, total stream power and specific stream power for the data sets of the pool-riffle reaches in Colombia, and Europe and the US combined. .......................................................................................................... 44

Figure 4-1 Localization of the two field sites: Santa Fe de Antioquia and Guarne .............. 51

Figure 4-2 DEM and map of geology and aquifer potential in the area of Santa Fe ........... 55

Figure 4-3 Site photo of the SantaFe well at the beginning of the study ............................ 56

Figure 4-4 Google earth view of the alluvial terrace and the SantaFe Well at the beginning of the work ........................................................................................................................... 56

Figure 4-5 Google earth view of the location where the SantaFe Well as located before being washed off by the Cauca River ................................................................................... 57

Figure 4-6 Average monthly values of precipitation, temperature, and discharge in the area of Santa Fe .................................................................................................................... 58

Figure 4-7 Physical characteristics in Santa Fe samples and their MAV for drinking water 59

Figure 4-8 Chemical characteristics with known adverse effects on human health in Santa Fe samples and their MAV for drinking water ..................................................................... 60

Figure 4-9 Chemical characteristics with implications in human health in Santa Fe samples and their MAV for drinking water ........................................................................................ 60

Figure 4-10 Chemical characteristics with major economic consequences and indirect effects on human health in Santa Fe (Ca, Alkalinity, Cl-, Hardness) and their MAV for drinking water ...................................................................................................................... 61

Figures

vii

Figure 4-11 Chemical characteristics with major economic consequences and indirect effect on human health in Santa Fe (Fe, Mg, SO4

2-, PO43-) and their MAV for drinking water

.............................................................................................................................................. 62

Figure 4-12 Microbiological characteristics (total coliforms and E.coli) in Santa Fe samples .............................................................................................................................................. 62

Figure 4-13 Concentrations of bicarbonates, potassium, silica and sodium in Santa Fe samples ................................................................................................................................. 63

Figure 4-14 Total dissolved solids and total suspended solids in Santa Fe samples ............ 64

Figure 4-15 Biochemical oxygen demand, chemical oxygen demand and dissolved organic carbon in Santa Fe samples .................................................................................................. 64

Figure 4-16 Relationship of ions concentrations with river discharge, and precipitation in Santa Fe during 2010 ............................................................................................................ 65

Figure 4-17 Piper diagram of major ions in Santa Fe samples ............................................. 68

Figure 4-18 Stiff diagrams of water from the river and the well in Santa Fe ....................... 68

Figure 4-19 La Isla Sector and dug-wells within it, and correspondent Piper diagrams ...... 70

Figure 4-20 Percentage of river-borne water in the well compared to discharge of the river and precipitation in Santa Fe for the year 2010 ................................................................... 71

Figure 4-21 Average log-removal of turbidity, E.coli, total coliforms and TSS in Santa Fe .. 72

Figure 4-22 DEM and map of Geology of the Saint Nicholas Valley nearby La Mosca Creek .............................................................................................................................................. 73

Figure 4-23 Google earth view of the Omya well near La Mosca Creek .............................. 73

Figure 4-24 Site photo of the Omya well .............................................................................. 74

Figure 4-25 Average monthly values of precipitation and discharge in Guarne .................. 75

Figure 4-26 Physical characteristics in Guarne samples and their MAV for drinking water 75

Figure 4-27 Chemical characteristics with implications in human health in Guarne samples and their MAV for drinking water ......................................................................................... 76

Figure 4-28 Chemical characteristics with major economic consequences and indirect effects on human health in Guarne samples (Ca, Alkalinity, Cl-, Hardness) and their MAV for drinking water ................................................................................................................. 77

Figure 4-29 Chemical characteristics with major economic consequences and indirect effects on human health in Guarne samples (Fe, Mn, Mg, SO4

2-) and their MAV for drinking water ..................................................................................................................................... 77

Figure 4-30 Microbiological characteristics (total coliforms and E.coli) in Guarne samples 78

Figure 4-31 Concentration of bicarbonates, potassium and sodium in Guarne samples .... 79

Evaluation of the Potential for Riverbank Filtration in Colombia

viii

Figure 4-32 Total dissolved solids, total suspended solids, biochemical oxygen demand and chemical oxygen demand in Guarne samples ...................................................................... 79

Figure 4-33 Relationship of calcium concentration with discharge and precipitation in Guarne in 2010 ..................................................................................................................... 80

Figure 4-34 Piper diagram of major ions in Guarne samples ............................................... 81

Figure 4-35 Stiff diagrams of the water from the river and the well in Guarne samples .... 81

Figure 4-36 Percentage of river-borne water in the well related to discharge and precipitation in Guarne during 2010 .................................................................................... 83

Figure 4-37 Average log-removal of some of the chemical parameters measured in Guarne .............................................................................................................................................. 83

Figure 5-1 Experiment setup for the column tests .............................................................. 90

Figure 5-2 Cumulative particle-size distribution curve from the sieve analysis .................. 92

Figure 5-3 BTCs for the tracer tests ...................................................................................... 94

Figure 5-4 Lab results E 1-1 (Column 1 at 20oC) ................................................................... 97

Figure 5-5 Lab results E 2-2 (Column 2 at 10oC) ................................................................... 97

Figure 5-6 Lab results E 3-1 (Column 1 at 10oC) ................................................................... 98

Figure 5-7 Lab results E 3-2 (Column 2 at 10oC) ................................................................... 98

Figure 5-8 Lab results E 4-1 (Column 1 at 20oC) ................................................................... 98

Figure 5-9 Lab results E 4-2 (Column 2 at 20oC) ................................................................... 98

Figure 5-10 Concentration of 3,4 Dichloroaniline in the outlet samples with time ............ 99

Figure 5-11 BTC and observed values E 1-1 ....................................................................... 104

Figure 5-12 BTC and observed values E 2-2 ....................................................................... 104

Figure 5-13 BTC and observed values E 3-1 ....................................................................... 104

Figure 5-14 BTC and observed values E 3-2 ....................................................................... 104

Figure 5-15 BTC and observed values E 4-1 ....................................................................... 104

Figure 5-16 BTC and observed values E 4-2 ....................................................................... 104

Figure 5-17 Study area in Lößnitztal, Germany .................................................................. 106

Figure 5-18 Bottom elevations of the aquifer .................................................................... 106

Figure 5-19 Boundary conditions for the groundwater flow model .................................. 107

Figure 5-20 Recharge from precipitation and from run-off (black) ................................... 108

Figure 5-21 Isolines from scenario 1 of the GW flow model ............................................. 109

Figure 5-22 Comparison of calculated and observed heads for scenario 1 ....................... 109

Figures

ix

Figure 5-23 Isolines from scenario 2 of the GW flow model .............................................. 109

Figure 5-24 Comparison of calculated and observed heads for scenario 2 ....................... 109

Figure 5-25 Flowlines and travel time marks (every 60 days) for clean water (R = 1) ....... 110

Figure 5-26 BTCs and contaminant plumes for each contaminant (αL = 12.2 m and a αT = 0.1αL) ................................................................................................................................... 118

Figure 5-27 BTCs and contaminant plumes for Ametryn for different values of αL, and αT = 0.1αL .................................................................................................................................... 120

Figure 5-28 BTCs and contaminant plumes for Atrazine for different values of αL, and αT = 0.1αL .................................................................................................................................... 121

Figure 5-29 BTCs and contaminant plumes for Carbofuran for different values of αL, and αT = 0.1αL ................................................................................................................................. 122

Figure 5-30 BTCs and contaminant plumes for Diuron for different values of αL, and αT = 0.1αL .................................................................................................................................... 123

Figure 5-31 BTCs for each pesticide for different values of αL, and αT = 0.1αL .................. 124

Figure 5-32 Maximum concentration of contaminants vs. time from spilling for different values of αL .......................................................................................................................... 125

Figure 5-33 Maximum concentration vs. time from spilling for Ametryn and Atrazine at different values of αT/αL ..................................................................................................... 126

Figure 5-34 Maximum concentration vs. time from spilling for Carbofuran and Diuron at different values of αT/αL ..................................................................................................... 127

Figure 5-35 Contaminant plumes for Ametryn at αL=4.3 and 17.9 m, and αT/ αL = 0.001 and 0.1 ................................................................................................................................ 128

xi

TABLES

Table 1-1 Water-quality parameters that did not meet the standards in the second semester of 2009 .................................................................................................................... 1

Table 1-2 Source and amount of organic matter and suspended solids discharged into the surface water bodies in 2008.................................................................................................. 2

Table 1-3 Percentage of Colombian cities and of total surface water supply within the five hydrologic areas in which the Colombian territory is divided ................................................ 2

Table 3-1 Riverbank filtration sites in USA and Europe used for comparison with the selected potential sites in Colombia ..................................................................................... 23

Table 3-2 Information about the used gauging stations in Colombia .................................. 28

Table 3-3 Average values for discharge and total suspended solids at the gauging stations selected in Colombia ............................................................................................................. 29

Table 3-4 Reach morphology type according to the classification tree of Flores et al. (2006) .............................................................................................................................................. 35

Table 3-5 Results of the stream-reach processing through AutoCAD macros and MatLAB routines ................................................................................................................................. 37

Table 3-6 Total stream power (Ω) and specific stream power (ω) for all the analyzed reaches .................................................................................................................................. 43

Table 4-1 Water sampling dates at both studied sites ......................................................... 52

Table 4-2 MAV for physical, chemical and microbiological characteristics of drinking water, and other measured variables .............................................................................................. 53

Table 4-3 Gauging stations nearby the two studied sites .................................................... 54

Table 4-4 Portion of river-borne water in the bank filtrate in the SantaFe well .................. 70

Table 4-5 Portion of river-borne water in the bank filtrate in the Omya well ..................... 82

Table 5-1 Volume of pesticides used in Colombia in 2010 discretized by type of control .. 88

Table 5-2 Most used pesticides by active ingredient in Colombia in 2010 .......................... 89

Table 5-3 Mass and percent cumulative fraction of sediments retained in the sieves ....... 92

Table 5-4 Grain-size distribution........................................................................................... 92

Table 5-5 Parameters of the tracer tests .............................................................................. 93

Table 5-6 Parameters calculated from the tracer tests ........................................................ 93

Table 5-7 Number of experiments performed at different temperature, for each column 95

Evaluation of the Potential for Riverbank Filtration in Colombia

xii

Table 5-8 General data related to the sampling .................................................................. 96

Table 5-9 Initial concentrations selected for each pesticide and experiment ..................... 96

Table 5-10 Porosity value used to calculate the equivalent pore volume ........................... 97

Table 5-11 Main parameters of the column experiments used during inverse modeling 102

Table 5-12 Results of the modeling for each pesticide and each experiment................... 102

Table 5-13 Basic statistic parameters for the fitted values for each pesticide .................. 103

Table 5-14 Water heads at the boreholes used as the initial heads in the simulation ..... 108

Table 5-15 Retardation factor and travel time from the river to the well for each pesticide and for clean water ............................................................................................................ 110

Table 5-16 Values of αL found from the equation for hydrodynamic dispersion coefficient (DL) ...................................................................................................................................... 113

Table 5-17 Values for longitudinal dispersivity calculated with different empirical equations ............................................................................................................................ 114

Table 5-18 Values of bulk density and distribution coefficient for each pesticide ............ 115

Table 5-19 Value of the first-order degradation coefficient for each pesticide ................ 116

Table 5-20 Time periods defined for the transient simulation .......................................... 116

Table 5-21 Scenarios considered during contaminant transport simulation .................... 117

Table 5-22 Results of simulation scenario 1 ...................................................................... 117

xiii

RESUMEN

Una primera aproximación al potencial de la técnica de filtración de ribera (o RBF por sus siglas en inglés) para ser utilizada en Colombia como alternativa en el tratamiento de agua para consumo humano fue el principal foco de esta tesis de doctorado. La técnica de RBF consiste en extraer agua de una corriente superficial a través de un pozo perforado en la llanura aluvial de la misma y utilizar el acuífero como filtro no sólo físico sino también químico y biológico. RBF se ha utilizado en Europa por más de un siglo pero su estudio sistemático es relativamente reciente y hasta ahora, además de Estados Unidos, son pocos los países que han aprovechado, o incluso explorado, su potencial. En vista de los altos costos asociados con las técnicas convencionales de tratamiento de aguas superficiales que se utilizan en el país, sumado al hecho de que más de la mitad de la población no cuenta con acceso a agua de buena calidad y a que, aparentemente, se cuentan con escenarios apropiados para la implementación de la técnica pero no hay documentación sobre su uso, se decidió entonces abordar el problema desde tres puntos de vista diferentes que permitieran una evaluación global a las posibilidades reales de aplicar RBF en el país de una manera exitosa. En primer lugar se utilizó la geomorfología fluvial de tramo con el objetivo de comparar sitios donde actualmente se utiliza RBF y determinar la existencia de características comunes que puedan ser utilizadas para evaluar el potencial de un sitio nuevo con información escasa. Adicionalmente, se trató de establecer si la alta carga de sedimentos en suspensión de los ríos Colombianos podría ser una limitante a la utilización de la técnica. En segundo lugar, se escogieron dos sitios en Antioquia donde existían pozos localizados en llanuras aluviales y donde se había establecido la existencia de una conexión directa entre el acuífero y las corrientes. Allí se realizaron campañas de muestreo de agua para análisis de calidad a la luz de la normatividad vigente para establecer el grado de remoción de ciertos compuestos presentes en el agua del río en comparación con el agua del pozo. Además, se determinó el porcentaje de agua del río en el pozo utilizando un trazador químico. Finalmente, se evaluó el potencial de RBF en la remoción de algunos pesticidas que se emplean actualmente en Colombia, mediante ensayos de laboratorio y modelación numérica de transporte de contaminantes. En términos generales se encontraron resultados positivos que apuntan a que la implementación y uso de RBF en el país son posibles y se considera que la técnica es una alternativa promisoria en el tratamiento de agua potable. Es necesario, sin embargo, realizar más trabajo de investigación para poder entender de una manera más precisa algunos de los mecanismos detrás de la técnica y las posibles limitaciones asociadas a características particulares del país y los ríos locales, en especial lo que tiene que ver con fenómenos climáticos, carga de sedimentos en suspensión y dinámica de las corrientes.

Palabras clave: Filtración de ribera, interacción acuífero-río, geomorfología fluvial, calidad de agua, pesticidas

xv

ABSTRACT

A first approach to the potential of riverbank filtration (RBF) to be used in Colombia as an alternative in the treatment of drinking water was the main focus of this dissertation. This technology involves removing water from a surface current through a well on the floodplain and use the aquifer as a physical, chemical and biological filter. RBF has been used in Europe for over a century, but its systematic study is relatively recent and so far, besides the United States, few countries have used, or even explored, its potential. In view of the high costs associated with conventional surface water treatment techniques used in the country, plus the fact that more than half the population has no access to water of good quality, and that apparently there are appropriate settings for the implementation of technical but there is no documentation about its use, it was decided to approach the problem from three different points of view that would enable an overall assessment to the real possibilities of applying RBF in the country in a successful manner. Stream-reach geomorphology was used in order to compare sites where RBF is currently operating to determine the existence of common features that could be employed to evaluate the potential of a new site with scarce information. In addition, an attempt was made to establish whether the high load of suspended sediments of Colombian rivers could be a limitation to the use of the technique. Secondly, two sites were chosen in Antioquia with wells located in floodplains and where the existence of a direct connection between aquifers and currents had been established. There, water sampling campaigns were carried out for quality analysis in the light of the current regulations to establish the degree of removal of certain compounds in the river water compared with the well water. In addition, the percentage of river-borne water in the well was determined using a chemical tracer. Finally, the potential of RBF in the removal of some pesticides currently used in Colombia was evaluated through column experiments and numerical modeling of contaminant transport. Overall positive results were obtained, suggesting that the implementation and use of RBF in the country are possible, and therefore the technique is considered a promising alternative in the treatment of drinking water. It is necessary, however, to do more research in order to understand more precisely some of the mechanisms behind the technique and the possible limitations associated with specific characteristics of the country and the local rivers, especially what has to do with climatic events, suspended sediment load and current dynamics.

Key Words: Riverbank filtration, aquifer-river interaction, fluvial geomorphology, water quality, pesticides

1

1 INTRODUCTION

1.1 MOTIVATION AND OBJECTIVE

Water scarcity affects one third of the population worldwide (WHO, 2009). In 2008 there were over 800 million people lacking access to improved drinking water sources, i.e. water sources that are protected from outside contamination, in particular fecal matter. This situation is much more critical in developing regions where the coverage in rural areas is only 76% compared to 94% in urban sectors (WHO – UNICEF, 2010). The circumstances are getting worse due to population growth, urbanization processes, and increased resource use in both homes and industries (WHO, 2009).

Colombia is one of the countries with higher availability of natural water in the world, with approximately 8000 km3, 28% corresponding to surface water reserves and 72% to groundwater (IDEAM, 2010). The country also has encouraging figures in terms of coverage of improved drinking water sources: in 2008 an average of 92% of the population had access to such sources, 99% in urban areas and 73% in rural regions (WHO, 2011), and the number of people who gained access to improved drinking water sources from 1990 to 2008 was over 12 million (WHO – UNICEF, 2010). However, according to the Colombian Ombudsman’s Office (Defensoría del Pueblo, 2010), in the second semester of 2009 only 32% of the cities received water that was safe for human consumption, i.e. when the water quality meets 95% of the standards established in the Colombian legislation. Table 1 shows some of the main water quality parameters that failed to reach the minimum admissible number (mAN) according to the legislation.

Table 1-1 Water-quality parameters that did not meet the standards in the second semester of 2009 (Defensoría del Pueblo, 2010)

Parameter % of cases IRCA

Total coliforms 71,1 15

E. coli 54,9 25

Residual chlorine 53,9 15

Color 34,5 6

Turbidity 33,8 15

pH 25,4 1,5

Total iron 11,5 1,5

Aluminum 2,0 3

Nitrites 1,9 3

TOC 1,0 3

As shown in Table 1-1, the most failed parameters are also those with higher risk scores according to the Drinking Water Quality Risk Index or IRCA for its Spanish acronym (República de Colombia, 2007), i.e. with higher risk for human health. Those parameters

Evaluation of the Potential for Riverbank Filtration in Colombia

2

are: total coliforms, E.coli, residual chlorine from the disinfection process, color, and turbidity.

One of the reasons for the high presence of total coliforms and E.coli in the drinking water in Colombia could be the fact that only 15% of the country’s waste water is somehow treated before its discharge into the environment (Defensoría del Pueblo, 2010). Table 1-2 presents the amount (in tons) of organic matter and total suspended solids (TSS) originated from domestic and industrial activities discharged in the rivers in 2008 (IDEAM, 2010). The organic matter is represented by the Biochemical Oxygen Demand (BOD), and the Chemical Oxygen Demand (COD).

Table 1-2 Source and amount (in tons) of organic matter and suspended solids discharged into the surface water bodies in 2008 (Source: IDEAM, 2010)

BOD COD TSS

Domestic 474.045 938.556 992.083

Industry 211.497 631.098 89.176

TOTAL 685.542 1.569.654 1.081.259

Other contaminants of consideration are nutrients, mercury and pesticides. In 2008, 117000 tons of total nitrogen, 29400 tons of total phosphorous, and 178 tons of mercury are believed to have been discharged in the rivers as a result of domestic and industrial activities for the first two cases, and from gold and silver mining for the last one (IDEAM, 2010). The sales of pesticides in 2009 were of the order of 12.8 kilotons in its solid form and 35000 m3 in its liquid form (ICA, 2010). However, the country does not have yet a system that permits to estimate the amount of pesticides that are being discharged annually into the water bodies (IDEAM, 2010).

Of special concern in Colombia is that the highest pressure in rivers and water bodies in terms of BOD, COD and TSS occurs in the hydrographic areas of Magdalena – Cauca and Caribe, where most of the country’s population is concentrated (63% of the Colombian cities) and only represent 28% of the total national surface water supply (IDEAM, 2010). A Hydrographic Area is defined as natural territory composed of big drainage systems where the lotic waters flow either to towards the ocean, a lake, or a main river (IDEAM, 2010). Colombia is divided in 5 hydrographic regions: Magdalena-Cauca, Caribe, Pacifíco, Orinoco and Amazonas. Table 1-3 shows the percentages of Colombian cities and the total surface water supply for the five hydrographic areas in Colombia.

Table 1-3 Percentage of Colombian cities and of total surface water supply within the five hydrologic areas in which the Colombian territory is divided (Source: IDEAM, 2010)

Hydrographic Area % of Colombian cities within the hydrographic area % Total surface water supply

Magdalena – Cauca 63 13,2

Caribe 10,8 7,8

Orinoco 11 27

Pacífico 9 12

Amazonas 4.7 38,7

Chapter 1 – Introduction

3

Other possible causes for the poor drinking water quality, according to the Colombian Ombudsman’s Office, besides the quality of the water sources, are (Defensoría del Pueblo, 2007):

Absence of drinking water treatment plants (DWTP): most of the sixth class municipalities do not have potable water treatment plants, and its inhabitants are supplied with raw water from different sources. Sixth class municipalities are those districts or municipalities with a population exceeding ten thousand (10000) inhabitants, and free destination income per year not exceeding fifteen thousand (15000) legal monthly minimum wages (Article 2nd of Law 617 of 2000). The legal monthly minimum wage in Colombia in 2011 was $536000 Colombian pesos ≈ US$280.

Treatment plants with very basic procedures: DWTP are not built with the specific technical regulations and are inefficient, or their capacity is not consistent with the requirements of the municipality.

Inadequate operation of DWTP: in some places the procedures at the DWTP are not adjusted to the water quality of the source, or to the conditions of the plant itself, but procedures from other plants or regions are implemented without making adjustments to the local conditions.

Low technical knowledge of the people operating the DWTP: in many cases the plumber is the same person that operates the plant, without knowing how to execute correctly the different procedures during the water treatment.

Problems in distribution networks: inefficient and obsolete. In general, the water that comes out of the DWTP in Colombia is of better quality than that analyzed in the pipe network or in households.

Scarce financial resources destined to DWTP: The resources for drinking water and basic sanitation are not always implemented or are diverted to different purposes. If the budget for water supply (aqueduct) had been properly executed, the 100% of the country’s population would have had access to improved sources of drinking water by the year 2005 (DNP, 2011).

The lower the water quality of the source, the more complex and expensive the treatment processes, increasing the risk of diarrhoeal diseases like cholera, typhoid and dysentery, and other water-borne diseases. Also, it forces people to search for different sources, even at great distances from population centers, encouraging the storage of water in the houses, thus incrementing the chance of breeding of mosquitoes carrying dengue, malaria and other diseases, and liquid contamination (WHO, 2009). In Colombia, the main water-related diseases are typhoid and paratyphoid fever, hepatitis A, leptospirosis, dengue and malaria (Defensoría del Pueblo, 2010). These diseases are related either to the treatment or water distribution, but also to storage conditions.

In order to prevent waterborne diseases is essential that people have access to water of adequate quality and sufficient quantity, as well as sanitation facilities. As the number of people with access to improved services increases, the incidence of water-borne diseases

Evaluation of the Potential for Riverbank Filtration in Colombia

4

will diminish. Preventing these diseases also contributes to poverty reduction, keeping in mind that the poorest people in the world are the ones that pose the greatest health problems of this type (WHO – UNICEF, 2008).

It is therefore necessary to evaluate alternatives that allow the population to have access to water of adequate quality, thus reducing the risks to public health, and the treatment and maintenance costs, since the conventional surface water systems are not always applicable due to the high costs they generate. In these aspects, the exploitation of alluvial aquifers, hydraulically connected to surface water, has advantages such as high production potential, ease and economy of extraction (Doussan et al., 1997).

By pumping, surface water can be forced through the bed and banks of the river, mixing with water in the aquifer, and experiencing a reduction in contaminant levels due to physical, chemical and biological processes that take place both between the two types of water, and between them and the substrate. This process, known as Riverbank Filtration (RBF) has proven to be an efficient technology in the pretreatment of surface water for over a century in Europe and half a century in the United States, replacing or reducing the number of processes commonly used for this task (Kuehn and Mueller, 2000).

In addition to its extensive use in Europe and the U.S., some cities in India, Brazil and South Korea have begun using RBF for water supply (Ray, 2008), and its potential has been studied in India (Sandhu et al., 2010) and Egypt (Abdalla and Shamrukh, 2011). In Colombia, however, there are no studies reporting the use of RBF technology as an alternative system for drinking water treatment, but the reasons for this to happen are still unknown, and there are apparently favorable scenarios for its implementation. The latter allows speculating that the RBF technique itself is being used in our country but empirically or even unconsciously.

The current understanding of the processes and mechanisms behind RBF is still very empirical (Tufenkji et al., 2002). Because the efficiency of RBF depends on specific site conditions (hydraulics, hydrogeology and hydrochemistry), there are limitations to define general guidelines on the conditions necessary for the protection and optimization of RBF sites (Hiscock and Grischek, 2002). The completion of more studies on the performance of RBF under a wide range of environmental conditions, may improve the design, operation and monitoring of RBF systems (Tufenkji et al., 2002).

Given the considerations presented above, it was decided to conduct a doctoral thesis with the main objective of exploring the potential of riverbank filtration in Colombia as a technology for the pretreatment of surface water for human consumption, and contribute to the state of global knowledge of the RBF technique, especially on its application in tropical regions where specialized studies are scarce.

Chapter 1 – Introduction

5

1.2 DOCUMENT OVERVIEW

The document is organized as follows:

This chapter (Chapter 1) presents a brief introduction on the motivations and main objective of this dissertation.

Chapter 2 provides a general theoretical framework on basic concepts about riverbank filtration, starting from its definition, and covering subjects such as the design of RBF wells, the attenuation processes that take place during the passage of water from the river to the well, the type of contaminants that can be removed by bank filtration and the physical, chemical or biological processes responsible for such reductions, and some information on the influence of the hyporheic zone in the quality of the bank filtrated water. Also, RBF is compared to the traditional steps in conventional water treatment plants, and a summary of the main limitations associated to the technique is presented at the end of the chapter.

The following three chapters, which represent the core of the thesis, are independent of each other and explore the potential of RBF in Colombia from different points of view.

Chapter 3 contains the results of geomorphological analysis performed to selected stream-reaches in potential RBF sites Colombia and in existent RBF sites in Europe and the United States. The chapter presents similitudes and differences that allow drawing some conclusions on the possibility of selecting a suitable site for riverbank filtration using readily available geo-processing tools and simple hydraulic geometry relationships, before on-site measurements are made. Furthermore, the chapter includes data on average discharge and sediment load, and comparison between the sites in Colombia and those abroad, in order to determine possible limitations of local sites due to physical clogging and the self-cleaning capacity of the rivers.

The effectiveness of RBF in the removal (or reduction in the concentration) of common water quality parameters is evaluated in Chapter 4. There, the results of the analysis of water samples collected in two different sites of the department of Antioquia, both from the river and the well located next to it, were compared to the maximum admissible values according to the current Colombian legislation for drinking water standards. Also, the portion of river-borne water in the bank filtrate or degree of mixing of the water from the river and the aquifer was estimated by using chloride as a conservative tracer.

The fate of agrochemicals during RBF is studied in Chapter 5. In this chase, five pesticides used in Colombia were selected to perform column tests to determine properties such as retardation factor, dispersion coefficient and degradation coefficient. Subsequently, the results were used as input of a contaminant transport model from an RBF site in Germany to determine that way the potential of RBF to remove these organic compounds from the river water.

Finally, Chapter 6 presents the most significant conclusions and contributions of this doctoral thesis and some important recommendations to take into account for future research.

7

2 THEORETICAL FRAMEWORK

2.1 INTRODUCTION

Alluvial aquifers are widely used as a groundwater source in many countries, mainly due to their high production potential, proximity to demand areas, their ease, and economy of extraction (Doussan et al., 1998). With the pumping of wells located in an alluvial plain hydraulically connected to a river it is possible to generate a hydraulic gradient so that surface water is forced to flow through the bed and the banks of the river (Figure 2-1). During this process, known as riverbank filtration (RBF), a reduction in the concentration of pollutants is achieved by physical, chemical, and biological processes that take place, between the surface water and groundwater, and with the substrate (Tufenkji et al., 2002).

Figure 2-1 Basic scheme of Riverbank Filtration

The reduction of pollution levels is accomplished by a number of processes including physical filtration, microbial degradation, ion exchange, precipitation, sorption, and dilution (Ray et al., 2002). Other factors that also contribute to the treatment are the river water and the groundwater quality, the porosity of the medium, the water residence time in the aquifer, temperature and pH conditions of water, and oxygen concentrations (Kuehn and Mueller, 2000).

In addition to the removal of pollutants (particles, microorganisms, organic, and inorganic compounds, etc.) there are two additional advantages of RBF. The first is relative to the fact that the flow through the aquifer acts as a barrier against concentration peaks that may result from accidental spills of pollutants. The second is the regulation on the temperature variations in the river water: during winter, when the air temperatures are

Evaluation of the Potential for Riverbank Filtration in Colombia

8

low, the filtered water is usually warmer than surface water, and in summer it is cooler. The lowest variation in temperature improves the quality and further processing of the bank filtrate (Hiscock and Grischek, 2002).

Riverbank filtration technology has been a common practice in Europe for over 100 years, particularly in countries such as Switzerland where 80 % of drinking water comes from RBF wells, 50 % in France, 48 % in Finland, 40 % in Hungary, 16 % in Germany, and 7 % in the Netherlands (Tufenkji et al., 2002). In Germany, for example, 75 % of the city of Berlin depends on RBF, and in Düsseldorf RBF has been used since 1870 as the main drinking water supply (Schubert, 2002a). In the United States, on the other hand, this technique has been used for nearly half a century, especially in the states of Ohio, Kentucky, Indiana, Illinois, among others (Ray et al., 2002). Other countries that have recently started implementing RBF for drinking water supply are India (Sandhu et al., 2010), China, and South Korea (Ray, 2008).

2.2 SITING AND DESIGN

Local factors such as river hydrology, hydrogeological site conditions (i.e. aquifer thickness and hydraulic conductivity), and the aims of water withdrawal determine not only the capacity of the wells, but also the travel time of the bank filtrate, and distance between the river and the well (Grischek et al., 2002).

According to Schubert (2006a), two principles have to be taken into account when designing RBF sites. The first one is the retention time of the river water in the aquifer which the author states should be at least three weeks, and depends on the distance of the wells form the riverbank and the pumping rate. The second principle is the verification of the position of the drawdown curve relative to the riverbed so the steeper part of the curve is completely confined to the aquifer, to minimize mechanical clogging.

Riverbank filtration wells can be designed either vertically (as the most common practice especially for the extraction of low water quantities) or horizontally (for higher extraction rates). Horizontal wells (sometimes with a radial pattern), also known as collector wells, are usually directed toward the river and extract water from beneath the riverbed, whereas vertical wells extract water along the riverbed (Ray et al., 2002). Also, RBF wells can be distributed parallel to the riverbank in galleries or groups (Grischek et al., 2002).

Geomorphology of the river is an important factor to determine the suitability of a site for RBF. Erosion regions or areas where very fine particles are deposited should be avoided (e.g. upstream of dams, river mouth regions) (Schubert, 2002b). It means the ideal reach is the one where transport occurs, where there is a rather broad flood plain, gentle slopes, and riverbed sediments in the sizes of sands and bigger. Bedrock rivers are not suitable. The hydraulic gradient of the river can provide information about grain-size distribution, mean flow velocity in the river, and bed load transport capacity (Schubert, 2002b). However, a first selection of river reaches apt for RBF can be made by looking at topographic and geologic maps, aerial photographs, satellite images, Google Earth images, among others.

Chapter 2 – Theoretical Framework

9

Experience has shown that the inside loop of a meander is the best place to do it to avoid mechanical clogging (Schubert, 2002a). For example, Schubert (2006b) presents two cases in the Lower Rhine region near Düsseldorf. The Flehe waterworks is located on the outer bend of the river and during flood events the clogged area spread out the driving head beneath the river, and the water level decreases abruptly. On the contrary, at the “Auf dem Grind” well, located inside the loop of a meander there is no significant variation in the driving head between the river and the groundwater during flood events. These differences between two sites separated no more than 3 km can be explained by the location of the well with respect to the meander. As Schubert (2006b) explains, the outer section of the bend, close to the wells, is almost impermeable because the ground is fully clogged and “paved”, while the inner side has a movable ground. Therefore, clogging is more strongly marked by bank filtration wells along the outer section of a bend.

Finally, Grischek et al. (2002) compiled available information from RBF systems in the United States and Europe, and concluded that the most important parameters for success during RBF are the flow path length, the thickness of the aquifer, and the infiltration area in the river. Finally, the authors conclude that the siting and design of an RBF system does not only depend on hydrogeological factors, but also on technical, economical, regulatory, and land-use factors.

2.3 ATTENUATION PROCESSES DURING RBF

Four attenuation processes are involved in RBF: hydrodynamic, mechanical, biological, and physicochemical (Doussan et al., 1998). Figure 2-2 summarizes the main attenuation processes during riverbank filtration.

Figure 2-2 Attenuation processes during RBF (Hiscock and Grischek, 2002)

Hydrodynamic processes include convective-dispersive transport, and dilution. The aquifer acts as a filter for the temporal variation of the pollutant compounds in the river caused by accidental (or intentional) spills, which, due to the connection between the river and

Evaluation of the Potential for Riverbank Filtration in Colombia

10

the aquifer, represent a risk of contamination to the groundwater. As a result, high frequency variations in the surface water quality are reduced in groundwater. Beyond smoothing fluctuations in river water quality, dilution takes place when the river water mixes with groundwater, which is usually of higher quality, further enhancing the quality of bank filtrate (Doussan et al., 1998).

The most important mechanical processes for the improvement of water quality is the natural filtration of fine sediments, particulate organic matter, and pathogens, especially in the first few meters from the river to the well (Hiscock and Grischek, 2002; Doussan et al., 1998). A disadvantage of physical filtration is associated with the obstruction or the clogging of the porous media, as will be explained later.

The biological processes that occur during RBF are directly dependent on the type of microorganisms that inhabit the aquifer (Doussan et al., 1998; Gollnitz et al., 2005). The metabolic processes of these microorganisms mainly determine the final quality of filtered water.

Finally, physicochemical processes are associated with sorption, precipitation reactions, flocculation, coagulation, and redox reactions (Doussan et al., 1998). All these processes govern the removal of particles from the porous media, affecting the concentration and the behavior of metals and other inorganic compounds, thus having implications for the chemical evolution of water.

2.4 CONTAMINANT REMOVAL

2.4.1 Organic Contaminants

Organic pollutants such as pesticides, herbicides, odorous compounds, oil sub-products, and pharmaceuticals are of great concern for water quality. Riverbank filtration has been extensively used for drinking water pretreatment in places with such pollution problems (Jüttner, 1995, 1999). The removal and the behavior of organic compounds during RBF depends on factors specific to pollutants such as the hydrophobicity of the compound, the potential for biochemical degradation, the amount of organic matter in the aquifer, microbial activity, infiltration rate, biodegradability, etc. (Tufenkji et al., 2002). Another aspect that apparently influences the removal of certain organic contaminants such as antimicrobial residues is the redox condition of the aquifer together with the travel time (Heberer et al., 2008).

Although RBF has proven to be a good pretreatment technique for a large number of organic compounds, it has been found that some certain pesticides, pharmaceuticals, and halogenated organic compounds are more resistant to removal (Kuehn and Mueller, 2000; Ray et al., 2002; Tufenkji et al., 2002).

2.4.2 Inorganic Compounds

The main and most common processes that control the transport and fate of inorganic compounds in RBF are (Tufenkji et al., 2002):

Chapter 2 – Theoretical Framework

11

Redox reactions: manganese and iron oxides are mobilized under reducing conditions and adsorbed, precipitated, or co-precipitated under oxidizing conditions.

Microbial degradation of organic matter: this can alter the geochemical conditions and mobilize metals usually associated with natural organic matter (NOM) such as copper and cadmium.

Dilution: high concentrations of inorganic compounds in river water are depleted by the mixing of surface and groundwater.

2.4.3 Microbial Pathogens

The primary processes for the removal of pathogens during soil passages are inactivation, attachment to the aquifer grains (adsorption), straining, and sedimentation (Schijven et al., 2002). These processes depend on the climate (temperature, rainfall, etc.), the nature of the porous media (clay content, moisture-holding capacity), and the type of microorganism (Yates et al., 1985).

2.4.3.1 Inactivation

Inactivation is a pathogen’s loss of ability to infect host cells. This happens with time because of the disruption of coat proteins and the degradation of nucleic acids (Gerba, 1984).

Among the main factors that influence virus inactivation rates are temperature, adsorption to particulate matter and soil, and microbial activity (Schijven et al., 2002). Many authors consider temperature to be the most significant factor for inactivation; however, temperature sensitivity and inactivation speed appear to be virus-dependent (de Roda Husman et al., 2009). For instance, at normal groundwater temperatures (8 to 25 ˚C), inactivation rates for pathogens are very low: 1.42/day for Shigella sp. (McFeters et al., 1974), 0.51/day for Salmonella sp. (Keswick et al., 1982), and 0.33/day for E. coli O157 (Rice et al., 1992).

2.4.3.2 Adsorption

Adsorption is defined as the sum of the electrostatic, hydrophilic, and steric interactions between viruses and the media itself (Aronino et al., 2009). The interactions that take place between a microorganism (especially a virus) and soil particles depend on their surface characteristics (Schijven and Hassanizadeh, 2000). Those characteristics may be altered by changes in pH, ion strength, multivalent ions, organic matter (Gerba, 1984), and temperature (Reed et al., 1999).

2.4.3.3 Straining

Straining is the physical removal of microbial particles, and depends on their size and that of the pore throats. According to McDowell-Boyer et al. (1986), if the ratio of the diameter of the media to the diameter of the particle is greater than 20, then the straining is not

Evaluation of the Potential for Riverbank Filtration in Colombia

12

considerable, but if the same ratio is between 10 and 20 the removal of particles is significant. Below a ratio of 10 there is not penetration of the particle through the porous media.

2.4.3.4 Settling

Also known as sedimentation in pores, settling is determined by Stoke’s settling velocity which states that the settling velocity directly depends on the mass density difference between the particle and the fluid, particle size, and gravity force, and inversely, on the fluid viscosity. In the case of the settling of pathogens in groundwater, the flow velocity is an important factor, as the settling is most likely to occur at low groundwater velocities (Schijven et al., 2002).

Several authors have confirmed the efficiency of RBF in the reduction of pathogenic microorganisms. Wang (2002), for example, reports a 3.8-log reduction for total coliform, and 2.0-log units for HPC bacteria in Louisville, Kentucky [% removal = 100 – 10^(2-x), where x is the number of log removal, i.e. 1 log = 90 % removal]. Weiss et al. (2002) found a removal of 2.9 to 3.4 logs for Clostridium, 2.3 to 3.0 logs for E. coli C, and 1.6 to 2.0 logs for E. coli Famp, in three RBF systems along the Ohio, Wabash, and Missouri rivers in the United States. Also, at the Greater Cincinnati Water Works in Ohio, a minimum of 4-log removal of Giardia and Cryptosporidium was reported by Gollnitz et al. (2003). During a 24-month study at the Central Wyoming Regional Water System, Gollnitz et al. (2005) found an average of 2.1-log reduction for surrogates of Giardia and Cryptosporidium (total coliform, E. coli, enterococci, total aerobic endospores, algae, diatoms, and turbidity). In Germany, Schubert (2002b, 2002c) reports a 5-log removal of bacteria, viruses, and parasites at the Flehe Waterworks in Düsseldorf (which began operation in 1870).

2.5 THE HYPORHEIC ZONE

The hyporheic zone is defined as the transition zone between surface water and groundwater in the alluvial aquifer. This area experiences biogeochemical activity which is much more intense than surface water or groundwater (Gibert et al., 1997). This biochemical activity is reflected in the complex and dynamic gradients of light, temperature, pH, redox potential, oxygen content, and organic matter content (Tufenkji et al., 2002).

The most important biochemical change that takes place in the hyporheic zone in some RBF sites is the creation of an anaerobic zone (Figure 2-3). This occurs as a result of the rapid oxygen consumption involved in microbial activity associated with the degradation of organic matter and organic contaminants. These anaerobic conditions increase the activity of denitrifying bacteria and sulfate-reducing bacteria, creating a highly reductive area, which causes the dissolution of manganese and iron oxides, thus affecting the quality of filtered water (Bourg and Bertin, 1993).

Chapter 2 – Theoretical Framework

13

Figure 2-3 Water chemistry changes in the hyporheic zone of some RBF sites (Tufenkji et al., 2002)

Another consequence of microbial activity in the hyporheic zone is the formation of biofilms that can block the pores of the aquifer and reduce its permeability. Clogging of the pores of the aquifer can also be induced by the retention of fine sediments (< 2 mm) in the hyporheic zone, especially the precipitation of sulfides and oxides (Gibert et al., 1997).

2.6 RBF VS. CONVENTIONAL WATER TREATMENT

Coagulation-flocculation followed by sedimentation, filtration and disinfection (often chlorination) is used worldwide in the treatment of surface water before storage and distribution to the consumers (Ndabigengesere and Narasiah, 1998).

Based on German experience with bank filtration along the Rhine River, Kuehn and Mueller (2000) compared the processes for the treatment of raw river water versus those needed for bank filtrate, and found that some conventional surface water treatment processes can be eliminated if RBF water is used, i.e. coagulation-flocculation, sedimentation and, sometimes, filtration (Figure 2-4). Bank-filtrate water usually requires additional treatment before disinfection, such as activated carbon filtration (ACF), ozonation → filtration → ACF, or aeration → filtration (Figure 2-5). This is especially common in rivers with high concentration of ammonia, organic compounds, and micro-contaminants.

Evaluation of the Potential for Riverbank Filtration in Colombia

14

Figure 2-4 Scheme of conventional steps during surface water treatment processes vs. RBF

The content of dissolved oxygen (DO) in river water is an important factor for determining the need for further bank filtrate treatment before disinfection. For example, if conditions become anaerobic either due to the low DO content in the river water or because of a high oxygen demand due to the presence of microorganisms in the soil, iron and manganese will undergo chemical reduction and solubilize in the water, requiring their removal by further treatment such as aeration and filtration, before disinfection (Mucha et al., 2006).

Figure 2-5 Processes for the treatment of raw river water and bank filtrate in Germany (Kuehn and Mueller, 2000)

Ammonia content in river water also determines the quality of bank filtrate (Kuehn and Mueller, 2000) and the necessity of performing other treatment processes before disinfection. This is because nitrification (the oxidation of ammonia into nitrites and then into nitrates) is an aerobic process that consumes oxygen in river water. If DO content in river water is already low, then anaerobic conditions will develop, and the iron and manganese which would otherwise be present as precipitated in their oxidized form (Fe3+ and Mn3+), will be solubilized (as explained above) (Tufenkji et al., 2002).

Chapter 2 – Theoretical Framework

15

Biological processes can also contribute to oxygen consumption and thus to the solubilization of iron and manganese. These processes depend on different factors such as pH, temperature, DO, and the content of organic compounds in the water (Kuehn and Mueller, 2000).

Kim and Corapcioglu (2002) and Kim et al. (2003) demonstrated that colloidal particles and dissolved organic matter (DOM) present in the water can increase the mobility of contaminants by the reduction of retardation, and that they can change the degree of sorption and microbial degradation. If this happens, the efficiency of RBF will be compromised since those contaminants might be present in the bank filtrate. Depending on the nature and final concentration of the contaminant, the bank-filtrate will need further treatment before disinfection.

When chlorination is used as the disinfection process, chlorine reacts with natural organic matter (NOM) and halides (Cl-, Br-) to form disinfection byproducts (DBPs) such as trihalomethanes (THMs). Those DBPs are harmful for human health, especially brominated trihalomethanes (THMs-Br) which are suspected to be much stronger carcinogens and mutagens than their chloride-containing analogues (Richardson, 2003).

Although the RBF has proven to be efficient in removing organic matter (total and dissolved organic carbon, TOC and DOC), as well as certain DBPs (Weiss et al., 2003), if chlorination is used as the disinfection method, there might be an increase in THM concentration. It could then be recommended to use ACF before disinfection to reduce the amount of TOC and then the formation of THMs.

2.7 LIMITATIONS

In addition to the inability of RBF to remove certain biological, inorganic, and organic contaminants, limitations associated with the hydrology and dynamics of the river and groundwater cannot be ignored. On the contrary, these aspects should be taken into account when RBF is considered as a pretreatment solution (Schubert, 2002a).

Changes in the hydraulic gradient from the river to the aquifer, and in the hydraulic conductivity of the alluvial deposits, generate changes in the pore water velocity as well as in the retention time, which may limit or change the biogeochemistry activity that takes place in the hyporheic zone. Finally, changes in water temperature affect not only the hydraulic conductivity due to the reduction of the viscosity of water, but also the rate of biogeochemical processes and microbial activity, which could weaken the final quality of the filtered water (Vanek, 1997).

Fluctuations in the river stage alter water saturation, biofilms content, geochemistry, and even the structure of the RBF system, thus affecting the performance of the treatment. Such variations can affect the flow and transport characteristics of the whole system because the unsaturated region, which may not have the same removal potential as the saturated zone, could be infiltrated by river water during an increased stage level and, therefore, the filtered water obtained will have a poorer quality (Schubert, 2000).

Evaluation of the Potential for Riverbank Filtration in Colombia

16

The efficiency and performance of RBF can also be compromised by scouring processes carried out on the river bed and banks when the flow rates are very high, which is a fairly common problem (Gollnitz et al., 2004). The consequent loss of fine sediments, responsible for the low permeability of the river bed (usually one to three orders of magnitude below the permeability of the aquifer) could be a problem in the treatment process since the efficiency of filtration decreases. Another effect of scouring is the removal of microorganisms which are essential for improving the quality of river water in the hyporheic zone.

Another limitation associated with RBF is the obstruction or clogging of the porous media. There are four types of clogging: mechanical, physical, biological, and chemical.

Mechanical clogging is defined as the blocking of flow through porous media due to the entrapped gas that is released from or dissolved into the porous media can change the permeability of media, thus preventing the water from making its way through the aquifer (Zhou et al., 2010).

Microorganisms produce a range of poorly soluble gases such as carbon dioxide (CO2), nitrogen (N2), hydrogen (H2), oxygen (O2), and methane (CH4). However, O2 and H2 have rarely been reported to accumulate in soil (Baveye et al., 1998). Nitrogen gas produced by denitrification, and methane produced during methanogenesis have been further identified as the most prevalent gases thus entrapped.

Physical clogging is caused by the continual percolation of river water containing suspended matter due to well pumping (Schubert, 2006c). This process is governed by the dynamics of the river (runoff, erosion, transport, and sedimentation), the quality of the river water, the location of the wells related to river geomorphology, and the distance between the well and the riverbank (Schubert, 2006b).

The settling velocity of fine particles inside the aquifer pores depends not only on the physical properties of the sediments (size, weight, etc.), but also on the water velocity (river discharge) and viscosity (Hubbs, 2006a). This means that a particle will settle easier on a low-flow low-viscosity river.

Sometimes reductions on the hydraulic conductivity of the aquifer (manifested by increasing drawdown) are related not only to clogging, but also to low water temperatures (Schafer, 2006; Hubbs, 2006b).

Hubbs (2006a) presents another possibility for mechanical clogging that has to do with the development of unsaturated conditions under the riverbed. In that case, the forces of the overlying water column compress the loose soil into a structure with much lower hydraulic conductivity.

Biological clogging is caused by excessive biomass accumulation in the riverbed (Engesgaard et al., 2006). The distribution of biomass has been assumed to be uniform as a biofilm covering the grain surfaces. However, it has been found that microorganisms grow in micro-colonies, and that plugs of biomass, rather than biofilms, are responsible for bioclogging (Seifert and Engesgaard, 2007).

Chapter 2 – Theoretical Framework

17

Biofilms are the normal environment for most microbial cells (especially bacteria) in many natural and artificial habitats. They are an aggregation of microorganisms growing in a matrix of expolysaccharides, by adhering either to each other or to a grain surface (Sutherland, 2001). The production of polysaccharides can be favored by factors such as temperature, redox potential, the availability and nature of organic substrate, nitrogen availability, O2 concentration, and the physiological status of microorganisms (Baveye et al., 1998).

Due to the fact that each type of microorganism has its own optimum growth temperature (Schijven et al., 2002) the extension and formation rates of biofilms in a riverbed will depend highly on the temperature of the water infiltrating the riverbed.

Chemical clogging is caused by the precipitation of compounds into the pores of the aquifer. Some factors thought to influence chemical clogging are iron, ammonia, and nitrate concentrations, and the hardness of the water (Caldwell, 2006).

High loads of biodegradable substances in the river water can lead to chemical clogging due to strong changes in redox-potential and pH values, which may cause the precipitation of substances into the pores of the aquifer (Schubert, 2002a). These changes are strongly related to the microbial activities in the riverbed, since it controls the redox conditions of the medium to a high degree, preventing or stimulating the precipitation of inorganic substances (Caldwell, 2006). That is the reason why some authors refer to biochemical clogging.

In general, biochemical clogging occurs beyond the infiltration area where mechanical clogging predominates (Schubert, 2006b).

Clogging may be limited or removed by the self cleaning potential of the river: scouring (Schubert, 2002a). The scouring process is the result of the shear forces imparted by the movement of water flowing into the river, and the resistance to the movement provided by the riverbed itself, which is a function of the river slope, the vertical velocity profile, and sediment transport (Hubbs, 2006a). However, in some cases, larger sediments form a sort of armor that prevents sediment resuspension from the bed by the action of flood waves (Stuyfzand et al., 2006).

2.8 CONCLUSIONS

During RBF, pumping pressure in the alluvial aquifer adjacent to the river can force the water to percolate from the river into the aquifer. In this path, a series of physical and biogeochemical processes take place, including physical filtration, adsorption, absorption, biodegradation, and dilution. Thus, riverbank-filtrate often shows better quality than river water, making its treatment for human consumption a lot easier and less expensive.

The removal of sediment, organic and inorganic compounds, and pathogens takes place during the first meters from the river in what is known as the hyporheic zone, which usually presents reducing conditions due to the high microbial activity which consumes

Evaluation of the Potential for Riverbank Filtration in Colombia

18

the oxygen in the water. Within this zone there are important biochemical processes and redox reactions that affect groundwater quality.

The efficiency of RBF depends on local conditions including the hydrology and hydrogeology of the site, the geochemistry of water (from both the river and the aquifer), the geochemistry of microbial populations, and associated metabolic activity. This is the reason why is difficult to define general procedures for identifying appropriate sites to implement the RBF technique, as well as the expected efficiency of the process.

One limitation on the efficiency of RBF is the clogging of the bed and the banks of the river, which decreases the hydraulic conductivity in the hyporheic zone. This clogging can be caused by the infiltration of fine sediments, gas entrapment, biofilm formation related to microbiological activity, or the precipitation and co-precipitation of inorganic compounds, being the first of these the most influential factor in clogging formation.

Although the practice of riverbank filtration has been used in Europe for more than a century, the current understanding of the processes and mechanisms behind this technique are still very empirical. Besides, its use in tropical countries is almost nonexistent. At the PARH (Posgrado en Aprovechamiento de Recursos Hidraulicos) of the Universidad Nacional de Colombia in Medellin, we believe there is a great potential for RBF in our country, and that is why we are exploring this issue with more detail.

2.9 ACKNOWLEDGMENTS

This chapter is the result of an extensive bibliographic review as part of the project “Integral management of joint use of surface and groundwater”, co-funded by COLCIENCIAS and the Universidad Nacional de Colombia. To both institutions the author wishes to express her most sincere gratitude.

19

3 POTENTIAL RIVERBANK FILTRATION SITES IN COLOMBIA – A GEOMORPHOLOGICAL ANALYSIS

3.1 INTRODUCTION

One of the problems that arises when looking for the right place for a riverbank filtration site is the fact that the efficiency of the technology depends on local conditions such as river hydrology (discharge, sediment load, etc.) and hydrogeology of the site (aquifer geometry, hydraulic conductivity, etc.), the geochemistry of water (from both, the river and the aquifer), and the biochemistry of microbial populations, just to name a few. This is the reason why it is difficult to define general procedures for identifying appropriate sites to implement RBF: because it is governed by site-specific conditions (Grischek et al., 2002).

It is also known, that the geomorphology of a stream is an important factor to determine the suitability of a site for RBF. Schubert (2002a), states that the hydraulic gradient of the river can provide information about grain-size distribution, mean flow velocity in the river, and bed load transport capacity. He also affirms that, based on the typical longitudinal river profile proposed by Shumm (1977), where three regions can be distinguished based on the dominant processes (erosion in the upper part, transport in the middle and deposition in the lower part), erosion reaches and areas where very fine particles are deposited should be avoided as RBF sites, due to the problems associated with scouring and clogging, respectively (Figure 3-1).

Figure 3-1 Typical longitudinal river profile and associated processes (Fitzpatrick et al., 2006)

Schubert (2006c) also suggests that the inside of meander loops are preferred locations for RBF sites, because of the higher proportions of river-borne water that can be extracted as a result of the natural cross flow between the upstream and the downstream of the loop, and the fact that the movable riverbed along the border of the river in such sites promote the self-cleaning process of clogged areas.

Evaluation of the Potential for Riverbank Filtration in Colombia

20

These appreciations about the most desirable geomorphological settings for an RBF site have been obtained in an empirical way, just as much of the knowledge that exists on the technology. It is important, then, to find out whether existing, successful and efficient RBF sites share similar geomorphological characteristics that could be easily measured through the use of geospatial data, aerial images, satellite images, etc. in order to help when looking for new, potential RBF sites.

In order to do this, some of the RBF sites in Europe and the United States where RBF is being used in a successful way were analyzed from a geomorphological point of view, and then compared to some pre-selected sites in Colombia. This was done by following the methodology proposed by Jimenez (2015), who, as part of his doctoral dissertation, analyzed the morphology of 123 stream reaches in Colombia and Venezuela, allowing the distinction of universal classes of downstream hydraulic geometry (DHG) for bankfull width. His work has a strong basis on the findings of Montgomery and Buffington (1997) and Flores et al. (2006), who studied channel-reach morphologies in mountain drainage basins and considered channel-reach as an important scale parameter in hydraulic geometry.

On one hand, because the assumption of a simple longitudinal river profile model does not really match most natural rivers, Montgomery and Buffington (1997) proposed an expansion of such channel-reach classification, as shown in Figure 3-2. In their scheme, alluvial channels can be classified in five different morphologies according to their slope and position within the channel network: cascade, step-pool, plane-bed, pool-riffle, and dune-ripple. Each morphology has its own particularities according to the confinement degree, the presence of a well-developed channel migration zone, and the relationship between sediment supply (Qs) and transport capacity (Qc). Thus, the downstream sequence shown in Figure 3-2 is accompanied by a progressive decrease in valley-wall confinement (hence, a better established floodplain) and transport capacity, whereas sediment supply increases with drainage area (Figure 3-3).

Figure 3-2 General distribution of alluvial channel types along a river longitudinal profile (Montgomery and Buffington, 1997)

Chapter 3 – Potential Riverbank Filtration Sites in Colombia

21

Figure 3-3 Relative trends in sediment supply (Qs) and transport capacity (Qc) in alluvial mountain channels (Montgomery and Buffington, 1997)

Flores et al. (2006), on the other hand, went a little bit further and investigated whether, besides channel-reach bed slope, the morphologies defined by Montgomery and Buffington (1997) also depended on basin scale, more specifically drainage area. The authors proposed a classification tree model based on the sediment transport capacity along the stream, expressed in terms of the index of specific stream power, S0A0.4, where S0 is the bed slope and A is the contributing watershed area associated to the channel-reach. They found a threshold for S0 equal to 0.025, below which the reaches can be classified as transport-limited (Qc << Qs), as is the case for pool-riffle and plane-bed morphologies, and above such value they can be considered supply-limited (Qc>>Qs), i.e. either step-pool or cascade reaches (Figure 3-4). Besides, they also found that the index S0A0.4 allows a better distinction of stream types into each coarse-category than using only the downstream slope.

Figure 3-4 Stream morphology classification framework proposed by Flores et al. (2006)

It is known that one of the limitations of RBF is the reduction of the hydraulic conductivity of the riverbed and banks due to clogging, which can be caused by the continual percolation of river water containing suspended sediments (Schubert, 2002a; 2006c). However, clogging may be reduced or removed by the self-cleaning potential of the river, i.e. scouring (Schubert, 2002a). Therefore, being able to determine such river potential to

Evaluation of the Potential for Riverbank Filtration in Colombia

22

erode the banks and bed is considered in this work as an important part of the determination of sites with potential for RBF.

According to Hubbs (2006a), the scouring process is the result of the shear forces imparted by the movement of water flowing into the river, and the resistance to the movement provided by the riverbed itself, which is a function of the river slope, the vertical velocity profile, and sediment transport. The author proposes an indirect way of evaluating the river capacity for scouring and sediment transport by using the downstream variations of the longitudinal profile, according to the equation presented by Julien (1998): 𝜏𝑏 = 𝛾𝑚ℎ𝑛𝑆0, where 𝜏𝑏 is the riverbed shear stress [N/m2], 𝛾𝑚 is the specific weight of the mixture of water and sediments [N/m3], which Hubbs (2006a) suggests can be simplified as the specific weight of clean water, ℎ𝑛 is the normal flow depth [m], and 𝑆0 is the bed slope [dimensionless].

Another way to measure the main driving forces acting in a channel, and thus the river’s capacity to transport sediment and to erode, is through the joint effect of channel gradient and discharge, also known as stream power (Bagnold, 1977; Newson et al., 1998; Ferguson, 2005; Bizzi and Lerner, 2013). Total stream power (in Wm-1) is defined as 𝛺 = 𝛾𝑄𝑆0, whereas specific stream power (in Wm-2) can be expressed as 𝜔 = 𝛺/𝑊, where 𝛾 is the unit weight of water (9800 N/m3), Q is discharge (m3/s), S0 is the bed slope (m/m), and W is the channel bankfull width (m).

The fact that the formula for shear stress that Hubbs (2006a) recommends encompasses a series of simplifications and requires knowledge of flow depth, which is not always available unless on-site measurements have been carried out, and given that stream power can be easily estimated using geo-processing tools, makes the latter a better variable to evaluate the self-cleaning potential of a river to counteract for possible clogging. However, one of the limitations of using stream power instead of shear stress is that it only takes into consideration the river driving forces but excludes the resisting forces such as sediment size distribution, or limiting factors such as sediment availability (Bizzi and Lerner, 2013).

This chapter has two main purposes then. The first one is to determine if, by using readily available geo-processing tools and simple hydraulic geometry relationships, it is possible to select a site suitable for riverbank filtration before on-site and detailed measurements need to be made. The second is to find out if, under those premises, Colombian rivers possess the necessary qualities to act as potential RBF sites.

3.2 METHODS

3.2.1 Site Selection

Colombia

Aiming at covering as much area of the Colombian territory as possible, and including different hydrological and geomorphological settings (elevation above sea level, stream width, discharge rates, sediment load, precipitation regime, etc.), a set of 14 stream

Chapter 3 – Potential Riverbank Filtration Sites in Colombia

23

reaches spread across the country, with a well-developed alluvial plain, was selected using Google Earth satellite images.

These sites were chosen because they were close to gauging stations belonging to the IDEAM (Colombian Institute of Hydrology, Meteorology and Environmental Studies), and where both discharge and sediment load were being measured. The idea behind this was to have reliable information in order to be able to make comparisons between the sites, not only on a geomorphological scale but also on a hydrological one.

Finally, the selected reaches were picked because there are population centers of different sizes nearby, so the results obtained here could provide the basis to implement a real alternative for drinking water supply in specific communities.

USA and Europe

The sites in the United States and Europe, against which the selected sites in Colombia were compared, were some of the RBF systems discussed by Caldwell (2006), and are presented in Table 3-1.

Table 3-1 Riverbank filtration sites in USA and Europe used for comparison with the selected potential sites in Colombia (Caldwell, 2006)

Location Waterworks Wellfield

USA

Great Miami River (Cincinnati, OH)

Greater Cincinnati Waterworks Charles M. Bolton

Missouri River (Kansas City, KS)

Kansas City Board of Public Utilities Nearman WTP

Ohio River (Louisville, KY)

Louisville Water Company B.E. Payne WTP

Platte River (Ashland, NE)

City of Lincoln Ashland

Raccoon River (Des Moines, IA)

Des Moines Waterworks Platte River

Russian River (Sonoma Co., CA)

Sonoma County Water Agency Wohler and Mirabel

EUROPE

River Enns (Steyr, Austria)

Ennskraft Company KRB02

Rhine River (Düsseldorf, Germany)

Flehe, Grind and Staad Flehe, Grind and Staad

3.2.2 Google Earth Digitalization

Using the satellite images in Google Earth, the margins and main axis of the selected reaches were digitalized, trying to follow the path of what would be the higher geomorphically-effective discharge, i.e. the discharge that, on average, transports the largest portion of a river’s annual sediment load (Wolman and Miller, 1959).

Evaluation of the Potential for Riverbank Filtration in Colombia

24

3.2.3 Reach Analysis

Different morphological features were found using the AutoCAD Visual Basic macros and MATLAB routines developed by Jimenez (2015). These include the reach-averaged width, W, sinuosity, and the median curvature radius, Rm.

To calculate the average width, W, an AutoCAD VB macro was used as follow: i) the digitalized margins and axis are transformed from a Google Earth format into a shape file; ii) the axis is resampled using a regular spacing, close to the expected width; iii) perpendicular transects are defined along the resampled axis intersecting the digitalized margins; iv) the macro estimates the width corresponding to each axis node and calculates the average and standard deviation of the width, as well as the length of the reach, and its sinuosity. Figure 3-5 shows one example of the different components that intervene in the estimation of W.

The estimation of a representative curvature radius, Rm, was made with a MATLAB routine

that iteratively re-meshes the axis using a regular spacing Xi within the interval [0.01W – 10W], where W is the reach-averaged bankfull width previously estimated. Afterwards, a radius of curvature is assigned to every three consecutive points along the re-meshed axis and, when the entire mesh interval is sampled (using incremental values of 0.01W) a

diagram of X versus Rm is obtained (Figure 3-6).

Figure 3-5 Example of a digitalized reach and some of the components in width estimation (Jimenez, 2015)

According Jimenez (2015) “The first maximum local variance and the second minimum local variance of the obtained diagram allow the identification of a W fraction for which the median radius estimation remains almost invariant. Such fraction is used to assess the Rm value corresponding to each stream reach.” The MATLAB routine also estimates the median reach amplitude and the wave length of the meander bends.

Chapter 3 – Potential Riverbank Filtration Sites in Colombia

25

Figure 3-6 Example of the diagram resulting from the estimation of the mean curvature radius, Rm, based on maximum and minimum local variances (Jimenez, 2015)

3.2.4 GIS Processing

Two different Geographic Information Systems were used during the development of this work: HidroSIG 4.0 and QGIS 2.2.

HidroSIG 4.0 is a set of plugins implemented into MapWindow GIS [www.mapwindow.org] and it was developed at the Graduate Program in Water Resources of the School of Geosciences and Environment at the Universidad Nacional de Colombia [http://www.medellin.unal.edu.co/~hidrosig/]. This tool package was utilized to delineate watersheds and the main water courses at the selected sites.

On the other hand, the free and open source Geographic Information System, QGIS 2.2, [http://www.qgis.org/en/site/] was used to extract, from digital elevation models, the longitudinal profiles of the analyzed rivers, and to calculate watershed areas (in km2). Those profiles were then corrected with an excel macro developed by Jimenez (2015). Figure 3-7 shows an example of the type of correction performed by the mentioned macro. The blue solid line shows the original elevations from the DEM, whereas the red dotted line is the profile resulting from the filtering.

The digital elevation models (DEM) and the flow direction maps used here corresponded to 15 seconds arc pixel size HydroSHEDS maps [http://hydrosheds.cr.usgs.gov/index.php].

Evaluation of the Potential for Riverbank Filtration in Colombia

26

Figure 3-7 Example of the correction performed by the macros to a river profile extracted from the digital elevation model.

3.2.5 Reach-Morphology Type Definition

The classification method proposed by Flores et al. (2006) was used to categorize each stream reach as either transport-limited (pool-riffle or plane-bed), or supply-limited (step-pool or cascade) (Figure 3-4).

The reach slope, S0, was calculated as the difference in heights between the upstream and downstream edges of the reach (taken after carrying out the filtering procedures), divided by the length of the reach calculated in QGIS. The watershed area, A, was calculated using GIS software as explained above.

3.2.6 Hydraulic Geometry Analysis

The planform features found during the stream-reach processing were then compared to each other (Rm, Amp and Lb vs W, Rm vs Lb, Rm vs Amp and Amp vs Lb), against drainage area (W, Rm, Amp and Lb vs A) and against discharge (W and Rm vs Q), in order to find hydraulic geometry relations that could allow making distinctions or generalizations between the RBF sites selected in the US and Europe, and those potential sites selected in Colombia for comparison. This was done mainly with the use of scatter graphs in log-log axis scale, and box and whisker plots.

3.2.7 Streams Self-Cleaning Potential

Total stream power (Ω) and specific stream power (ω) were used to compare the reaches according to their self-cleaning potential. The reach slope and the bankfull width were those values obtained previously during the reach-processing. The stream discharge used for the Colombian rivers was the flood having a 2.33 years recurrence interval, whereas for the sites in the US the discharge was the one reported by Caldwell (2006) as bankfull discharge. In the case of the two RBF sites in Europe, the mentioned author does not report the bankfull discharge but only the average discharge, so the calculations were made using such value.

Chapter 3 – Potential Riverbank Filtration Sites in Colombia

27

3.3 RESULTS AND DISCUSSION

3.3.1 Site Selection in Colombia

The gauging stations, and hence channel reaches, selected in Colombia for the analysis and comparison with functioning RBF sites in Europe and the US, are shown in Figure 3-8. The name of the gauging stations (which will be used to identify the site from now on), the name of the river where they are located, the daily average variables that are measured in each station (discharge, sediment concentration and/or sediment load), and the geographical coordinates of the sites are shown in Table 3-2. In those stations where suspended sediment concentration, or total suspended solids (TSS), was not measured, the value was calculated dividing the suspended sediment load by the discharge (ṁ/Q), previously having converted ṁ to kilograms per second.

Figure 3-8 Localization of the selected gauging stations in Colombia

At the time the information was received, the station with the longest registry period was La Virginia, with 41 years for discharge and 37 years for both, sediment concentration and sediment load; whereas the station with the shortest registry period was Sta Rita with 14 years for discharge, 8 years for sediment load and no registry for sediment concentration. However, the parameter with less data was sediment load for Peñas Blancas station, with only 4 years of registry.

Evaluation of the Potential for Riverbank Filtration in Colombia

28

Table 3-2 Information about the used gauging stations in Colombia

ID Station Name Department River Parameters* Lat. Long.

1201706 Apartado Antioquia Apartado Q, C, ṁ 7.88 -76.62

2612701 El Alambrado Valle del Cauca La Vieja Q, C, ṁ 4.41 -75.88

3207711 El Limon Meta Guejar Q, C, ṁ 3.37 -74.03

2119715 El Limonar Tolima Sumapaz Q, C, ṁ 4.23 -74.62

2624702 La Coquera Antioquia Cauca Q, C**, ṁ 7.96 -75.20

2617703 La Virginia Risaralda Cauca Q, C, ṁ 4.89 -75.88

2316701 Peñas Blancas Antioquia Magdalena Q, C**, ṁ 6.95 -73.95

3501702 Pte Lleras Meta Meta Q, C, ṁ 4.10 -72.94

2308715 Pte Real Antioquia Negro Q, C, ṁ 6.14 -75.38

4701716 Pte Texas Putumayo Putumayo Q, C, ṁ 0.58 -76.58

3521701 Pte Yopal Casanare Cravo Sur Q, C, ṁ 5.37 -72.42

1602706 Pto Leon Norte de Santander Zulia Q, C, ṁ 8.37 -72.43

2703703 San Juan Antioquia Nechí Q, C**, ṁ 7.48 -74.92

3307701 Sta Rita Vichada Vichada Q, C**, ṁ 4.87 -68.36 *Q: discharge [L

3/T], C: suspended sediment concentration [M/L

3], ṁ: suspended sediment load [M/T]. **C

not measured but calculated as ṁ/Q.

The daily data for discharge (Q) and total suspended solids (TSS) was taken to a monthly resolution for all the available years, and then the average multiannual monthly values were calculated and plotted against time (in months), as shown in Appendix A. In the graphs, the blue bars represent Q in cubic meters per second, and the orange lines denote TSS in kilograms per cubic meter.

Two examples of the graphs contained in Appendix A are presented in Figure 3-9. From the figures, it becomes clear that some stations present a bimodal hydrological behavior of Q whilst others display a unimodal distribution of the same parameter, which goes in concordance with the findings of Mesa et al. (1997) about Colombian hydrological cycles. In general terms, the TSS graphs show a similar trend than the ones for discharge.

Figure 3-9 Examples of multiannual monthly values of discharge in m3/s (blue bars) and total suspended solids in kg/m3 (orange line) for two of the selected stations in Colombia.

The highest discharge values were those measured at Peñas Blancas station (located on the Magdalena River), whereas the lowest were those for Apartado station, on the river that has the same name (see Table 3-3). Also, the highest sediment concentration values

Chapter 3 – Potential Riverbank Filtration Sites in Colombia

29

were registered at La Coquera station, on the Cauca River, and the lowest at Sta Rita, on the Vichada River. Thus, the highest average discharge is of the order of 2900 m3/s, whilst the lowest is close to 5 m3/s. In terms of TSS, the maximum concentration measured was 1.4 kg/m3, whereas the lowest was 0.02 kg/m3. The fact that the data is covering a wide range of both discharge and TSS, allow making comparisons at different scales.

Table 3-3 Average values for discharge and total suspended solids at the gauging stations selected in Colombia

Station Q (m3/s) TSS (kg/m

3)

Apartado 4.65 0.241

El Alambrado 55.07 0.781

El Limon 32.06 0.216

El Limonar 43.18 0.098

La Coquera* 1,380.98 1.409

La Virginia 545.75 0.435

Peñas Blancas* 2,912.96 0.469

Pte Lleras 434.04 0.677

Pte Real 7.22 0.111

Pte Texas 500.15 0.082

Pte Yopal 89.34 0.525

Pto Leon 120.8 0.236

San Juan* 479.2 0.390

Sta Rita* 1,032.38 0.024 *TSS estimated as ṁ/Q

3.3.2 River Profiles

The longitudinal profiles of the rivers, after filtering as explained in section 3.2.4, are presented in Figure 3-10. Due to the amount of rivers, it was decided to divide them according to river length. Thus, rivers with a longitude up to 200 km are shown in Figure 3-10 (a), those longer than 200 km but shorter than 500 km are presented in Figure 3-10 (b), rivers with length between 500 and 1400 km can be seen in Figure 3-10 (c), and finally, those longer than 1400 km are in Figure 3-10 (d). The square blue markers (and associated blue lines) represent those sites where there are at present, successfully functioning RBF sites, while the round black markers (and associated black lines) denote those potential RBF sites selected in Colombia where the gauging stations are located.

The longest river is the Missouri with a total length of about 3900 km, followed by the Putumayo with 2110 km and the Ohio with 2030 km. The shortest, on the other hand, is the Apartado River with barely 35 km in longitude. As expected, the head drop between the highest point of the river profile and the RBF site, either existent or potential, is evidently higher for the Colombian rivers than for those in the US and Europe.

In all cases, the existing or potential RBF sites are located in reaches with low slopes, hence where transport or sedimentation processes are the most important. However, this simple classification does not allow differentiations between such processes and therefore is not useful as a classification tool.

Evaluation of the Potential for Riverbank Filtration in Colombia

30

Figure 3

-10 R

iver pro

files, after filterin

g, divid

ed b

y river length

: (a) < 2

00

km, (b

) 20

0 – 5

00

km, (c) 50

0 – 140

0 km

, and

(d) >1

400

km

(c)

(a)

(d)

(b)

Chapter 3 – Potential Riverbank Filtration Sites in Colombia

31

3.3.3 Stream Reaches Digitalization

For each selected gauging station, a stream-reach was digitalized in Google Earth, ensuring that the length of the reach was higher than 15 times its width since that is considered to be a useful scale over which to relate stream morphology to channel processes (Montgomery and Buffington, 1997; Flores et al., 2006; Jimenez, 2015). The selected reaches include straight, meandering and braided, in order to cover different types of channels. Figure 3-11 to Figure 3-13 show the Colombian reaches chosen at every station and the populated sites nearby. The yellow line represents the main axis, and the light blue ones are the margins. The same procedure was done for the RBF sites in Europe (Figure 3-14) and the United States (Figure 3-15) used for comparison. Note that at the Rhine River, as well as at the Russian, there is more than one well field (three on the former and two on the latter); however, the processing and analysis was made for one reach as a whole.

Figure 3-11 Digitalized stream reaches in Colombia: Apartado, El Alambrado, El Limon and El Limonar

Apartado (Apartado R.)

El Alambrado (La Vieja R.)

El Limon (Guejar R.)

El Limonar (Sumapaz R.)

Evaluation of the Potential for Riverbank Filtration in Colombia

32

Figure 3-12 Digitalized stream reaches in Colombia: La Coquera, La Virginia, Peñas Blancas, Pte Lleras, Pte Real and Pte Texas

Peñas Blancas (Magdalena R.)

Pte Lleras (Meta R.)

Pte Real (Negro R.)

Pte Texas (Putumayo R.)

La Coquera (Cauca R.)

La Virginia (Cauca R.)

Chapter 3 – Potential Riverbank Filtration Sites in Colombia

33

Figure 3-13 Digitalized stream reaches in Colombia: San Juan, Sta Rita, Pte Yopal and Pto Leon

Figure 3-14 Digitalized stream reaches in Europe: Enns River and Rhine River

Enns R. (Steyr, Austria)

Rhine R. (Dusseldorf, Germany)

Pte Yopal (Cravo Sur R.)

Pto Leon (Zulia R.)

San Juan (Nechi R.)

Sta Rita (Vichada R.)

Evaluation of the Potential for Riverbank Filtration in Colombia

34

Figure 3-15 Digitalized stream reaches in the US: Great Miami River, Missouri River, Ohio River, Platte River, Raccoon River and Russian River

Great Miami R. (Cincinnati, OH)

Missouri R. (Kansas City, MO)

Ohio R. (Louisville, KY)

Platte R. (Ashland, NE)

Raccoon R. (Des Moines, IA)

Russian R. (Sonoma County, CA)

Chapter 3 – Potential Riverbank Filtration Sites in Colombia

35

3.3.4 Morphologic Classification

The stream-reaches were morphologically classified following the scheme posed by Flores et al. (2006), shown in Figure 3-4. The reach slope was calculated with data from the DEM and the filtered longitudinal profile of the river, and the basin area was defined using the HidroSIG software, as explained in section 3.2.4. The watershed for the Rhine and Russian rivers were determined at the point of the furthest wellfield downstream, i.e. Staad PW I and Mirabel, respectively.

The morphology type displayed by all the reaches in Europe and the US, where there are functioning RBF sites, corresponds to Pool-Riffle (Table 3-4). This is the same morphology for most of the stream-reaches in Colombia, with the exception of the Sumapaz River, where El Limonar station is located, the Cravo Sur River, where Pte Yopal station is found, and the Nechi River at San Juan gauging station. These reaches turned out to be classified as plane-bed.

Table 3-4 Reach morphology type according to the classification tree of Flores et al. (2006)

Location River RBF Site Reach Slope

(m/m) SA0.4

Morphology Type

Colombia

Apartado Apartado 0.0026 0.0160 Pool-Riffle

La Vieja El Alambrado 0.0011 0.0215 Pool-Riffle

Guejar* El Limon 0.0013 0.0124 Pool-Riffle

Sumapaz El Limonar 0.0049 0.1107 Plane-bed

Cauca La Coquera 0.0002 0.0174 Pool-Riffle

Cauca La Virginia 0.0004 0.0202 Pool-Riffle

Magdalena Peñas Blancas 0.0003 0.0300 Pool-Riffle

Meta Pte Lleras 0.0002 0.0063 Pool-Riffle

Negro Pte Real 0.0008 0.0075 Pool-Riffle

Putumayo Pte Texas 0.0010 0.0266 Pool-Riffle

Cravo Sur* Pte Yopal 0.0091 0.1505 Plane-bed

Zulia Pto Leon 0.0003 0.0076 Pool-Riffle

Nechi San Juan 0.0029 0.1070 Plane-bed

Vichada Sta Rita 0.0001 0.0056 Pool-Riffle

Europe Enns KRB02 0.0017 0.0510 Pool-Riffle

Rhine Grind/Flehe/Staad 0.0002 0.0219 Pool-Riffle

USA

Great Miami Charles M. Bolton 0.0008 0.0302 Pool-Riffle

Missouri Nearman 0.0002 0.0464 Pool-Riffle

Ohio B.E. Payne 0.0001 0.0111 Pool-Riffle

Platte Ashland 0.0003 0.0457 Pool-Riffle

Raccoon Maffitt 0.0006 0.0237 Pool-Riffle

Russian Mirabel/Wohler 0.0008 0.0196 Pool-Riffle * Reaches visually classified as braided

It is important to notice that two reaches that can be visually categorized as braided, El Limon and Pte Yopal, were classified as pool-riffle and plane-bed, respectively, proving to be one of the limitation of the methodology, as mentioned by Jimenez (2015). Also, although the analyzed reach of the Enns river at the RBF site KRB02 was classified as pool-riffle according to the Flores et al. (2006) criteria (S0 ≤ 0.025 and S0A0.4 ≤ 0.055), this could

Evaluation of the Potential for Riverbank Filtration in Colombia

36

be arguable due to the fact that the product S0A0.4 in this case is 0.051, just a difference of 4 x 10-3 with the plane-bed category, which could be attributed to uncertainties induced by the methods used to calculate the slope in the first place. Consequently, the reach could also be interpreted as a plane-bed channel and, therefore, it will be considered as both in the analysis of hydraulic geometry.

Montgomery and Buffington (1997) report that pool-riffle, as well as plane-bed channels, show a combination of both, supply- and transport-limited characteristics, depending on the bed sediments mobility thresholds, i.e. the degree of bed-surface armoring. Thus, armored channels represent persevering supply-limited conditions, whereas unarmored channels indicate a balance between transport capacity and sediment supply. However, the authors report a general trend towards transport-limited conditions.

The main difference between these two types of morphology is that plane-bed channels lack the characteristic rhythmic bedforms present in pool-riffle channels (bars, pools), and that pool-riffle channels are usually unconfined, whereas plane-bed channels may be either unconfined or confined by valley walls. Also, plane-bed reaches occur at moderate to high slopes in relatively straight channels, while pool-riffle reaches happen at moderate to low gradients and have well-established flood-plains. Lastly, the typical substrate in pool-riffle streams is gravel sized, and in plane-bed channels is gravel to cobble. (Montgomery and Buffington, 1997)

Finally, Figure 3-16 presents a box plot of the slope (in m/m) for the stream-reaches classified as pool-riffle, for the data sets of Colombia (textured boxes) and the United States and Europe together (blank boxes). The graph shows that, for the selected reaches in Colombia, the median of the values is slightly lower than the median for the slope reaches in Europe and the US, but the interquartile values for both data sets are very similar with reach slope ranging in average from 2.5 x 10-4 to 8.5 x 10-3. The maximum value of slope in Colombia was 0.0026 and in Europe and the US together it was 0.0017. The minimum value of slope in both data sets was around 0.0001.

Figure 3-16 Box plots of the slopes for the pool-riffle reaches in Colombia (hashed boxes), and Europe and the United States (blank boxes)

Chapter 3 – Potential Riverbank Filtration Sites in Colombia

37

3.3.5 Stream-Reach Features

The results of the stream-reach processing using the AutoCAD macros and MatLAB routines are shown in Table 3-5. Appendix B presents the resulting diagrams for the estimation of mean curvature radius based on maximum and minimum local variances.

Table 3-5 Results of the stream-reach processing through AutoCAD macros and MatLAB routines

Location RBF Site W (m) Sinuous Rm (m) Amp (m) Lb (m)

Colombia

Apartado 34.6 1.36 234.53 111.50 361.15

El Alambrado 57.1 1.20 469.59 187.00 866.46

El Limon* 152.7 1.07 1,440.78 -9999 -9999

El Limonar 81.5 1.20 984.75 253.35 1,296.37

La Coquera 323.1 1.28 1,669.02 1,232.32 3,465.07

La Virginia 140.5 1.43 2,310.77 880.36 3,273.35

Peñas Blancas 1,413.8 1.20 10,228.02 1,762.01 9,953.90

Pte Lleras 346.4 2.24 931.96 1,588.64 1,762.98

Pte Real 16.6 1.36 123.93 63.97 221.85

Pte Texas 282.7 1.58 1,273.03 963.33 1,822.85

Pte Yopal* 289.1 1.09 3,220.82 -9999 -9999

Pto Leon 175.0 3.09 825.12 801.57 797.78

San Juan 148.2 2.06 1,311.90 1,388.82 2,493.16

Sta Rita 641.5 1.75 1,637.42 1,346.66 3,186.38

Europe Enns 141.5 1.28 717.39 557.86 1,746.96

Rhine 349.1 1.71 2,039.81 2,606.83 4,717.31

USA

Great Miami 128.6 1.56 672.4 636.75 1,148.31

Missouri 259.5 1.26 2,736.7 -9999 -9999

Ohio 760.4 1.06 12,127.2 -9999 -9999

Platte 567.9 1.09 5,415.6 -9999 -9999

Raccoon 107.2 1.85 400.2 212.05 917.14

Russian 47.3 1.62 589.2 688.58 1,512.21 * Reaches visually classified as braided

The calculated width values, W, for the Colombian reaches vary from 16.6 m at Pte Real station on the Negro River to 1.4 km for the Magdalena River at Peñas Blancas station. The European rivers do not exceed 350 m in width, whereas the value of W for the stream-reaches in the US ranges between 47 m for the reach of the Russian River located at the Mirabel and Wohler wellfields, and 760 m for the Ohio River reach in Louisville, KY, at B.E. Payne.

Sinuosity, or the ratio between the stream-reach length along its axis and the distance in a straight line between the two extreme points of the reach, is presented in the numerical column number two. The lowest value of sinuosity for the entire data set is 1.1, displayed by two reaches in Colombia and another two in the US, whereas the maximum is 3.1 for the reach of the Zulia River at Pto Leon station, followed by 2.2 and 2.1 at Pte Lleras (Meta River) and San Juan (Nechi River), respectively. For the data set corresponding to Europe and the USA, the highest sinuosity is presented by the Raccoon River at Maffit wellfield reach (1.8), followed by that of the Rhine River at Düsseldorf (1.7).

Evaluation of the Potential for Riverbank Filtration in Colombia

38

The obtained median curvature radii, Rm, are also presented in Table 3-5 (Appendix B contains the resulting graphs of median curvature radius and associated local variance diagrams for all the reaches). Just as for W, the highest value for the Colombian sites is for the reach of the Magdalena River at Peñas Blancas (10.2 km), and for the sites in the US the highest is that of the Ohio River (12.1 km). The lowest value of the whole data set is for the Negro River at Pte Real with 123.9 km. The high value of Rm at Peñas Blancas could be explained due to the fact that this reach visually presents characteristics of both, braided channel in some parts, but meandering overall (Figure 3-12).

Finally, for the entire data set the values of amplitude, Amp, range between 64 m and 2.6 m, whereas those for bend wave length, Lb, vary from 221.85 m and 9.95 km. It is important to notice that for the Guejar River at El Limon, Cravo Sur River at Pte Yopal, and the Ohio, Platte and Missouri river-reaches, the Matlab routine could not find a value for Amp and Lb and that is the reason of the -9999 values that appear in the table. For the first four cases, this is explained by the fact that they are also the reaches with the lowest values of sinuosity, meaning that they are almost straight channels. The Missouri river-reach, on the other hand, presents meanders (see Figure 3-15), and the inability of the routine to find a value for both Amp and Lb could be related to the reach length not being long enough for the calculations.

In order to morphologically compare the reaches in Colombia with those in Europe and the US, a series of hydraulic geometry relationships were assessed. In the first place, all the parameters in Table 3-5 were plotted against the drainage area (see Figure 3-17), which has been suggested as a surrogate variable for formative discharge (Jimenez, 2015). The blue-filled triangles represent the stream-reaches in Colombia and the orange-filled circles represent those in Europe and the US. The two reaches in Colombia visually classified as braided (El Limon and Pte Yopal) are shown as blue triangles inside a red circle, whereas the two channels classified as plane-bed (El Limonar and San Juan) are presented as blue triangles inside a black square. The orange-filled circle inside a black square corresponds to the Enns river-reach which, as explained before, could have also been classified as plane-bed. The power trend line corresponds only to the data from the reaches classified as pool-riffle.

Although there is a clearly-defined trend between the watershed area (A) and the planform features W, Rm, Amp and Lb, the correlation is strong only for the last case where (R2 = 0.8344). It is not possible to determine a unique behavior for the stream-reaches in Europe and the US, as they can be found to either side of the trend lines. Braided channels, on the other side, do appear to have a distinctive tendency way above the power trend lines, as can be seen in Figure 3-17 (a) and (b); however, there is not enough data to support this hypothesis. Reaches with plane-bed morphology behave similarly to the pool-riffle channels, since they are also transport-limited stream-reaches.

Chapter 3 – Potential Riverbank Filtration Sites in Colombia

39

(a) (b)

(c) (d)

Figure 3-17 Average width (a), median curvature radius (b), bend amplitude (c), and bend wavelength (d), related to watershed area

The fact that is not appropriate to define a unique power law relationship to drainage area coincides with the findings of Jimenez (2015) who, like other authors (Leopold and Wolman, 1960; Williams, 1986; Nicoll and Hickin, 2010), propose the use of bankfull width for making dimensionless morphological comparisons. Therefore, the next step was to compare the planform features of mean curvature radius, Rm, amplitude, Amp, and bend wavelength, Lb, to bankfull width, W (see Figure 3-18 a, b and c).

In all cases, the relationship with the bankfull width for the pool-riffle reaches is a strong one (R2 > 0.75), especially for the curvature radius (R2 = 0.8061) (Figure 3-18 a). However, based only on the scatter diagrams, it is not possible to distinguish between the stream-reaches in Europe and the US, of those in Colombia.

When looking at the boxplots of the ratios Rm/W, Amp/W and Lb/W (Figure 3-18 d, e and f), and despite the fact of the highly different data ranges and interquartile values, it is possible to see that the median values for the data sets in Colombia (textured boxes) and Europe and the US (blank boxes) are similar, with a median in Colombia of 6 for Rm/W, 3.6 for Amp/W, and 8.7 for Lb/W; for the US and Europe the median values were, in the same order, 7.7, 5, and 12.3.

Evaluation of the Potential for Riverbank Filtration in Colombia

40

(a) (b) (c)

(d) (e) (f)

Figure 3-18 Curvature radius (a), amplitude (b) and bend length (c), related to bankfull width, and the box plots correspondent to the ratios Rm/W (d), Amp/W (e) and Lb/W (f)

Additionally, the planform features Rm, Amp and Lb were compared with each other, resulting in the graphs shown in Figure 3-19 a, b and c. In general terms, a strong relationship was found among those parameters, being the highest the one between the median curvature radius and the bend wavelength, with an R2 equal to 0.9195. For cases (a) and (b), the RBF sites in Europe and the US seem to be distributed along the lower side of the trend lines, whereas for case (c) no differentiation can be made.

The boxes and whiskers graphs presented in Figure 3-19 d, e and f, show the comparison between the data set values for the selected reaches in Colombia (textured boxes) and those RBF sites in Europe and the US (blank boxes). For all the cases, the data ranges differ a lot between the compared sets of values. For cases (d) and (e) this is also true for the interquartile and the median values, especially for the relation Rm/Lb. On the other hand, the relation Amp/Lb (Figure 3-19, f) displays a very similar interquartile ratio for both data sets ranging between 0.3 and 0.5, and a combined median around 0.4.

From the previous comparisons, a graph of median curvature radius, Rm, and bankfull width, W, was made showing only the pool-riffle stream-reaches (Figure 3-20). It was found that the power trend line is separating reaches according to their Lb/W ratio, which can be associated to the confinement degree of the channels, according to Jimenez (2015) and Nicoll and Hickin (2010). In this particular case, the reaches with an Lb/W ratio higher than 13.3, located above the trend line (pink-filled triangles), can be considered as free meandering reaches according to the findings of Jimenez (2015) who established a threshold of 13.23 for transport-limited reaches with free meandering. On the other hand, those reaches with a ratio equal or lower than 13.3 are found below the trend line, represented as green markers. The three orange-filled circles represent the three reaches in the US for which the Matlab routine was unable to find a bend wavelength value. The

Chapter 3 – Potential Riverbank Filtration Sites in Colombia

41

green marker above the trend line and furthest to the right represents the reach of the Magdalena River at Peñas Blancas station which, as explained before, has the highest value of Rm probably due to its characteristics of both braided and meandering channel.

(a) (b) (c)

(d) (e) (f)

Figure 3-19 Scatter graphs of Rm to Lb (a), Rm to Amp (b) and Amp to Lb (c), and the correspondent box plots

Figure 3-20 Relation between curvature radius, Rm, and bankfull width, W, and degree of confinement expressed in terms of the ratio between bend wavelength and bankfull width, Lb/W

Finally, the bankfull width, W, and the median curvature radius, Rm, were compared against the bankfull discharge, Q, (Figure 3-21). For the Colombian reaches, the maximum discharge for a 2.33 year-recurrence period was selected as bankfull discharge, while for the rivers in the US, it was used the value of bankfull discharge presented by Caldwell (2006) even though it is not clear how such value was found. For the two streams in Europe (Enns and Rhine rivers) Caldwell does not present a bankfull discharge so the

Evaluation of the Potential for Riverbank Filtration in Colombia

42

average discharge was used instead. Again, the power trend line is for those values correspondent to pool-riffle reaches only.

In both cases a strong relationship to bankfull discharge was found, being especially high for the Rm/Q ratio, with a correlation coefficient of 0.85. The box plots displayed in Figure 3-21 (c) and (d) show that the Colombia and the EU+US data sets are very similar, not only on the value ranges (maximums and minimums) but also on the interquartile values, and the median values. Thus, the W/Q ratio in Colombia shows an interquartile that ranges from 0.15 to 0.31 with a median of 0.27, and in the US and Europe combined the interquartile goes from 0.09 to 0.21, with a median of 0.15. Lastly, for the Rm/Q ratio the interquartile values for the Colombian reaches vary from 0.81 to 1.78, with a median of 1.42, and for the reaches in the United States and Europe the interquartile is between 0.92 and 1.44, with a median value of 1.26.

(a) (b)

(c) (d)

Figure 3-21 Bankfull width (a) and median curvature radius (b) compared with bankfull discharge, and their correspondent box plots (c) and (d)

3.3.6 Stream Power as an Indicator of the River Self-Cleaning Capacity

The results of the estimation of total stream power (Ω) and specific stream power (ω) are shown in Table 3-6, along with the discharge used for the calculations as explained in section 3.2.7.

The maximum value of Ω for the reaches in Colombia was 43,660 Wm-1 at Pte Yopal (Cravo Sur River), followed by 35,890 Wm-1 at San Juan (Nechi River) and 20,770 Wm-1 at Pte Texas (Putumayo River). On the other hand, the maximum values for ω in Colombian

Chapter 3 – Potential Riverbank Filtration Sites in Colombia

43

reaches were for San Juan (242 Wm-2), El Limonar at Sumapaz River (238 Wm-2), and Pte Yopal (151 Wm-2). The specific stream power for the rest of the stream-reaches did not exceed 75 Wm-2, with the lowest values being for Sta Rita (Vichada River) with 3 Wm-2, and Pte Lleras (Meta River) with 5.4 Wm-2. Thus, the reaches morphologically classified as plane-bed were the ones with highest values of stream power.

Table 3-6 Total stream power (Ω) and specific stream power (ω) for all the analyzed reaches

Location RBF Site Q(m3/s) Ω(Wm-1

) ω(Wm-2

) Newson et al. (1998)

Colombia

Apartado 48.7 1,230.5 35.5 Erosion

El Alambrado 313.7 3,382.3 59.2 Erosion

El Limon+ 226.1 2,864.8 18.8 Equilibrium

El Limonar* 405.1 19,461.8 238.8 Erosion

La Coquera 2,956.8 7,203.3 22.3 Both

La Virginia 1,403.0 5,221.0 37.2 Erosion

Peñas Blancas 5,622.3 17,371.9 12.3 Equilibrium

Pte Lleras 1,124.5 1,869.4 5.4 Deposition

Pte Real 41.0 308.8 18.6 Equilibrium

Pte Texas 2,122.3 20,768.3 73.5 Erosion

Pte Yopal+* 488.0 43,659.1 151.0 Erosion

Pto Leon 612.2 1,765.0 10.1 Equilibrium

San Juan* 1,262.1 35,893.9 242.2 Erosion

Sta Rita 2,049.2 1,931.6 3.0 Deposition

Europe Enns 160.0 2,659.0 18.8 Equilibrium

Rhine 2,100.0 3,850.9 11.0 Equilibrium

USA

Great Miami 900.0 6,746.2 52.4 Erosion

Missouri 5,100.0 8,862.7 35.2 Erosion

Ohio 8,500.0 6,579.5 8.7 Equilibrium

Platte 3,700.0 12,038.5 21.2 Equilibrium

Raccoon 300.0 1,821.3 17.0 Equilibrium

Russian 500.0 3,731.9 78.9 Erosion * Reaches morphologically classified as plane-bed + Reaches visually classified as braided

In the US and Europe the reach with a highest total stream power is the Platte River (12,040 Wm-1), followed by the Missouri River (8,860 Wm-1), whereas the river-reach with the highest specific stream power is the Great Miami (52.4 Wm-2). The lowest value of ω, on the other hand, is 8.7 Wm-2 at the B.E. Payne wellfield in Louisville, KY.

Bizzi and Lerner (2013) demonstrated that there is a significant relationship between Ω and ω and distinctive features of erosion a deposition. According to their work, a total stream power of 1,648 Wm-1 and a specific stream power of 34Wm-2 are the threshold for the minimum energy necessary to trigger erosion processes, mobilize sediments, and to activate bank erosion and lateral channel migration. The authors also suggest that ω is an indicator of a “stability” threshold, whereas Ω is one for channel patterns transition and sediment budget analysis. Previously, Newson et al. (1998) considered specific stream power as the most important predictive variable in classification of channel types and found that channels with ω values lower than 7.5 Wm-2 showed a tendency to

Evaluation of the Potential for Riverbank Filtration in Colombia

44

depositional processes, whereas those with ω higher than 35 Wm-2 reflected signs of instability, i.e. erosion.

Considering the findings of Bizzi and Lerner (2013) on the threshold for total stream power (Ω), most of the stream-reaches analyzed would have the energy to start erosion processes. Furthermore, when looking at the values of specific stream power (ω), and taking into account the thresholds suggested by Bizzi and Lerner (2013) and Newson et al. (1998), it would be possible to define three main groups: those reaches with ω lower than 7.5 Wm-2 where deposition would be dominant, the ones with ω higher than 34 Wm-2

where the leading process is erosion, and the rest in between where no prevailing process is expected.

Finally, the data set for the reaches in Colombia classified as pool-riffle was compared to the data set for Europe and the US together, in terms of total suspended solids (TSS), bankfull discharge (QBF), total stream power (Ω) and specific stream power (ω), (Figure 3-22).

(a) (b)

(c) (d)

Figure 3-22 Box plots of the total suspended solids (a), discharge (b), total stream power (c) and specific stream power (d) for the data sets of the pool-riffle reaches in Colombia, and Europe and

the US combined.

From the upper part of Figure 3-22, which shows the distributions for TSS and QBF for Colombia (textured boxes) and the US and Europe (blank boxes), it is clear that Colombian rivers carry a much higher concentration of suspended sediments than the rivers in Europe and the United States but, at the same time, the bankfull discharges are

Chapter 3 – Potential Riverbank Filtration Sites in Colombia

45

significantly lower. This could suppose a higher potential to mechanical clogging and a limitation in the implementation of the RBF technology in Colombia, since one of the main causes for such clogging is the concentration of suspended sediments in the river water. However, when looking at the lower part of Figure 3-22, which presents the box plots of Ω and ω, it can be seen that the distributions for both sets of data (Colombia and EU+US) are very similar, which indicates that the Colombian rivers possess the require energy to clean themselves from clogging via erosion of banks and beds.

It must be taking into account that bankfull stage was calculated assuming a 2.33-years recurrence period, which it is not necessary true for Colombian rivers taking into account that higher floods can be triggered by “La Niña” phase of the ENSO. That is why Ω is also slightly lower in Colombian rivers than in those from abroad, and that is the reason why ω would be more appropriate in order to make comparisons.

3.4 CONCLUSIONS AND RECOMMENDATIONS

The potential of having successful and functioning riverbank filtrations sites in Colombia, just as the ones currently working in some places in Europe and the United States, was evaluated from the point of view of downstream hydraulic geometry and the self-cleaning capacity of the rivers to counteract for possible mechanical clogging due to the high concentrations of suspended sediments in Colombian rivers. The main idea behind this goal was to find elements, easily measurable through the use of readily available geo-processing tools, AutoCAD and Excel macros, Matlab routines, and information facile to obtain, such as river discharge and sediment load, that could help not only find similarities among those sites where RBF is currently the water treatment technology being used, but to act as a filter in the selection of future RBF sites in prefeasibility studies (especially in places with scarce information) before any on-site measurements have to be made. In order to do this, a set of 14 stream-reaches in Colombia was compared to stream-reaches in Europe and the US in the vicinity of RBF existing wellfields, following the methodologies previously proposed by other authors.

The first step was to visually compare the location of the current RBF sites in the US and Europe with the proposed location of potential RBF sites in Colombia within the river longitudinal profile. It was found that both, the existing and potential sites are located in sections of the profile with low gradient, which is where transport or sedimentation are the dominant processes according to the typical profile model. However, as it was expected, this simple classification did not allow making any differentiations between the analyzed stream-reaches.

The slope of the digitalized reaches was then calculated, along with the drainage area, in order to morphologically classify them according to their index of specific stream power. It was found that all the reaches in Europe and the US fell into the category of pool-riffle channels where transport-limited conditions prevail. This type of channels are usually unconfined by valley walls, present characteristic rhythmic bed forms such as bars and pools, have well-established channel migration zones, and the substrate is usually formed

Evaluation of the Potential for Riverbank Filtration in Colombia

46

by gravel sized sediments. Most of the stream-reaches in Colombia were also classified as pool-riffle, with the exception of three that were classified as plane-bed.

Even though this classification scheme is very attractive because of its facility of use, it has some limitations that need to be considered, such as the fact that it was defined for single-channel reaches, which means that braided channels cannot be included in such analysis. Another limitation is that it only goes so far as to separate reaches according to their planforms but it does not take into account the bedforms that can be present inside the channels.

The main axis and margins of the reaches digitalized in Google Earth were the input to an AutoCAD macro, which estimated length of the reach, bankfull width, and sinuosity. The width and length were then used with a Matlab routine to calculate the median curvature radius, the bend wavelength and the amplitude. The planform features thus calculated were compared through the use of hydraulic geometry relationships (mainly scatter graphs and box plots).

As it had been stated by other authors, it was found that a better power law relationship was obtained when comparing the planform parameters to bankfull width than to drainage area, since the former allows a dimensionless morphological comparison. Because all the reaches where RBF sites exist were classified as pool-riffle, the analysis of the results of hydraulic geometry focused especially on those streams. The strongest relationship was found when the curvature radius was compared to the bend wavelength, which could mean that such ratio is an important parameter to be considered when evaluating a potential RBF site. Other strong relationships found were curvature radius vs. bankfull discharge, bend wavelength vs. drainage area, and curvature radius vs. channel width.

From the results of the hydraulic geometry analysis can be concluded that important parameters to consider when evaluating the potential of a site for riverbank filtration are bankfull width, curvature radius, bend wavelength and bankfull discharge. However, in order to improve the results presented here, it is necessary to expand the information with more data, especially in places where there are currently RBF sites that have proven to be efficient. Overall, sites in Europe and US shared quantitative fluvial geomorphological characteristics and similarities with Colombian streams which could be interpreted as a positive result from the point of view of the suitability of those local sites for riverbank filtration

The last parameter evaluated here was the self-cleaning capacity of the rivers as to counteract possible mechanical clogging. This was done by estimating the total stream power and the specific stream power of the reaches. In the first place, the fact that the concentration of total suspended solids in Colombian rivers is a lot higher than in rivers in Europe and the United States, but the discharges are lower, could implicate higher potential to develop mechanical clogging, which might be a limitation of the technique if it progresses to the point where the connection between the river and the aquifer is lost. However, the results obtained here showed that the stream power, and further specific

Chapter 3 – Potential Riverbank Filtration Sites in Colombia

47

stream power, for the Colombian rivers is very similar to those in Europe and the United States, which means that they possess the necessary energy to start erosion processes of the bed and banks, thus removing the clogged layer. Despite the limitations associate to using stream power instead of shear stress, the former can be considered a good indicator of the self-cleaning potential of the river (scouring) as a counteraction for clogging, especially at the first stages of evaluation of a possible site for RBF, when on-site measurements are not possible.

As an extra measurement of the self-cleaning potential of the rivers, it would be recommendable to study recurrent floods in terms of magnitude, duration, and intensity, not only at those places where RBF is being used but also in locations where the technique is considered a possibility. Such analysis was beyond the objectives of this work but it is encouraged because scouring can also be a limitation of the technique since the constant removal of the clogging layer can represent a reduction in the efficiency of the treatment.

According to the results obtained during this work, it can be concluded that despite the limitations exposed above, Colombian rivers have the potential to be suitable to implement riverbank filtration sites and that the high concentration of suspended solids does not necessarily represent a limitation to the technique.

3.5 ACKNOWLEDGMENTS

The author wishes to express her immense gratitude to Dr. Mario Jimenez from Gotta Ingenieria, in Medellin, Colombia, for his constant collaboration during the work that led to the elaboration of this particular chapter.

49

4 RIVERBANK FILTRATION AND WATER QUALITY AT TWO SITES IN ANTIOQUIA

4.1 INTRODUCTION

Riverbank filtration (RBF) has proven to be an effective technology in the pretreatment of surface water for more than a century in Europe, and half a century in the United States, replacing or reducing the number of processes commonly used for this task, at the point that sometimes disinfection is the only treatment needed before consumption (Kuhen and Mueller, 2000). These two authors analyzed the steps needed to make raw water from the Rhine River in Germany apt for human consumption and found that with bank filtrate water, coagulation/flocculation, sedimentation and filtration could be avoided. Weiss et al. (2003) also reported that, with a 99.9% removal of turbidity, coagulation/flocculation, sedimentation and physical filtration would not be needed.

The final process in the treatment of drinking water is disinfection, whose sole objective is the removal, deactivation or killing of pathogenic microorganisms. A number of authors have documented the efficiency of RBF in the removal of this type of organisms, or their surrogates, with a reduction of over 99% with respect to the river water (Gollnitz et al., 1997; Schubert, 2002b and 2002c; Tufenkji et al., 2002; Wang, 2002; Weiss et al., 2002; Gollnitz et al., 2003; Gollnitz et al., 2005).

The chemistry of the river water not only determines to a great extent the quality of bank filtrate (Kuehn and Muller, 2000; Tufenkji et al., 2002) but has also been used to study the surface-groundwater interaction employing environmental tracers such as stable isotopes of water like deuterium and Oxygen-18 (Aseltyne et al., 2006; Baskaran et al., 2009), chloride (Bourg and Bertin, 1993; Bertin and Bourg, 1994; Trettin et al., 1999; Aseltyne et al., 2006; Cox et al., 2007), Radon-222 (Hoehn et al., 1992; Bertin and Bourg, 1994; Baskaran et al., 2009), bromide (Constantz et al., 2003), and temperature (Constantz, 1998; Constantz et al., 2003; Constantz et al., 2006), to name a few.

The use of conservative natural tracers to determine the portion of bank filtrate in a production well has also been reported in the literature. Wang (2002), for example, states that the ideal method for determining the amount of dilution, i.e. physical mixing of the river and aquifer water, is to find a tracer that exists in one source at a constant concentration but is absent in the other, and is conservative during the filtration process.

Bourg and Bertin (1993), on the other hand, used chloride to determine the mixing in a well field in Bordeux (France), where an aquifer rich in Cl- was recharged by water from the Lot River, which had a poor concentration of the anion. In their work, an expression is proposed to calculate the fraction of water originating directly from the river (x), by comparing the Cl- concentrations in each well (Cwell) with the measured values in the River

Evaluation of the Potential for Riverbank Filtration in Colombia

50

(Criv) and in the alluvial aquifer prior to operation of the well field (Caq), as shown in Eq. 4-1 and Eq. 4-2.

𝐶𝑤𝑒𝑙𝑙 = 𝑥𝐶𝑟𝑖𝑣 + (1 − 𝑥)𝐶𝑎𝑞 Eq. 4-1

𝑥 = (𝐶𝑤𝑒𝑙𝑙 − 𝐶𝑎𝑞)/(𝐶𝑟𝑖𝑣 − 𝐶𝑎𝑞) Eq. 4-2

Even though this equation was developed for chloride, the authors suggest that it can be used with any given parameter, as long as the substance is a conservative tracer so the only process involved is the physical mixing of the river and aquifer water.

The portion of river-borne water in bank filtrate can also be estimated through numerical modeling, according to the methods proposed by Chen (2001) and Schön (2006). The first method is purely geometric and assumes that the totality of water extracted in the well corresponds to particles coming from a complete circule around it, i.e. capture zone defined using backward particle tracking, and the angle between the particles that are coming from the river that are furthest apart from each other is measured and divided by 360o, thus:

𝑄𝑟𝑖𝑣𝑒𝑟/𝑄𝑤𝑒𝑙𝑙 = 𝑚𝑒𝑎𝑠𝑢𝑟𝑒𝑑 𝑎𝑛𝑔𝑙𝑒/360𝑜 Eq. 4-3

The second method, on the other hand, is based on stoichiometry. In MODFLOW (McDonald and Harbauch, 1988), a condition called Point Source is applied to the RIVER-cells and a known concentration of an imaginary tracer is set. Meanwhile, the well is represented as an observation well, and the portion of river water in the bank filtrate is calculated as the ratio between the calculated concentration of the tracer in the observation well (Cwell) and the concentration assigned to the RIVER-cells (Criv), thus:

𝑄𝑟𝑖𝑣𝑒𝑟/𝑄𝑤𝑒𝑙𝑙 = 𝐶𝑤𝑒𝑙𝑙/𝐶𝑟𝑖𝑣𝑒𝑟 Eq. 4-4

In order to evaluate the efficiency of RBF in terms of water quality for human consumption, two sites in Antioquia were selected and sampling campaigns of water from the rivers and the wells were carried out during the second half of the year 2010. The collected water was tested at certified laboratories for different quality parameters according to the current Colombian legislation for drinking water (República de Colombia, 2007).

The sampling campaigns were done as part of the project “Comprehensive Management of Joint Use of Surface Water and Groundwater” (Gestión Integral de la Utilización Conjunta de Aguas Superficiales y Subterráneas), co-funded by the Administrative Department of Science, Technology and Innovation (COLCIENCIAS) and the National University of Colombia (UNAL by its acronym in Spanish). Included in this project, a Master’s thesis in Engineering-Hydraulic Resources was developed with the objective of implementing numerical models of groundwater flow to represent the aquifer-river interaction at the two formerly mentioned sites, using data from previous, general,

Chapter 4 – Riverbank Filtration and Water Quality at Two Sites in Antioquia

51

groundwater studies (UNAL – CORNARE, 2000; UNAL – CORANTIOQUIA, 2004) and information collected on-site.

Escobar (2011) found, through the use of numerical modeling of groundwater flow, that the water extracted from the wells both sites was indeed a mixture of surface water and groundwater, with proportions and travel times that differed with the rate and duration of the pumping, the distance from the river to the well, and the governing climatological conditions. He also estimated the portion of river-borne water extracted in the wells at the sites using the methods of Chen (2001) and Schön (2006), described above.

Since the direct interaction between the rivers and the wells had been established using indirect methods, the main purpose behind the water sampling was to establish the actual degree of improvement of the quality of the bank filtrate in comparison with the river water. Consequently, the main objective of this chapter is to present the results and interpretation of the hydrochemistry laboratory analysis, the comparison of the water quality from the rivers and wells with the maximum admisible values according to the Colombian legislation for drinking water standards, and to determine the portion of river-borne water in the bank filtrate using chloride as a conservative tracer following the method proposed by Bourg and Bertin (1993).

4.2 METHODS

4.2.1 Site Selection

Two sites were chosen from an inventory of wells located in floodplains of selected rivers, with special attention to those wells that were suspected to be drawing water from the shallow aquifer. The location of these sites is shown in Figure 4-1. Site number one corresponds to a well in Santa Fe de Antioquia, west of the Department, adjacent to the Cauca River, and site number two is a well in Guarne next to La Mosca creek, in the region known as Eastern Antioquia (Oriente Antioqueño).

Figure 4-1 Localization of the two field sites: Santa Fe de Antioquia (1) and Guarne (2)

Evaluation of the Potential for Riverbank Filtration in Colombia

52

4.2.2 Water Sampling and Analysis

Between June and October of the year 2010, water samples were taken from the two sites, both from the rivers and the extraction wells (5 campaigns in Guarne and 8 in Santa Fe de Antioquia), with time intervals of approximately 20 days between campaigns, as shown in Table 4-1.

The water, which was never filtered at the site but taken as it was to the laboratories, was collected using two kinds of containers depending on the parameters that were going to be analyzed: a 2-liter plastic container for physico-chemical parameters, and a 250-mm glass bottle, autoclaved at 121oC, for the microbiological. The vessels were provided by the laboratories where the water was analyzed. Additionally, on-site measurements were conducted to determine water levels, temperature, electric conductivity, pH, and sometimes dissolved oxygen, depending on the availability of the equipment.

The water samples were tested in certified laboratories in Colombia for different quality parameters for drinking water, based on the Colombian legislation (Resolution No. 2115 of June 22nd 2007, Republica de Colombia). The samples from Santa Fe de Antioquia were analyzed at the laboratory of CORANTIOQUIA (Regional Autonomous Corporation of the Center of Antioquia) and those from Guarne at the laboratory of CORNARE (Regional Autonomous Corporation of the watersheds of the rivers Negro and Nare). Those are the two environmental agencies in the Department of Antioquia that had jurisdiction over the regions where the wells were located. The analytical methods each laboratory employed are shown in Appendix C.

Table 4-1 Water sampling dates at both studied sites

Date Santa Fe

de Antioquia Guarne

2-Jun-10 X

23-Jun-10 X

7-Jul-10 X

28-Jul-10 X

11-Aug-10 X

8-Sep-10 X

14-Sep-10 X

22-Sep-10 X

28-Sep-10 X

6-Oct-10 X

13-Oct-10 X

20-Oct-10 X

27-Oct-10 X

4.2.3 Quality of the Bank Filtrate vs. the River Water

The first step was to compare the results of the water quality analysis with the maximum accepted values (MAV) for human consumption according to the Colombian legislation. Resolution No. 2115 of 22nd June 2007 divides the characteristics to be considered for water quality for human consumption in three types: physical, chemical, and

Chapter 4 – Riverbank Filtration and Water Quality at Two Sites in Antioquia

53

microbiological. At the same time, the second classification is subsequently divided into chemical characteristics with known adverse effects in human health, chemical characteristics with implications in human health, and chemical characteristics with major economic consequences and indirect effects on human health. Table 4-2 presents the MAV of the parameters measured at either one or the two sites, according to the Colombian legislation (Res. 2115, 2007), and other variables that have no MAV but were also analyzed due to their importance in the determination of water quality.

Table 4-2 MAV for physical, chemical and microbiological characteristics of drinking water, and other measured variables

Characteristics Parameter Expressed as MAV

Physical

Color Platinum-cobalt units (PCU)

15

Turbidity Nephelometric turbidity units (NTU)

2

Conductivity Microsimens per centimeter (μS/cm)

1000

pH pH units 6.5 – 9.0

Chemical with known adverse effects in human health

Cyanide mg/l of CN- 0.05

Mercury mg/l of Hg 0.001

Chemical with implications in human health

Total Organic Carbon mg/l of TOC 5

Nitrites mg/l of NO2- - N 0.1

Nitrates mg/l of NO3- - N 10

Chemical with major economic consequences and indirect effects on human health

Calcium mg/l of Ca 60

Total Alkalinity mg/l of CaCO3 20

Chloride mg/l of Cl- 250

Total Hardness mg/l of CaCO3 300

Total Iron mg/l of Fe 0.3

Magnesium mg/l of Mg 36

Sulphate mg/l of SO42-

250

Phosphates mg/l of PO43-

0.5

Microbiological

Total Coliforms and E.coli Colony-forming unit (CFU) per 100 ml

0

Total Coliforms and E.coli Most probable number (MPN) per 100 ml

0

Other measured parameters not contained in the legislation but considered important for water quality

Bicarbonate mg/l of HCO3- -

Potassium mg/l of K -

Silica mg/l of Si -

Sodium mg/l of Na -

Total Dissolved Solids (TDS) mg/l -

Total Suspended Solids (TSS) mg/l -

Dissolved Organic Carbon (DOC)

mg/l -

Biochemical Oxygen Demand (BOD)

mg/l O2 -

Chemical Oxygen Demand (COD)

mg/l O2 -

Evaluation of the Potential for Riverbank Filtration in Colombia

54

4.2.4 Hydroclimatological Information

With the purpose of further understanding possible trends in the concentrations of the analyzed parameters, monthly hydroclimatological information was gathered from different gauging stations belonging to the IDEAM (Institute of Hydrology, Meteorology and Environmental Studies). Table 4-3 presents the information on the stations that were used in each site, including the parameters that are measured, the years of registry and its localization.

Table 4-3 Gauging stations nearby the two studied sites

Site ID Station Name Parameters Years Lat (N) Lon (W) Elev

Santa Fe de Antioquia

26225030 Cotove Hda Temperature (

oC)

Precipitation (mm) 1976 - 2013 06.53 75.83 530

26217050 Canafisto Discharge (m3/s) 1978 - 2010 06.42 75.82 534

Guarne 23087670 Riotex Discharge (m

3/s) 1994 - 2013 06.18 75.37 2017

23080640 Marinilla Precipitation (mm) 1974 - 2013 06.17 75.33 2028

4.2.5 Hydrochemistry

The water chemistry of the samples, both from the river and the well, was determined in terms of major ions (Ca2+, Na+, K+, Mg2+ and HCO3

-, Cl-, NO3-, SO4

2-) by plotting their concentration in meq/l in the diagrams of Piper and Stiff. The free software Diagrammes version 6.1, developed at the Laboratory of Hydrogeology of the University of Avignon in France (http://www.lha.univ-avignon.fr/LHA-Logiciels.htm) was employed for such purpose.

4.2.6 Portion of River-Borne Water in the Bank Filtrate

With the objective of validating the results obtained by Escobar (2011), the percentage of river water extracted from the wells was estimated using chloride as a natural tracer, as previously accomplished by Bourg and Bertin (1993).

Because the campaigns did not involve the sampling of other wells besides the ones mentioned at the beginning of this section, the Cl- concentration in the aquifer was obtained from previous works in the area of Santa Fe de Antioquia (UNAL – CORANTIOQUIA, 2004) and the Saint Nicholas Valley (UNAL – CORNARE ,2000), as it will be properly explained later for each particular site.

4.3 SANTA FE DE ANTIOQUIA

4.3.1 Site Description

4.3.1.1 Generalities

Santa Fe de Antioquia (from now on Santa Fe) is one of the main touristic areas in the Department of Antioquia. It is located in the western part of the Department, and presents an average elevation of 550 m.a.s.l. (Figure 4-2). Santa Fe has a semi-arid climate,

Chapter 4 – Riverbank Filtration and Water Quality at Two Sites in Antioquia

55

with an average temperature of 27oC, an average precipitation of 967 mm/year, and a real evapotranspiration between 547 and 827 mm/year (UNAL – CORANTIOQUIA, 2004).

The main river in the area is the Cauca River, which is the second most important fluvial artery in the country. Other main streams on the western margin of the Cauca River include the Tonusco River, the Seca Creek, and the Juanes Creek (Figure 4-2). The rest of the surface currents in the region are short creeks of small drainage areas or intermittent streams that only transport water during rainy events. Among those intermittent currents is the Cañaveral Creek, next to which the SantaFe well is located.

Figure 4-2 DEM (left) and map of geology and aquifer potential (right) in the area of Santa Fe

In the study by UNAL – CORANTIOQUIA (2004), the aquifer where the SantaFe well is located is described as an isolated alluvial terrace of the Cauca River, 10 to 12 m above the river, with high hydrogeological potential, and a thickness between 15 and 20 m (Figure 4-2 and Figure 4-4). To the west, the aquifer limits with igneous rocks of the Barroso Formation and the Sabanalarga Batholith, and to the east it is controlled by the Cauca River (Figure 4-2, right). In general, the area is composed of typical alluvial deposits, with intercalations of gravel, sand and silt (UNAL – CORANTIOQUIA, 2004).

The selected well is a dug well with a diameter of 90 cm and a depth of 18 m located on a small farm (Figure 4-3). This well had a submerge pump that was turned on every morning for about 6 hours at an approximate rate of 3.1 l/s, and its water was mainly used for household chores and for irrigation of the pastures and the few crops that grow there. The people of the farm used to put a tin sheet on top of the well to prevent contamination from outside or animals from going inside.

Evaluation of the Potential for Riverbank Filtration in Colombia

56

Figure 4-3 Site photo of the SantaFe well at the beginning of the study

By the time of the field work the well was located 15 m from the Cauca River (Figure 4-4) However, the scouring of the river eventually washed away the well. In a field trip made eighteen (18) months after finishing the water sampling, the well was only about 2 m from the river, and at the time this document was being written, the well did not exist anymore (Figure 4-5).

Figure 4-4 Google earth view of the alluvial terrace and the SantaFe Well at the beginning of the work

Escobar (2011) found through modeling that a very direct interationship between the river and the aquifer exists even without pumping, as the river will give water to the aquifer when the river stage ascends; otherwise, the aquifer will be the one providing water to the river.

Chapter 4 – Riverbank Filtration and Water Quality at Two Sites in Antioquia

57

Figure 4-5 Google earth view of the location where the SantaFe Well as located before being washed off by the Cauca River

4.3.1.2 Hydroclimatology of the area

Information from two hydroclimatological stations was obtained from the IDEAM: Cotove Hda records temperature and precipitation, whereas Cañafisto gauges discharge. The average multiannual values (blue bars) and the monthly values for the year 2010 (red bars) are shown in Figure 4-6. Those months that do not have red bars in the graphs of temperature and discharge lacked such information for the year 2010.

A bimodal cycle is readily apparent from the average multiannual graphs (blue bars) with two peaks in precipitation in April-May-June and September-October-November. The behavior of temperature is almost the opposite as precipitation, with the highest values recorded during the months of January, February and March. Discharge, presents a similar trend to that of precipitation but the second peak is during the months of October, November and December, i.e. a month lag with respect to precipitation.

From the graphs it is also clear that the monthly values of precipitation and discharge in the year 2010 were lower than the average for the first five months of the year and higher than the average for the remaining months, whereas the contrary occurs with temperature.

According to the Climate Prediction Center at NOAA, the year 2010 registered the transition from the end of a moderate El Niño (2009 – 2010) and the beginning of a strong La Niña conditions (2010 – 2011), with an Oceanic Niño Index (ONI) that reached a maximum of -1.5 from the period of September-October-November till the end of the year (www.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/ensoyears.shtml). In Santa Fe de Antioquia, this shift happened sometime around June, when the highest precipitations were recorded, as shown in Figure 4-6 (top left), with their associated discharge registered within a month-lag (Figure 4-6, bottom).

Evaluation of the Potential for Riverbank Filtration in Colombia

58

Figure 4-6 Average monthly values of precipitation, temperature, and discharge in the area of

Santa Fe

4.3.2 Results and Discussion

4.3.2.1 Quality of the bank filtrate vs. the river water

The graphs shown below present the concentration of a given parameter for each sampling campaign. The blue line represents the water in the river, the red line corresponds to the water extracted from the well, and the green line with no markers is the MAV for drinking water, for those cases where it applies. Additionally, for displaying purposes, the concentrations reported by the lab as lower than certain number, i.e. the detection limit, were graphed with the value of such limit. For color, turbidity, and total coliforms the lab also reported values higher than 150 PCU, 1000 NTU, and 1.6 x 108 NMP/100 ml, respectively, which are the highest values shown in the graphs. The complete table with the results of the water analysis is presented in Appendix D.

Physical Characteristics

The graphs with the results of the analysis of the physical characteristics listed in the legislation are shown in Figure 4-7. There is a notorious improvement in the water quality of the water in the well with respect to true color and turbidity, both parameters being way above the MAV in the river water but overall lower than such value in the water from the well, even sometimes lower than the detection limit (5 PCU for color). Electrical conductivity and pH comply with the MAV both in the river and the well. The increase in EC, together with other parameters that showed the same tendency, during the second half of the sampling campaigns, will be explained in the following section.

Chapter 4 – Riverbank Filtration and Water Quality at Two Sites in Antioquia

59

Figure 4-7 Physical characteristics in Santa Fe samples and their MAV for drinking water

Chemical Characteristics

The only two parameters belonging to the category of chemical characteristics with adverse effects on human health were total cyanide and mercury (Figure 4-8). These two compounds were not included in the lab analysis at the beginning of the sampling campaigns, but later on the people at CORANTIOQUIA considered them important enough as to include them, so they were taken into consideration in the last two sampling campaigns (October 6th and 10th 2010).

In the case of cyanide, the MAV is only exceeded in the river water collected on the 6th of October, whereas the water from the well complies with the norm. On the other hand, the concentration of mercury in the river water is both times above the MAV of 0.001 mg/l, but below the detection limit (0.001 mg/l) in the well water. The river water was indeed expected to present mercury due to the gold mining activities that take place in different locations within the Cauca River watershed, which use both mercury and cyanide in the process. These results show the potential that RBF might have in the removal of Hg and CN- even for lower concentrations in the river water.

Evaluation of the Potential for Riverbank Filtration in Colombia

60

Figure 4-8 Chemical characteristics with known adverse effects on human health in Santa Fe

samples and their MAV for drinking water

Among the category of chemical characteristics with implications in human health, the Colombian legislation lists total organic carbon (TOC), nitrites (NO2

-) and nitrates (NO3-).

These three compounds were measured in all the samples and the results are shown in Figure 4-9, along with their respective MAV.

Figure 4-9 Chemical characteristics with implications in human health in Santa Fe samples and

their MAV for drinking water

As it can be seen, the only parameter that exceeds its MAV is TOC in the river water, but the water from the well complies with the legislation. Nitrites and nitrates are below their MAV both in the water from the river and the well. In the case of nitrites, the concentration in the well water was almost always below the detection limit of 0.003 mg/l NO2-N), and nitrates were below their detection limit (1.5 mg/l NO3-N) for most of the samples both from the well and the river. Since the presence of nitrogen compounds is

Chapter 4 – Riverbank Filtration and Water Quality at Two Sites in Antioquia

61

usually associated with agricultural activities, it makes sense that their concentrations are low in the water from the well because the land there is not really used for such purposes. As for the river water, the Cauca is such a big river that natural mixing and dilution could easily decrease the concentrations of those compounds, especially if the discharge sites are far enough from the sampling site.

The parameters that represent the third group of chemical characteristics, i.e. those with major economic consequences and indirect effects on human health, are presented in Figure 4-10 and Figure 4-11. The two parameters in the water from the well that clearly exceed the MAV, at least in the last sampling campaigns, are calcium and hardness (Figure 4-10, top left and bottom right, respectively), with values above the MAV ranging from 70 to 107 mg/l for Ca, and 393 to 404 mg/l of CaCO3 for hardness. The total alkalinity in the well water (Figure 4-10, top right) only exceeds the MAV twice but with very small differences of 2 and 11 mg/l in the samples from the 20th of October and 22nd of September, respectively.

Figure 4-10 Chemical characteristics with major economic consequences and indirect effects on

human health in Santa Fe (Ca, Alkalinity, Cl-, Hardness) and their MAV for drinking water

The only one of these parameters whose values in the river water is higher than in the water from the well, is iron (Figure 4-11, top left), and it is also the only chemical characteristic in the river water with major economic consequences and indirect effects on human health that is above the MAV. In the case of Mg, the concentration in the well water is very close to the MAV in the samples form the 22nd of September and 20th of October. Even though the concentration of total phosphates is below the MAV, this one is most of the times lower in the river than in the well water, sometimes with values even below the detection limit of 0.153 mg/l.

Evaluation of the Potential for Riverbank Filtration in Colombia

62

Figure 4-11 Chemical characteristics with major economic consequences and indirect effect on human health in Santa Fe (Fe, Mg, SO4

2-, PO43-) and their MAV for drinking water

Microbiological Characteristics

The concentrations of Total Coliforms and E. coli as most probable number (MPN) per 100 ml are shown in Figure 4-12.

Figure 4-12 Microbiological characteristics (total coliforms and E.coli) in Santa Fe samples

Even though there is a substantial diminution in the concentration of total and fecal coliforms in the well water with respect to the river water, the values are still way above the MAV according to the Colombian legislation that places them at zero. The high concentrations of E.coli in the well water might be due to contamination from the sceptic tank of the farm, as this is a very poor house and no proper maintenance would be expected.

Chapter 4 – Riverbank Filtration and Water Quality at Two Sites in Antioquia

63

Other Compounds

Even though the Colombian legislation does not establish a maximum accepted value for the parameters presented in Figure 4-13, Figure 4-14 and Figure 4-15, they are considered important from the stand point of water quality as they represent other major ions such as HCO3

-, K and Na, occurrence of organic compounds (Biochemical Oxygen Demand -BOD-, Dissolved Organic Carbon -DOC- and Chemical Oxygen Demand -COD-), and presence of particulate matter (Total Dissolved Solids -TDS- and Total Suspended Solids -TSS-).

All the parameters shown in Figure 4-13 present higher concentrations in the well water than they do in the water from the river, being the greatest difference that of bicarbonates, which can reach in the well water concentrations as high as four times those in the river.

Figure 4-13 Concentrations of bicarbonates, potassium, silica and sodium in Santa Fe samples

As expected, the well water has a TDS concentration larger than the river water, but showing a similar trend in time (Figure 4-14). On the other hand, the same figure shows the significant improvement in the quality of the well water in terms of TSS as their concentration is around 100 times lower than in the river water, many times even below the detection limit (7 mg/l).

Evaluation of the Potential for Riverbank Filtration in Colombia

64

Figure 4-14 Total dissolved solids and total suspended solids in Santa Fe samples

Finally, Figure 4-15 shows that BOD in the river and well presents a similar trend, being overall slightly higher in the river. During the second half of the sampling campaigns the concentration of BOD were lower than the detection limit (2 mg/l O2), both in the river and in the well. Contrarily, COD in the river reaches concentrations of almost 180 mg/l, whereas in the well it does not exceed 20 mg/l. The DOC in the river shows a similar behavior to the COD graph but with a maximum value of 5.53 mg/l O2.

Figure 4-15 Biochemical oxygen demand, chemical oxygen demand and dissolved organic carbon in Santa Fe samples

4.3.2.2 Effects of climate and geology in water chemistry

Some of the graphs presented above display a marked increment in the concentration of the measured parameter in the well water, some times as high as a fourfold after the

Chapter 4 – Riverbank Filtration and Water Quality at Two Sites in Antioquia

65

sampling campaign of the 28th of July and lasting for the remaining time of the measurements. This is especially noticeable for ions of calcium, sodium, magnesium, chloride and sulphate, and in a less abrupt way for bicarbonate and potassium, all of which at the same time represent a rising of electrical conductivity, total dissolved solids, hardness and alkalinity (Figure 4-7, bottom left; Figure 4-10; Figure 4-11, top right and bottom left; Figure 4-13, top left and bottom right; and Figure 4-14, left).

When comparing the trends in concentrations of the aforementioned ions with the discharge of the Cauca River in 2010 (Figure 4-16, left), the increments in concentration are directly related to the discharges, implying that the water in the well came from the Cauca River. However, the fact that the concentrations of those parameters remained somehow constant in the river water is an indicator that such ions did not enter the aquifer when the Cauca river levels went up during the rainy months of La Niña, but had to come from elsewhere and by a different transport mechanism.

The relationship between the ions concentration (in mg/l) and the precipitation (in mm) during the year 2010 is also shown in Figure 4-16 (right). The graph suggests that the rising in concentrations is indeed related to the rainy events but with a month-lag. With that in mind, began the search for an explanation of the origins and transport mechanisms that caused the increase in concentrations of the ions in the river water.

Figure 4-16 Relationship of ions concentrations with river discharge (left), and precipitation (right)

in Santa Fe during 2010

When looking at the composition of the rocks located to the west of the alluvial terrace where the well is located (Figure 4-2, right), the origin of the ions can be explained. According to INGEOMINAS (2005), these are a set of volcanic rocks called Barroso Formation by Alvarez and Gonzalez (1978), predominantly basalts and diabases, whose main minerals are plagioclase, augite, amphibole and pyroxene, with chlorite, epidote and calcite as alteration minerals, and apatite and quartz as accessories.

To the west of the Barroso Formation appears the Sabanalarga Batholith, an intrusive body, primarily tonalites, gabbros and quartz-diorites, whose major constituents are plagioclase (intermediate to calcic), hornblende, pyroxene, biotite and accessory minerals such as zircon, apatite and sphene. Both igneous bodies are heavily affected by the Tonusco and Cauca Faults, facilitating the percolation of rain water and the chemical

Evaluation of the Potential for Riverbank Filtration in Colombia

66

weathering of the minerals. It is common to find calcite and quartz inside the fractures if these rocks.

Most of the minerals in these rocks are silicates with different contents of Ca, Na, Mg, Fe and Al (not measured in the water samples) and, therefore, their weathering and further dissolution could be accountable for the increments of calcium, sodium and magnesium in the well water. Furthermore, bicarbonate is produced during weathering of silicates (Appelo and Postma, 2005). Intercalated with the volcanic rocks of the Barroso Formation exist lenses of sediments of marine origin some of which are calcareous (INGEOMINAS, 2005), which could also explain the increase in bicarbonates in the well water.

On the other hand, an increment in iron concentration will not be expected because the iron that is present in silicate minerals, like biotite or hornblende, may form Fe-oxide as an insoluble weathering product (Appelo and Postma, 2005). The decrease in dissolved iron concentration in the well water throughout part of the second half of the sampling campaigns (Figure 4-11) is most probably showing that during that time the aquifer was fed by a source different than the Cauca River.

As for the chloride concentration in groundwater, its rising could be due to the accessory mineral apatite, a calcium phosphate mineral that can present high concentrations of either Cl- or F- ions, but the latter was not measured in the samples.

Finally, according to INGEOMINAS (2005) on the fractures of the basalts of the Barroso Formation and in the intercalated sedimentary are mining deposits of gypsum and pyrite, found to the west of the alluvial terrace. The weathering of both minerals could be the source of sulphate ions that also presented a rising in the well water samples collected during the second have of the campaigns.

Once the origin of the ions was established, the question remained as to how they got into the aquifer. One of the hypotheses was that they came from the mountains to the west via runoff and subsurface flow associated to the precipitation events. However, could not be solely responsible for an increase of the kind that was observed, mainly because seepage velocities are normally low and would not lead to such an immediate response in the concentrations of the ions in the well water. Therefore, another mechanism had to be missing from the equation.

As described in section 4.3.1.1, most of the secondary surface currents in the area of Santa Fe de Antioquia are intermittent, including the Cañaveral Creek, located approximately 70 m northwest from the well (Figure 4-2 and Figure 4-4 ). These stream flows over rocks from the Sabanalarga batholith and Barroso Formation for approximately 5 km, and receives the water of the La Yagual Creek which also runs over the same type of rocks. Because these two streams only carry water during precipitation events, it was not possible to take samples of the Cañaveral Creek as there was nobody trustworthy for this task at the site at the moments those events where occurring.

It was decided then to take samples of the sediments of the banks and bed of the creek and had them optically analyze them for their composition. This job was conducted by a Geology student who prepared the samples at the Laboratory of Sample Preparation at

Chapter 4 – Riverbank Filtration and Water Quality at Two Sites in Antioquia

67

the Faculty of Mines, National University in Medellin. First, he made a sieve analysis in order to determine the size distribution, roundness and selection of the particles. Then he used an electronic binocular magnifier with light directed to examine the gravel-size grains and a petrographic microscope for some thin sections prepared at the same laboratory. The thin sections were analyzed at the Petrography Laboratory at the National University of Colombia. The finer sediments were going to be seen with a scanning electron microscope but at the end there was no funding available for such a task.

Minerals such as quartz, plagioclase, feldspar, pyroxene, muscovite, biotite, and glauconite were found both in the samples of the bed and banks of the Cañaveral Creek, along with some rock fragments, particularly plutonic and volcanic, metamorphic, and sedimentary. Furthermore, based on the sieve analysis and the roundness and sphericity of the grains, it was concluded that they were transported a very long way, by a high energy stream with sufficient capacity as to carry such a variety of materials and sizes. The entire report is presented as Appendix E (in Spanish).

These findings strongly suggest that the enrichment of ions in the well water is related to the discharges in the Cañaveral Creek due to the above-average precipitation events registered during the second half of the sampling campaigns, implying that during those months the main recharge of the aquifer came from this stream.

Back to the Cauca River, the BOD graph (Figure 4-15) shows a marked increment in the samples taken during the second and third campaigns (June 23rd and July 7th) just to return to almost its initial values in the subsequent samples, a similar behavior to that of nitrates (Figure 4-9). Conversely, the COD and DOC graphs present an increment in the river water after the 7th of July to a peak on the 8th of September for COD and the 22nd of September for DOC.

The opposite behavior between the BOD and the DOC graphs, and the concordance of the last one with the COD, is an indication that, even though there was an increment in the dissolved phase of organic matter in the river water, most of it represented non-biodegradable organic compounds (very low BOD/COD ratio), probably pesticides or industry sub-products that during the rainy season were washed off into the Cauca River, whose watershed is inhabited by almost 19 million people (around 41% of the total Colombian population).

4.3.2.3 Hydrochemistry

The concentration in meq/l of the four major cations (Ca2+, Na+, K+, Mg2+) and anions (HCO3

-, Cl-, NO3- and SO4

2-) was plotted using the Stiff and Piper diagrams in order to determine the type of water both in the river and the well.

The piper diagram in Figure 4-17 reveals that in terms of cations, the water extracted from the well is mainly calcic, whereas the water from the river is more magnesic. On the other hand, according to the anions, both the river and well waters are bicarbonate-rich but the latter migrates towards the sulphate apex, showing an enhancement in the concentration of such ion as explained in the previous section.

Evaluation of the Potential for Riverbank Filtration in Colombia

68

Figure 4-17 Piper diagram of major ions in Santa Fe samples

Furthermore, the Stiff diagrams in Figure 4-18 allow visualizing the changes in concentration in major ions during the time of the sampling. Thus, it can be seen that the chemistry of the river water (light blue) remains very constant through time, whereas the water from the well (light red) clearly shows the abrupt increase in concentrations of calcium, magnesium, bicarbonate and sulphate in the second half of the sampling campaigns, aspect that was explained in section 4.3.2.2.

Figure 4-18 Stiff diagrams of water from the river (light blue) and the well (dark blue) in Santa Fe

Chapter 4 – Riverbank Filtration and Water Quality at Two Sites in Antioquia

69

4.3.2.4 Portion of river water in the bank filtrate

The portion of river-borne water extracted in the well was estimated using the concentrations of chloride measured during the sampling campaigns, following the procedure by Bourg and Bertin (1993), as explained in section 4.2.6.

Because the sampling did not include any well different from the one being studied, the background concentration of chloride in the aquifer was obtained from a regional study conducted by UNAL – CORANTIOQUIA (2004) on the evaluation of the aquifer potential of some areas western of the department of Antioquia, among which Santa Fe was included. They divided the area in sectors according to geological, geomorphological and hydrogeological characteristics, and gave the name of La Isla to the sector that comprises the alluvial deposits of the Cauca River with high aquifer potential, as shown in Figure 4-19, left), where the SantaFe well is located a little bit north to the southernmost well, A22. Part of their work consisted on characterizing the water quality of the alluvial aquifers, and the water from seven dug-wells was sampled in La Isla for such purpose. The sampling was done in May of 2004, which presented normal hydrological conditions according to NOAA’s Climate Prediction Center (www.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/ensoyears.shtml).

Before deciding which well, or wells, to use as background values for chloride, the concentration of major ions was plotted in a piper diagram (Figure 4-19, right). In terms of cations, only the water from A58 presents a dominant ion, in this case Na+K, whereas the rest of the wells have water richer in calcium and magnesium. The anion triangle, on the other hand, shows that all wells have bicarbonate water, but wells A25 and A26 are somehow less enriched in such anion and moving towards the limit of no dominant ion. Hence, the water from well A58 can be classified as Na-K-HCO3, whilst the rest of the wells have water of the type Ca-Mg-HCO3, which is in general terms the type of water found both in the Cauca River and the SantaFe well. Even though wells A25 and A26 present the same type of water than the rest, they have higher concentrations of sulphate and chloride, like some of the samples from the SantaFe well during the second half of the campaigns.

With the aim of having a representative chloride concentration of the aquifer, the wells closest to the river were eliminated, remaining then A25, A26, A58 and A22. However, the differences in the type of water of A58 with respect to the rest of the wells (including SantaFe), and the fact that this well presented suspiciously high concentrations of total dissolved and suspended solids, were reasons enough not to use it in this exercise. Well A22 was also discarded because it is located in the vicinity of the SantaFe well and the differences in chloride concentrations were very small, most probably indicating the influence of the river water. Finally, it was decided to use the average of the chloride concentrations of wells A25 (30.8 mg/l) and A26 (22.9 mg/l), as the background value for the aquifer, since they are both far enough from the river (> 300 m) to guaranty no mixing and chemically they have the same type of water.

Evaluation of the Potential for Riverbank Filtration in Colombia

70

Figure 4-19 La Isla Sector and dug-wells within it (left), and correspondent Piper diagrams (right)

The percentage of river-borne water in the SantaFe well was then estimated for each sampling date using Eq. 4-2, and the results are shown in Table 4-4.

Table 4-4 Portion of river-borne water in the bank filtrate in the SantaFe well

Date Cl

- (mg/l) % River-

borne water River Well avg A25-A26

02-Jun-10 4.13 6.33 26.85 90.3

23-Jun-10 3.73 5.62 26.85 91.8

07-Jul-10 3.92 6.78 26.85 87.5

28-Jul-10 3.85 5.78 26.85 91.6

08-Sep-10 3.91 12.80 26.85 61.2

22-Sep-10 4.52 19.20 26.85 34.3

06-Oct-10 3.49 14.60 26.85 52.4

20-Oct-10 5.00 18.20 26.85 39.6

The results show that during the first half of the sampling campaigns, when El Niño conditions prevailed, there was, in average, an input of river water of 90%. However, when rainy events started and La Niña conditions were the norm, the portion of water from the Cauca River in the well decrease significantly by 30 to 60%.

The relationship between the percentage of river-borne water in the well and the discharge of the Cauca River for the year 2010 is shown in Figure 4-20 (left). There, it is clear that about a month after the discharge rate of the river increased, the portion of river water in the aquifer decreased, a similar behavior to that observed with the precipitation (Figure 4-20, right).

Chapter 4 – Riverbank Filtration and Water Quality at Two Sites in Antioquia

71

Figure 4-20 Percentage of river-borne water in the well compared to discharge of the river (left)

and precipitation (right) in Santa Fe for the year 2010

These results support the hypothesis that the enrichment in certain ions during the second half of the campaigns, as stated in numeral 4.3.2.2, was the product of recharge by another source different than the Cauca River, most probably from the Cañaveral Creek, and subsurface flow from the rocks to the west of the terrace, which experienced chemical weathering, and hence dissolution of minerals, due to the heavy precipitation events registered during La Niña conditions.

Using the geometrical method of Chen (2001), Escobar (2011) found that if the SantaFe well pumped continuously during 365 days, the maximum percentage of river-borne water would be around 43%, which means about half of the estimations made using chloride as a tracer. However, when the cited author used the method of Schön (2006), he determined that even without pumping the aquifer is recharged by the river in a proportion of 50 to 60%, number that increases by 20 to 30% if the pumping occurs.

The problem with Escobar’s model is that he did not consider the Cañaveral Creek and completly ignored its interaction with the aquifer. Even though he found a direct relationship between the water levels in the well and some reconstructed water levels in the river (Escobar, 2011), he assumed that the water entering the aquifer came solely from the Cauca, which has proven here to be correct only during months of no rain, but not true when precipitation events are prevailing.

4.3.2.5 Removal rates

As hypothesized in the previous sections, the main percentage of the recharge to the aquifer varied during the time of the sampling campaigns (from the Cauca River during the first half and from the Cañaveral Creek during the second), and because it was not possible to take samples of the Cañaveral Creek, then the removal rates with respect to such stream can not be established.

However, using the results from the first four campaigns, where there was an average of 90% of Cauca River water in the well, some parameters presented a removal rate over 90%. These were turbidity, E. coli, total coliforms, total suspended solids (TSS), volatile suspended solids (VSS), dissolved iron, and color. Figure 4-21 shows the log-removal for those parameters Figure 4-21 [% Removal = 100 – 10^(2-x), where x is the log removal].

Evaluation of the Potential for Riverbank Filtration in Colombia

72

Other parameters not included in the graph, but showed important removal rates were total phosphorous (85.8%), total solids (80.5%), nitrites (75.9%), and total COD (75.5%),

Figure 4-21 Average log-removal of turbidity, E.coli, total coliforms and TSS in Santa Fe

The findings reported here show that RBF from the Cauca River can be promising alternative in Santa Fe de Antioquia during dry months when surface water is scarce as most streams are dried, and the water for human consumption comes from water treatment plants that directly take the water from the river which increases the costs of treatment, especially of the removal of the turbidity.

4.4 GUARNE

4.4.1 Site Description

4.4.1.1 Generalities

The second site studied is located in the Oriente Antioqueño (Eastern Antioquia), in the region known as Valle de San Nicolás (Saint Nicholas valley), which is characterized for having one of the fastest growing populations in the department of Antioquia. The elevation in the study area ranges between 2086 and 2145 m.a.s.l. (Figure 4-22, left), the precipitation oscillates between 1800 and 2600 mm/year, the average temperature is 17oC, and the evapotranspiration is in average 825 mm/year (UNAL – CORNARE 2000). The main streams in the area are the Negro River, with its tributary creeks: La Marinilla, La Pereira and La Mosca (Figure 4-22).

The well is located on the aquifer associated to the fluvial deposits of La Mosca creek. Bellow these deposits, the quartz-diorite of the Antioquian Batholith, the main igneous body of the area, is found (Figure 4-22, right). The alluvial deposits of La Mosca Creek

Chapter 4 – Riverbank Filtration and Water Quality at Two Sites in Antioquia

73

were defined by UNAL – CORNARE (2000) as having a high hydrogeological potential, according to the regional study conducted in the area.

Figure 4-22 DEM (left) and map of Geology (right) of the Saint Nicholas Valley nearby La Mosca Creek

The well is located on the premises of the company Omya Andina S.A. (Figure 4-23), a producer of industrial minerals, mainly fillers and pigments derived from calcium carbonate and dolomite. It is a vertical well with a depth of 25 m, situated 30 m from the creek, properly built and maintained (Figure 4-24), and it pumps continuously at a rate of 1.2 l/s.

Figure 4-23 Google earth view of the Omya well near La Mosca Creek

Evaluation of the Potential for Riverbank Filtration in Colombia

74

At a local scale, the technical report by AGUA (2000), the company responsible for the drilling of the well, states that the place where the well was drilled is an unconfined aquifer with an intercalated sequence of sand, gravel and stones in a sandy clay matrix. At a depth of about 19 m the weathered igneous basement was found, and the fresh rock appeared at 25 m. The report also shows that the filter was located at a depth between 9 and 18 m below the surface. Finally, with the pumping test performed after the drilling, it was found that the aquifer has a hydraulic conductivity of about 2.3 x 10-5 m/s.

Escobar (2011) found that, just as it happens in Santa Fe, there is a direct interaction between La Mosca creek and the well and that, with or without pumping, an increment in the water levels in the creek represents an entry of river water into the aquifer, and viceversa.

Figure 4-24 Site photo of the Omya well

4.4.1.2 Hydroclimatology of the area

The values of precipitation and discharge were obtained from registries of two IDEAM hydroclimatological stations in the area of La Mosca creek: Marinilla and Riotex, respectively. The average multiannual values (blue bars) and the monthly values for the year 2010 (red bars) are shown in Figure 4-25.

Both parameters present a bimodal multiannual cycle with peaks in April-May-June and September-October-November. In general terms, precipitation during most months of the year 2010 was higher than the average, with values in September and October significantly higher than the mean in as much as a threefold. During those months, four of the five sampling campaigns in La Mosca well and creek were carried out. On the other hand, discharge for 2010 presents a one-month-shift to the right on the bimodal cycle, with values lower than the average during the first five months of the year, and then higher than the mean for the rest of the months. The trends in precipitation and discharge agree with the shift in hydrological conditions from a moderate El Niño to a strong La Niña

Chapter 4 – Riverbank Filtration and Water Quality at Two Sites in Antioquia

75

occurring towards the middle of year 2010 according to NOAA’s Climate Prediction Center, as explained in section 4.3.1.2.

Figure 4-25 Average monthly values of precipitation and discharge in Guarne

4.4.2 Results and Discussion

4.4.2.1 Quality of the bank filtrate vs. the river water

Physical Characteristics

The results of the analysis of the physical characteristics of the water from La Mosca creek and the Omya well, and the maximum admissible values according to the requirements of the Colombian legislation for drinking water are shown in Figure 4-26. The complete table with the results of the water analysis is presented in Appendix F.

Figure 4-26 Physical characteristics in Guarne samples and their MAV for drinking water

Evaluation of the Potential for Riverbank Filtration in Colombia

76

Just as what happens at Santa Fe, there is a notable improvement in color and turbidity (y axes in a logarithmic scale) in the water from the well with respect to the river, where both parameters are considerably above the MAV. Color even reaches values below the detection limit of 2.5 UPC for three of the water samples from the well. Electrical conductivity for both waters is safely below the MAV. On the other hand, pH of the water from the well is below the minimum accepted value in Colombia which is 6.5, and the river water only complies once with the requirements of the legislation.

Chemical Characteristics

In Guarne, unlike Santa Fe, none of the parameters listed as chemical characteristics with known adverse effect on human health were measured. As for the chemical characteristics with implications in human health, only the concentration of nitrites and nitrates together (NO2

- + NO3-) was analyzed at the laboratory of CORNARE (Figure 4-27).

The graph also shows on the right y axis the MAV value for nitrates (10 mg/l NO3-N) as a reference.

Figure 4-27 Chemical characteristics with implications in human health in Guarne samples and

their MAV for drinking water

It can be seen that the concentrations of these ions both in the well and river water is very low. These values (an average of 0.275 mg/l for the river water and 0.136 mg/l for the well) mean scarce agricultural activity in the basin of the creek up from the point of sampling and at the lot where the well is situated, which makes sense being that the location of the company Omya Andina, a plant for the processing of industrial minerals, and upstream the watershed is barley intervened by anthropogenic activities.

The parameters from the group of chemical characteristics with major economic consequences and indirect effect on human health are presented in Figure 4-28 and Figure 4-29. In general, all the parameters both in the water from the river and the well comply with the maximum admissible values established in the legislation.

The concentrations of calcium, magnesium, and hence alkalinity and hardness, are higher in the well than in the river. Chloride and sulphate, on the other hand, are significantly higher in the river water than in the well, but the values do not exceed the MAV.

Chapter 4 – Riverbank Filtration and Water Quality at Two Sites in Antioquia

77

Figure 4-28 Chemical characteristics with major economic consequences and indirect effects on human health in Guarne samples (Ca, Alkalinity, Cl-, Hardness) and their MAV for drinking water

Figure 4-29 Chemical characteristics with major economic consequences and indirect effects on human health in Guarne samples (Fe, Mn, Mg, SO4

2-) and their MAV for drinking water

Evaluation of the Potential for Riverbank Filtration in Colombia

78

Total iron and manganese were only analyzed in the last three samples (Figure 4-29) and both compounds have higher concentrations in the river water than in the well. In the case of iron, the difference between the river and the well can be as high as 35 times, and in all samples this parameter in the water from the creek was considerably above the MAV. It is important to remember that, in the case of iron the analysis included both the dissolved and particulate fraction (which also explains the very high concentrations of this ion in the stream water, > 6 mg/l), even though the legislation is only concerned with the former.

Microbiological Characteristics

The concentrations of Total Coliforms and E. coli, as colony forming units (CDU) per 100 ml, are shown in Figure 4-30. The concentration of both total coliforms and E. coli in the well water compared to the river water is significantly lower but still substantially higher than the MAV, which in Colombia is zero. This means that the water is not suitable for consumption as it is but it needs to be treated to remove the remaining pathogens.

Figure 4-30 Microbiological characteristics (total coliforms and E.coli) in Guarne samples

Other Compounds

Other major ions that were analyzed in the water samples were bicarbonate, potassium, sodium, total dissolved solids (TDS), total suspended solids (TSS), biological oxygen demand (BOD) and chemical oxygen demand (COD) (Figure 4-31 and Figure 4-32).

The concentrations of HCO3- in the well are a lot higher than in the river whereas sodium

is slightly higher in La Mosca creek than in the well, and potassium presents a more erratic behavior, but in general with concentrations in the well higher than in the river.

As in Santa Fe, the well water shows a larger concentration of TDS than the river water, but of similar trend in time, whereas the TSS are significantly higher in the river (Figure 4-32). As for BOD and COD, both parameters are below the detection limit in the samples of well water (1.43 mg/l O2 and 10.82 mg/l O2, respectively), while the river presents a higher BOD concentration in all the samples, meaning greater biological activity than in the aquifer. Only one sample of river water shows an obvious higher COD concentration with respect to the well water, but for the rest of the samples this parameter is below the detection limit.

Chapter 4 – Riverbank Filtration and Water Quality at Two Sites in Antioquia

79

Figure 4-31 Concentration of bicarbonates, potassium and sodium in Guarne samples

Figure 4-32 Total dissolved solids, total suspended solids, biochemical oxygen demand and chemical oxygen demand in Guarne samples

Evaluation of the Potential for Riverbank Filtration in Colombia

80

4.4.2.2 Effects of climate and geology in water chemistry

The explanation for the large increase in the concentration of calcium ions in the well water (Figure 4-28, top left), could be linked to the climatological conditions in the region during the time of the water sampling. When comparing the calcium concentrations in the well with the discharge of La Mosca Ckreek during the year 2010 (Figure 4-33, left) it can be seen that the initial trend is very similar, i.e. when the discharge rises, so does the ion concentration. However, for the last two water samples, even though the discharge of the stream was still rising, the content of calcium abruptly decreased. That, and the fact that the water from La Mosca had a pretty constant concentration of calcium, is an indicator that the ion had to come from somewhere else.

Figure 4-33 Relationship of calcium concentration with discharge (left) and precipitation (right) in

Guarne in 2010

On the other hand, when comparing the calcium content with the precipitation in the area (Figure 4-33, right), the relationship between both parameters became very clear. Thus, during the months of extremely high precipitation (between 700 and 800 mm) the calcium concentration in the well increased significantly (more than twice compared with the first sample), but when the heavy rain ceased and the values of precipitation came down, the concentration of calcium significantly decreased. It became clear that the increment in calcium was due to the rainy events, but the origin of the ions remained to be explained.

As explained in section 4.4.1.1, the well is situated in the grounds of a company, producer of industrial minerals derived mainly from limestone, and the remnants fragments are accumulated on the open field, nearby the well as shown in Figure 4-24. The dissolution with rain water of these sedimentary rocks, which are mainly composed of calcite (CaCO3), and further percolation of the dissolved ions, can explain the increase in calcium concentration during the months of high precipitation.

Finally, the higher concentrations of HCO3- and Mg in the well water compared to the river

water could also be related to the same phenomenon of dissolution of the limestone and dolomite deposits and the percolation into the saturated zone, the same as calcium.

Chapter 4 – Riverbank Filtration and Water Quality at Two Sites in Antioquia

81

4.4.2.3 Hydrochemistry

The type of river and well water was determined by plotting the concentrations in meq/l of the four major cations (Ca2+, Na+, K+, Mg2+) and anions (HCO3

-, Cl-, NO3- and SO4

2-) using the Stiff and Piper diagrams (Figure 4-34 and Figure 4-35).

Figure 4-34 Piper diagram of major ions in Guarne samples

Figure 4-35 Stiff diagrams of the water from the river (light blue) and the well (dark blue) in

Guarne samples

In terms of cations, the samples present a disperse behavior from Ca to Na+K with an intermediate content of Mg, while the anions seem to be much more stables, especially in the well water which is almost completely carbonate water. The river water, even though it is also bicarbonated, has a higher contribution of chloride and sulphate (Figure 4-34). What the Stiff diagrams are showing (Figure 4-35) is that the chemistry of the river water (light blue) is somehow constant throughout time with concentrations of major ions remaining always below 0.5 meq/l, whereas the well water is evidently richer in bicarbonate (around 1.5 meq/l) and has a distinct variable concentration of calcium in a range from below 0.5 to above 1.5 meq/l).

These variations in the amount of Ca, as explained in the previous section, are the result of local precipitation events and the consequent dissolution of the limestone rocks that are deposited near the well. The fact that the concentration of bicarbonate in the well water does not present the same variations as calcium could be indicating that the

Evaluation of the Potential for Riverbank Filtration in Colombia

82

background concentration of such ion is usually that much higher in the aquifer than in the creek and it is not as influence by rain as the aforementioned cation.

4.4.2.4 Portion of river water in the bank filtrate

The water sampling campaigns at Guarne, as at Santa Fe, did not include any other well different that the one being studied, therefore it was necessary to find a background value of chloride in the aquifer to be able to apply Eq. 4-2 proposed by Bourg and Bertin (1993). Unfortunately, unlike in Santa Fe, the major project on the characterization of the aquifers in the San Nicolas Valley (UNAL-CORNARE, 2000) did not include information on water quality, and no other studies of this type have been conducted in the area ever since, at least not for public access.

Taking into consideration the differences in concentration of chloride of one order of magnitude between La Mosca creek and the well, and the fact that chloride in the water from the well was very low (four of the samples were under 1 mg/l and three of them were below the detection limit of 0.68 mg/l), it was decided to apply the equation assuming that the concentration of chloride in the aquifer was negligible. Thus, the maximum possible percentage of river-borne water in the bank filtrate would be between 10 and 24% (Table 4-5).

Table 4-5 Portion of river-borne water in the bank filtrate in the Omya well

Date Cl

- (mg/l) % River-

borne water River Well 11-Aug-10 7.50 0.75 10.0

14-Sep-10 5.00 0.68 13.6

28-Sep-10 3.74 0.68 18.2

13-Oct-10 4.49 0.68 15.1

27-Oct-10 5.24 1.25 23.9

A very direct relationship between the percentage of river-borne water and the discharges of La Mosca during the time of the sampling is shown in Figure 4-36 (left), which goes in concordance with what Escobar (2011) found through numerical modeling about how an increment in water levels in the stream meant an entry of river water into the aquifer. The fact that the highest values of percentage of river-borne water are at the end of the months of September and October, also shows the direct river-aquifer interaction in correspondence with the hydroclimatology of the area, which in those months presented the highest precipitation values of the year and the second and third highest discharges of La Mosca creek (Figure 4-36, right).

The values obtained here are very similar to the results of Escobar (2011), who indirectly estimated the portion of river water from the results of numerical modeling. He found with the methods of Chen (2001) and Schön (2006) that, under continuous pumping as is the case for the Omya well, the river-borne water in the bank filtrate was about 20%.

Chapter 4 – Riverbank Filtration and Water Quality at Two Sites in Antioquia

83

Figure 4-36 Percentage of river-borne water in the well related to discharge (left) and precipitation

(right) in Guarne during 2010

4.4.2.5 Removal rates

In conclusion, the parameters that on average presented a removal over 95% in their concentration in the well water compared to the water of La Mosca creek were E. coli, color, total suspended solids, turbidity and iron (log-removal between 1.9 and 1.4), and those whose concentration was reduced between 75 and 85% (log-removal from 0.6 to 0.8) were sulphate, chloride, and total coliforms (Figure 4-37).

Figure 4-37 Average log-removal of some of the chemical parameters measured in Guarne

4.5 CONCLUSIONS AND RECOMMENDATIONS

As part of the project “Comprehensive Management of Joint Use of Surface Water and Groundwater”, co-funded by COLCIENCIAS and the Universidad Nacional of Colombia, two sites in Antioquia (Santa Fe de Antioquia and Guarne) were selected to test the efficiency of riverbank filtration in terms of water quality for human consumption. Previously, under the same project, the direct interaction between the streams and the wells at both sites

Evaluation of the Potential for Riverbank Filtration in Colombia

84

had been established through numerical modelling, and therefore the analysis of the water samples had as main purpose to establish the actual degree of improvement of the quality of water extracted from the well in comparison with the river water, and to determine the degree of mixing as the portion of river-borne water in the bank filtrate using chloride as a tracer.

Several water sampling campaigns, both from the rivers and the wells, were carried out during the second half of the year 2010, at time intervals of approximately 20 days, and the samples were analyzed at the laboratories of CORANTIOQUIA and CORNARE. Most of the parameters including in the tests corresponded to those that the Colombian legislation for drinking water consider important and have an established maximum accepted value. Hydroclimatological data (precipitation, discharge and temperature) was also available and it was use to understand the main trends in the concentrations of the analyzed parameters and the degree of interaction between the rivers and the wells.

When comparing the water quality of the bank filtrate with the Colombian legislation for drinking water, it was found that in Guarne, the only parameter that exceeded the MAV was the pathogens; although the reduction on average was of almost 99% in E.coli and 80% in total coliforms. Other parameters that in average achieved a reduction of over 90% were color, total suspended solids, turbidity, and iron.

On the other hand, in Santa Fe de Antioquia, parameters such as turbidity, alkalinity and magnesium slightly surpassed the norm once or twice during the time of the sampling, but in general the quality of the bank filtrate complied with the legislation, with the exception of calcium (and hence hardness), whose concentration significantly exceeded the MAV for the last four samples, and pathogens.

The marked increase in calcium concentration at Santa Fe de Antioquia, only in the well water, during the second half of the sampling campaigns was not an isolated event. The same behaviour was observed for other major ions such as sodium, magnesium, chloride and sulphate, and in a lower degree for bicarbonate and potassium, which at the same time represented an increase in electrical conductivity, total dissolved solids, hardness and alkalinity. Since the increments occurred during the rainy months, the results demonstrate the close relationship between local geology and climate in the water quality of the aquifer, and evidenced the influence of Cañaveral creek, an intermittent stream that flows in the vicinity of the well, and which only carries water during precipitation events, together with dissolved ions product of the weathering and dissolution of minerals of the rocks to the west of the alluvial plain. These results imply that the recharge to the aquifer during the months of heavy rains comes predominantly from the Cañaveral creek instead of the Cauca River.

Even though most of the parameters from Guarne presented a somehow constant behaviour in time, the fact that the concentrations of calcium, magnesium, bicarbonate, potassium, and hence alkalinity and hardness, are higher in the well than in the water from La Mosca creek are evidences of the impact that the dissolution during rainy events of the local deposits of limestone near the well have on the hydrochemistry of the aquifer.

Chapter 4 – Riverbank Filtration and Water Quality at Two Sites in Antioquia

85

Finally, using chloride as a tracer, the portion of river-borne water in the bank filtrate was estimated. Because the sampling campaigns did not involve any other well different from the ones being studied, the chloride concentration of the aquifer had to be acquired indirectly. In the case of Santa Fe de Antioquia, there was a regional study on the potential of the aquifers in the area which included the water quality analysis of some wells, so the chloride concentration in the aquifer could be established. Despite the fact that in Guarne the only major groundwater study in the San Nicolas Valley did not involved water quality, given the marked differences in the concentration of chloride between the creek and the well, and the fact that the values for this ion in the well water were in three out of five occasions lower than the detection limit, the maximum possible percentage of creek water in the well was determined by assuming that the chloride concentration in the aquifer was so small it could be neglected. Thus, the estimated portion of river-borne water at the well in Santa Fe de Antioquia coming from the Cauca River ranged from 34 to 91%, proving the hypothesis of the different origins of recharge at the site, whereas in Guarne was between 10 and 24%. The lower values of river-borne water for Guarne, compared to those for Santa Fe de Antioquia, could be related to the lower hydraulic conductivity at the former site (one order of magnitude), along with the larger distance from the river to the well (twice the distance), and the lower pumping rate (half). most likely due to its lower hydraulic conductivity and the larger distance from the river to the well. Because the Cañaveral creek in Santa Fe de Antioquia only carries water during precipitation events, the percentage of surface water coming from the stream and the removal rates could not be determined as there was no opportunity to sample its waters.

Given the differences in recharge sources at Santa Fe, and the lack of information on the water quality of the Cañaveral creek, only the higher removal rates from the first four months, when in average 90% of the water from the well came from the Cauca River, were determined. The parameters that in average showed a removal of over 99% were turbidity, E. coli, total coliforms and total suspended solids. Furthermore, there was a significant reduction of over 94% in volatile suspended solids, iron and color in the water from the well compared to the water from the creek.

The results presented here proof that riverbank filtration is efficient enough as to replace some of the treatment steps in conventional water treatment plants, such as coagulation/flocculation, sedimentation and physical filtration, being disinfection the only treatment needed before human consumption. As a result, RBF represents a very promising alternative in Santa Fe de Antioquia during dry months when surface water is scarce as most secondary streams are dried, difficulting the access of drinking water to the population.

However, there were some compounds that were not analyzed for, and need to be taken into consideration because their presence could imply additional treatment. Such is the case of ammonia, trihalomethanes, hydrocarbons, pesticides, heavy metals, water-soluble pharmaceuticals and, in general, any substance of importance that is suspected to be present in the river water. Furthermore, certain parameters should always be measured because their content may determine the need of further bank filtrate treatment before

Evaluation of the Potential for Riverbank Filtration in Colombia

86

disinfection. Among these are dissolved oxygen and total and/or dissolved carbon. The concentration of the former (to be measured on-site) in the water regulates the solubility of iron and manganese; whereas the latter, which represent the organic substances in the water, is important especially if chlorination is used as the disinfection method (as it is done in Colombia), because the chlorine can react with the organic matter to form disinfection by-products such as trihalomethanes, which are known to be harmful for human health. Temperature and pH should also been measured on-site since, together with dissolved oxygen, can determine the nature of biological processes or chemical reactions to be expected to happened during the travel of water from the river to the well.

During the time of the water sampling, a transition from moderate El Niño to strong La Niña conditions occurred and it was accompanied by a significantly increment in precipitation and, therefore, discharge of the rivers. This shift in hydroclimatological conditions clearly affected the water chemistry in Santa Fe de Antioquia, where not only the concentration of major ions increased significantly for the second half of the samples, but also the river water presented a rising of dissolved organic carbon and chemical oxygen demand of non-biodegradable compounds, probably pesticides or industry sub-products that during the rainy season were washed off into the Cauca River. The change in hydrological conditions was also evident in the portion of river-borne water in the well which decreased by 30 to 60% during the months of La Niña, when the recharge to the aquifer came mainly from the Cañaveral creek.

In a Country like Colombia, where the effects of climate phenomena such as El Niño and La Niña are experienced in a direct way, the consequences in the water quality of both rivers and aquifers could be a limitation in the implementation of RBF technique. However, depending on the degree of impact, it would be possible to take the necessary measures to minimize repercussions.

4.6 ACKNOWLEDGMENTS

Special thanks to CORANTIOQUIA and CORNARE, institutions that provided the services of water analysis at no cost, and to COLCIENCIAS that co-funded my studies during the time the necessary work for this chapter was carried out.

87

5 PESTICIDES REMOVAL DURING RBF

5.1 INTRODUCTION

Pesticides are organic chemical compounds intentionally released into the environment to control, eliminate or destroy pests in agricultural corps or livestock (Rashid et al., 2010). Despite the fact that pesticides were created to benefit human beings by improving agricultural production, they have been found to be toxic for living organisms, difficult to degrade and with persistent bioaccumulative effects (Ormad et al., 2008). Pesticides can reach surface water bodies by runoff, groundwater discharge, or atmospheric deposition, and their presence in groundwater can be attributed to either direct leaching from agricultural fields or recharge from rivers contaminated with them (Verstraeten et al., 2002a).

The fate and transport of pesticides in the subsurface depend on factors such as the physical and chemical properties of the compounds (solubility, size, polarization and volatility), hydrologic conditions, types of soil and sediments, and on physical, chemical, and biological processes in the soil and in the vadose and saturated zones (Verstraeten et al., 2002a; Rashid et al., 2010). Among those processes are advective transport, hydrodynamic dispersion, precipitation, sorption, volatilization and degradation by biotic and abiotic mechanisms, being sorption and biodegradation the two main dominant processes to determine the occurrence of pesticides in a water-sediment system (Son, 2010).

Pesticides can undergo chemical transformation and the respective sub-products are usually more soluble than their parent compounds and leach more rapidly (Scribner et al., 2000). Despite the fact that the transformation products tend to be less toxic (Sinclair and Boxall, 2003), they are expected to be more persistent (Verstraeten et al., 2002b), and hence represent a greater contamination threat to the environment (Baluch et al., 1993; Scribner et al., 2000).

Even though the removal of pesticides cannot be fully achieved by bank filtration, and additional treatment techniques may be needed to remove byproducts as well as parent compounds (Verstraeten and Heberer, 2002), the concentration of organic compounds in bank filtrate have been reported to be lower than in the river water (Schaffner et al., 1987; Ray et al., 1998; Verstraeten et al., 1999; Kuehn and Mueller, 2000), and the potential of the technique in the complete removal of specific pesticides has been investigated by Son (2010).

For more information on the presence of pesticides in the environment, on the adverse effects of these compounds on the ecosystem and human health, and on particular experiences with riverbank filtration in the removal of organic pollutants, the reader is

Evaluation of the Potential for Riverbank Filtration in Colombia

88

encouraged to refer to the comprehensive works of Buchanan et al. (2010), Rashid et al. (2010), and Verstraeten et al. (2002a), respectively.

Colombia is considered to be an agricultural country and such activity constitutes the most common industry, both for internal use and export. Due to poor agricultural practices, an excess of fertilizers is applied to degraded soils and, by run-off, those fertilizers end up in the surface water bodies (IAvH et al., 2011). Besides the fertilizers, agricultural practices also provide pesticides, among which there are substances with high levels of toxicity, carcinogens, and teratogens (IDEAM, 2010).

Unfortunately, it is not possible to know the real amount of agrochemicals that end up in the water bodies, since the monitoring network of water quality in Colombia includes only dissolved oxygen, total suspended solids, electric conductivity, pH, biological oxygen demand (BOD), chemical oxygen demand (COD) and, in some cases, total nitrogen and total phosphorous (IDEAM, 2010).

However, according to the Colombian Agricultural Institute (ICA by its Spanish acronym), the volume of pesticides used in the country in 2010 was about 11.9 millions of tons for their solid presentation and 32 millions of liters for their liquid formulation (ICA, 2011). Table 5-1 shows the distribution of such volumes per type of control.

Table 5-1 Volume of pesticides used in Colombia in 2010 discretized by type of control (ICA, 2011)

Type of control Kg Lt

Volume % Volume %

Fungicide 6,973,173 58.72 4,693,494 14.63

Herbicide 2,066,583 17.40 20,892,975 65.12

Insecticide 2,835,545 23.88 6,497,693 20.25

Total 11,875,301 100 32,084,162 100

In 2010 the most common fungicide used was Mancozeb in its solid form and Chlorothanolil in its liquid one; for herbicides, the most used was Diuron in its solid form and Glyphosate in its liquid equivalent; and the most common insecticide was Chlorpyrifos, both in its solid and liquid forms (ICA, 2011). Table 5-2 presents some of the most used pesticides in Colombia in 2010.

While in Colombia the pesticides shown in Table 5-2 are still currently in use, in the US and the European Union the use of Carbofuran, Methamidophos and Methyl Parathion has been prohibited, and the EU has also banned Ametryn, Atrazine, Propanil and Paraquat (SAN, 2011).

In order to determine the fate of agrochemicals during RBF five active ingredients from Table 5-2 were used to perform column tests at the Geohydraulics laboratory at HTW-Dresden, where the author was conducting her internship. The pesticides chosen for these experiments were Ametryn, Atrazine, Carbofuran, Diuron and Propanil. These compounds were selected based on their availability and the possibility of being analyzed at the Institute of Water Chemistry at TU-Dresden, which generously provided its service and support during the time of the experiments and results analysis.

Chapter 5 – Pesticides Removal During RBF

89

Table 5-2 Most used pesticides by active ingredient in Colombia in 2010 (ICA, 2011)

Type of Control Active ingredient Kg % Lt %

Fungicide

TOTAL 6,973,173 100 4,693,494 100

Chlorothanolil 1,584 0.02 1,058,470 22.55

Mancozeb 4,314,059 61.87 335,612 7.15

Herbicide

TOTAL 2,066,583 100 20,892,975 100

2,4-D 81,340 3.94 1,361,509 6.52

2,4-D amine - - 2,062,228 9.87

Ametryn 14,435 0.70 453,804 2.17

Atrazine 178,322 8.63 27,168 0.13

Diuron 430,493 20.83 150,134 0.72

Glyphosate 226,014 10.94 9,443,589 45.20

Paraquat - - 1,743,032 8.34

Pendimethalin 387,960 18.77 174,310 0.83

Pichloram + 2,4-D - - 2,191,100 10.49

Propanil 386,184 18.69 666,436 3.19

Insecticide

TOTAL 2,835,545 100 6,497,693 100

Buprofezine 358,274 12.64 - -

Carbofuran 291,176 10.27 362,219 5.57

Chlorpyrifos 1,364,646 48.13 2,368,213 36.45

Cypermethrin 81,344 2.87 278,350 4.28

Methamidophos - - 780,543 12.01

Methyl Parathion - - 155,659 2.40

With the outcomes of the column tests, the values for retardation factor, dispersion coefficient, and degradation coefficient for each pesticide were found using the software CXTFIT 2.0 (Toride et al., 1995). Subsequently, the results were used to feed a contaminant transport model from the Lößnitztal RBF site in Germany, where the sediments for the columns came from, using MODFLOW (McDonald and Harbaugh, 1988) and MT3DMS (Zehng and Wang, 1999). The idea behind it was to test the potential of RBF to remove these pesticides from the river water.

Even though it was not possible to work with sediments from Colombian rivers, the size of the sediments from the riverbed of the Lößnitztal River at the RBF site of the same name were similar enough to some places in Colombia, so that their use in the experiments was considered appropriate to represent Colombian conditions. As for the water used during the column tests, it came from the Elbe River in Dresden (Germany), despite the possible differences in the characteristics between the Elbe and Colombian river waters that could affect chemical or biological reactions. However, since the purpose of this study was to establish first approximations to the potential of RBF to be used in Colombia as a water treatment technology, those possible differences were not considered a limiting factor. Additionally, as it will be explained later, further research is needed in this area of pesticide removal and contaminant transport in order to determine with a higher degree of detail the necessary conditions and limitations of using RBF in the removal of pesticides at a given site.

Evaluation of the Potential for Riverbank Filtration in Colombia

90

5.2 COLUMN EXPERIMENTS

5.2.1 Laboratory Work

Soil columns experiments in the study of hydrogeological properties have been documented since the eighteenth century with the work of De la Hire (1703), although their use in the chemistry and movement of solutes in pore water started until two centuries later (Lewis and Sjöstrom, 2010).

The classification of soil columns can be done either by their level of saturation or according to the method of construction (Lewis and Sjöstrom, 2010). Thus, soils operating under unsaturated regime (commonly referred as lysimeters) are characterized by having both air and water in their pore spaces, simulating the vadose zone, whereas in columns running under saturated regime water occupies all the voids in order to reproduce the conditions found in an aquifer. On the other hand, packed columns are built by encasing and compacting disturbed soil in a rigid container, unlike monoliths which are extracted whole and intact from natural soil. Usually, packed columns are more homogeneous than monoliths, hence easier to reproduce but also less realistic.

5.2.1.1 Experiment setup

Two (2) stainless steel packed columns were run under saturated regime, each one with a diameter of 7.6 cm and a length of 50 cm. The sediments came from the riverbed of the river Lößnitztal in East Germany and the water was collected from the Elbe River at Dresden. The columns were filled with dry sediments, which were compacted with a metal pestle every 1 or 2 cm of soil had been added, and then saturated with water. The whole experiment was set up inside of a thermobox so the temperature could be controlled (Figure 5-1).

Figure 5-1 Experiment setup for the column tests

Two temperatures were used, 10oC and 20oC, since those are common water temperatures found in Colombia. The water was made to flow through the columns from top to bottom at a steady-state rate with the use of a membrane pump. A solution

Chapter 5 – Pesticides Removal During RBF

91

containing all five pesticides and dissolved with river water was then introduced into the soil column, and samples were taken at the outlet at different time intervals.

5.2.1.2 Soil characterization

A sieve analysis was performed to the material used to fill the columns, according to the standard DIN 18123 (Deutsches Institut fur Normung, 2011). The collected soil was first sieved to remove the grains with a diameter bigger than 2 mm to make sure the sediment sizes were smaller than sand-size and to be able to accomplish a more uniform packing of the column. A soil sample of 530.05 g was then flushed with water through the 0.063 mm sieve to remove the silt and clay present in the sediments. The gradation test was then performed to 517.56 g of sediments using the sieve sizes 1, 0.71, 0.5, 0.25, 0.125, and 0.063 mm.

In order to determine the type of soil, the method described by the Unified Soil Classification System ASTM D2847 (ASTM, 2006) was used. This standard method allows the classification of mineral and non-mineral soils based on the sieve analysis, i.e. the sediment size. The first step is to determine whether the soil is coarse-grained or fine-grained. If more than 50% of the sediments are retained on a No. 200 sieve (0.075 mm), then the soil is classified as coarse-grained, otherwise it is classified as fine-grained soil. The coarse-grained soils can be further divided into gravel or sand. Gravel is defined as particles that will pass a 75 mm sieve but will be retained on a 4.75 mm sieve, and sand are particles that pass the 4.75 mm sieve but cannot go through a 0.075 mm sieve.

The next step is to determine the cleanliness of the sample. In both cases, gravel and sand, are classified as clean if there is less than 5% of fine sediments (smaller than 0.075 mm) in the sample. Finally, it is necessary to determine if the sample is well-graded or poorly graded. In order to do this the coefficient of curvature (Cc) and the coefficient of uniformity (Cu) have to be calculated as shown in Eq. 5-1:

𝐶𝑐 =(𝑑30)2

𝑑10 × 𝑑60 𝐶𝑢 =

𝑑60

𝑑10 Eq. 5-1

where d60, d30, and d10 are the particle sizes corresponding to 60, 30, and 10% finer on the cumulative particle-size distribution curve, respectively. If Cu ≥ 6 and 1 ≤ Cc ≤ 3, then the sample is well-graded, otherwise the soil is poorly graded.

The amount of organic matter (TOC), and of iron and manganese were determined at the Institute of Water Chemistry of TU-Dresden, following the standards DIN EN 1593-C and DIN EN ISO 11885 E22, respectively (Deutsches Institut fur Normung, 2012 and 2009).

The results from the sieve analysis are shown in Table 5-3, Table 5-4, and Figure 5-2. According to the sieve analysis, the effective particle sizes d10, d30, and d60 are 0.23, 0.47, and 0.87 mm, respectively. Following the Unified Soil Classification system (ASTM, 2006), Cc = 1.11 and Cu = 3.84. Thus, and since the content of fines is less than 5%, the sample can be classified as poorly graded sand. The TOC of the soil was 0.29%, the iron content 1.56%, and the manganese content 0.0337%.

Evaluation of the Potential for Riverbank Filtration in Colombia

92

Table 5-3 Mass and percent cumulative fraction of sediments retained in the sieves

Sieve Size (mm) Mass Fraction (g) Cum. Fraction (%)

2 0 100.00

1 172.09 67.53

0.71 92.32 50.11

0.5 92.69 32.62

0.25 113.62 11.17

0.125 35.44 4.49

0.063 10.74 2.46

Rest 0.55 0.11

Sum Weight 517.45 -

Sieve Loss -0.11 0.02

Figure 5-2 Cumulative particle-size distribution curve from the sieve analysis

Table 5-4 Grain-size distribution

Size Term Size Range (mm) Percentage (%)

Very Coarse Sand 1 – 2 32.47

Coarse Sand 0.5 – 1 34.91

Medium Sand 0.25 – 0.5 21.44

Fine Sand 0.125 – 0.25 6.69

Very Fine Sand 0.063 – 0.125 2.03

Silt and Clay < 0.063 2.46

Total 100

5.2.1.3 Tracer tests

A conservative tracer, in this case table salt, was used during the column experiments to determine, based on the resulting breakthrough curves (BTCs), parameters such as porosity, volume of the pores (or pore volume), and residence time, among other things.

In total, five (5) tracer tests were performed in the columns at two different temperatures, 10 and 20oC. The first tracer test experiment (TT 1-1) was carried on before the pesticides experiment started mainly to check the BTC for any anomalies such as isolated peaks that could be due to problems in the packing or saturation (Lewis and Sjöstrom, 2010).

Chapter 5 – Pesticides Removal During RBF

93

The four remaining tracer tests, TT 2-2, TT 3-1, TT 4-1 and TT 5-2, were performed in association with pesticides experiments E 2-2, E 3-1, E 4-1 and E 4-2, respectively. The name of the tracer test experiment is given by the letters TT (Tracer Test) followed by the sequence number of the experiment (1 to 5), and ending in the column number (1 or 2). Thus, tracer test TT 3-1 was the third tracer test performed and it was run in column number 1. For the pesticides experiments the nomenclature is similar, only TT is changed by the E of experiment.

The main parameters of the tracer tests and the associated pesticide experiments, if applicable, are shown in Table 5-5, where d is the diameter of the columns, L is the length of the columns, A is the cross-sectional area of the columns, ∀ is the volume of the columns, T is the temperature at which each test was performed, C0 is the initial tracer concentration in water in the columns, Ctracer is the concentration of the tracer in the reservoir that fed the columns, t is the total length of time of the experiment, and Q is the average discharge at the bottom of the column.

Table 5-5 Parameters of the tracer tests

Parameter TT 1-1 TT 2-2 TT 3-1 TT 4-1 TT 5-2

d (cm) 7.6

L (cm) 50

A (cm2) 45.36

∀ (cm3) 2268.23

T (oC) 20 10 10 20 20

Associated pesticide experiment N/A E 2-2 E 3-1 E 4-1 E 4-2

C0 (S/cm) 564 360 350 450 450

Ctracer (S/cm) 702 635 871 759 759

t (min) 2585 12635 11755 8120 8120

Q (ml/min) 0.98 0.56 0.58 0.46 0.55

The concentration of the tracer in the outlet, C, was measured at different times, and the values of normalized concentration (C/C0) were then plotted against time (in minutes). The breakthrough curves (BTCs) resulted from the tracer tests are shown in Figure 5-3. The time at which C/C0 was equal to 0.5 (t50), i.e. when half of the solute has left the column, is shown next to each BTC. The parameters calculated for each tracer test are presented in Table 5-6.

Table 5-6 Parameters calculated from the tracer tests

Parameter TT 1-1 TT 2-2 TT 3-1 TT 4-1 TT 5-2

Temperature (oC) 20 10 10 20 20

Q (ml/min) 0.98 0.56 0.58 0.46 0.55

t50 (min) 793 1307.5 1372.5 1668 1498

Pore-water velocity (cm/min) 0.063 0.038 0.036 0.03 0.033

Darcy’s vel (cm/min) 0.022 0.012 0.013 0.010 0.012

n 0.34 0.32 0.35 0.34 0.36

Pore Volume (cm3) 781 727 797 760 822

The pore-water velocity was calculated as the length of the column, divided by t50, and the Darcy’s velocity as the average discharge divided by the cross-sectional area of the

Evaluation of the Potential for Riverbank Filtration in Colombia

94

column, A. The porosity, n, was then calculated dividing the Darcy’s velocity by the pore-water velocity. Finally, the pore volumes, PV, for each column were determined as the volume of the column times the porosity.

Figure 5-3 BTCs for the tracer tests

From the results of the tracer tests can be seen that the porosity of column 1 remains pretty much constant for all the experiments, independently of the temperature (0.34 at 20oC, and 0.35 at 10oC). However, for column 2 the change of porosity between tracer test TT2-2 and TT5-2 is more significant (0.32 at 10oC and 0.36 at 20oC). Taking into account

Chapter 5 – Pesticides Removal During RBF

95

that the porosity was calculated as the Darcy’s velocity divided by the linear velocity, and that for both columns the average discharge, Q, is almost the same (hence the same Darcy’s velocity), then the increase in porosity is due to the greater t50 in the latter experiment with respect to the former. One reason for the rising in porosity has to do with the rearrangement of the sediments over the time of the experiments, which can be caused by the difficulties in uniformly packing soil columns. It is possible that a similar process took place in column 1 but in a lesser degree that did not affect the final value of porosity.

5.2.1.4 Sampling and analysis

In total, six (6) experiments with pesticides were performed at 2 different temperatures (Table 5-7). The first two experiments (E 1-1 and E 2-2) involved only one column, while the last four (3-1, 3-2, 4-1 and 4-2) were run in both columns at the same time, receiving water from the same reservoir.

A solution mix of approximately 10 mg of each pesticide (Ametryn, Atrazine, Carbofuran, Diuron and Propanil) in 1 liter of filtered Elbe water was prepared at the Institute of Water Chemistry at TU-Dresden. For experiments E 1-1 and E 2-2, five hundred milliliters of such solution were mixed with 10 liters of filtered Elbe water (filter size = 1.2 μm) for the latter, and unfiltered Elbe water for the former. For experiments E 3-1, E 3-2, E 4-1 and E 4-2, the whole liter of the pesticide solution was mixed with 20 liters of filtered Elbe water, since both columns were fed from the same reservoir.

Table 5-7 Number of experiments performed at different temperature, for each column

Experiment No.* Column 1 Column 2

E 1-1 20oC -

E 2-2 - 10oC

E 3-1 10oC -

E 3-2 - 10oC

E 4-1 20oC -

E 4-2 - 20oC

* The first number represents the experiment number and the second number represents the column.

This mixture was slowly added to the columns using a membrane pump, and then the samples were collected at the outlet and filtered (filter size = 0.45 µm) before saving them in the vials provided by the Institute of Water Chemistry at TU-Dresden. Sometimes a sample of the inlet water was taken for control effects. The samples were preserved in a refrigerator at 4oC until it was time to take them to TU. Also, a sample from the Elbe river water, with no pesticide in it, was analyzed to determine if any of the pesticides under investigation was present. None of them were found. The analysis of all the samples was performed by members of the Institute of Water Chemistry at TU-Dresden, using high-performance liquid chromatography (HPLC) technique.

The average discharge, the time length of each experiment, the total number of samples collected, the number of inlet samples, and the number of samples analyzed, are presented in Table 5-8.

Evaluation of the Potential for Riverbank Filtration in Colombia

96

Table 5-8 General data related to the sampling

Experiment No. E 1-1 E 2-2 E 3-1 E 3-2 E 4-1 E 4-2

Q (ml/min) 0.83 0.56 0.58 0.57 0.46 0.55

Duration time (min) 10560 12640 11750 11750 12350 12350

No. of outlet samples taken 13 36 22 22 22 22

No. of inlet samples taken 4 9 6 6 5 5

No. of inlet samples analyzed 4 8 6 6 5 5

No. of outlet samples analyzed 13 29 22 22 22 22

In order to obtain the breakthrough curves for the components, it was necessary to define an initial concentration, C0, for each pesticide in each experiment. Since the concentration values obtained for the inlet samples during the lab analysis varied in different degrees depending on the component, it was decided to eliminate the obvious outliers, i.e. those that were one order of magnitude lower than the majority of the samples for the same pesticide. With the rest of the values, an average initial concentration for each pesticide and experiment was calculated, and the results are shown in Table 5-9. Notice that experiments E3-1 and E 3-2 share the same initial concentration, just as it happens with experiments E 4-1 and E 4-2 because each time both columns were fed from the same reservoir and that is the reason of the names given to the experiments (E3 and E4). It is important to mention that the differences in the inlet concentrations (represented by the standard deviation, σ, in Table 4-9) for a given pesticide and experiment did not seem to be dependent on temperature, or at least no pattern was found that could prove otherwise.

Table 5-9 Initial concentrations selected for each pesticide and experiment

C0 E1-1 (20

oC) E2-2 (10

oC) E3 (10

oC) E4 (20

oC)

Avg σ Avg σ Avg σ Avg σ

Ametryn 243.3 24.2 238.7 27.4 991.7 14.8 427.5 35.5

Atrazine 252.8 20.8 255.9 23.0 475.9 25.0 476.0 6.2

Carbofuran 535.8 63.9 560.0 27.9 591.0 12.3 564.0 27.9

Diuron 438.0 62.4 626.6 55.3 379.5 26.1 458.7 15.8

Propanil 278.7 56.1 292.9 57.5 452.1 5.6 178.0 1.4

Once the initial concentrations were calculated, the normalized concentration, C/Co, of each pesticide in the outlet samples was plotted against the pore volumes equivalent to the time at which each sample was taken. As it was stated earlier, one pore volume is the volume of the column multiplied by the porosity (ALn). In order to express the dimensionless time, tR, in terms of pore volumes, Eq. 5-2 was applied (Fetter, 1999):

𝑃𝑉𝑖 =𝑡𝑖𝑄

𝑃𝑉= 𝑡𝑅 Eq. 5-2

where, PV is the pore volume found during the tracer test [L3], t is time [T], Q is the average discharge rate from the column [L3/T], and i is the sub index for each time a sample was taken. The previous equation can also be expressed in a dimensionless way, as shown in Eq. 5-3:

Chapter 5 – Pesticides Removal During RBF

97

𝑃𝑉𝑖 =𝑣𝑥𝑡𝑖

𝐿 Eq. 5-3

where, 𝑣𝑥 is the linear groundwater velocity or pore-water velocity [L/T] (equal to Q/An), and L is the length of the column [L]. The values of porosity used for each column experiment are presented in Table 5-10.

Table 5-10 Porosity value used to calculate the equivalent pore volume

Column Experiment n value Associated Tracer Test

E 1-1 0.34 TT 1-1

E 2-2 0.32 TT 2-2

E 3-1 0.35 TT 3-1

E 3-2 0.32 TT 2-2

E 4-1 0.34 TT 4-1

E 4-2 0.36 TT 5-2

Finally, the BTCs were plotted and the most obvious outliers (caused by either human errors during the sampling process or instrumental errors during the analysis) were removed (Figure 5-4 to Figure 5-9). An initial value of the retardation factor, R*, was calculated for each component in each one of the experiments as the time (in pore volumes) at which the relative concentration (C/Co) was equal to 0.5. This value was used later on as initial retardation factor during the inverse modeling.

The results for Carbofuran, correspondent to experiment #4 (Figure 5-8 and Figure 5-9), presented some inconsistencies due to problems being experimented at the laboratory at TU-Dresden, and therefore it was decided not to include them in the next steps of analysis.

Figure 5-4 Lab results E 1-1 (Column 1 at 20oC)

Figure 5-5 Lab results E 2-2 (Column 2 at 10oC)

Evaluation of the Potential for Riverbank Filtration in Colombia

98

Figure 5-6 Lab results E 3-1 (Column 1 at 10oC)

Figure 5-7 Lab results E 3-2 (Column 2 at 10oC)

Figure 5-8 Lab results E 4-1 (Column 1 at 20oC)

Figure 5-9 Lab results E 4-2 (Column 2 at 20oC)

In all the cases the concentration in the outlet never reached the same value as the concentration in the inlet, most probable due to degradation and sorption processes, as it will be explained later on this chapter. This is especially noticeable for Propanil, which is almost completely missing in the outlet samples since it transformed into its metabolite 3,4 Dichloroaniline, as can be seen in Figure 5-10 where the concentration of this compound (plotted as area under the curve against time in minutes) in the outlet samples increases with time.

According to the Hazardous Substances Data Bank (HSDB), a database of the National Library of Medicine's TOXNET system (http://toxnet.nlm.nih.gov), 3,4-Dichloroaniline is not only the main degradation product of the herbicide Propanil and one of the metabolites of Diuron but also a more toxic substance than its two parental compounds.

Taking into consideration that the time required for the chemical concentration of Propanil to decline to 50% of the amount of application (DT50) in a water-sediment system is 4.4 days, whereas that of Diuron is 75.5 days, it can be inferred that most of the Dichloroaniline found in the inlet water came from the degradation of Propanil. This and more information on the pesticides used in this study can be found in the Pesticides Properties Data Base (PPDB), which is a comprehensive relational database of pesticide

Chapter 5 – Pesticides Removal During RBF

99

physicochemical and ecotoxicological data developed by the Agriculture and Environment Research Unit (AERU) at the University of Hertfordshire (England) that can be accessed online at http://sitem.herts.ac.uk/aeru/ppdb/en/index.htm.

Figure 5-10 Concentration of 3,4 Dichloroaniline in the outlet samples with time

5.2.2 Inverse Modeling

5.2.2.1 Conceptual framework

The code CXTFIT 2.0, included in the freeware STANMOD v. 2.8. (Simunek et al., 1999) was used to determine the transport parameters associated to the column tests. STANMOD (STudio of ANalytical MODels) is a public domain computer software package for evaluating solute transport in porous media using analytical solutions of the advection-dispersion equation (ADE). The software includes a modified and updated version of the CXTFIT code of Toride et al. (1995) for the estimation of solute transport parameters from observed concentrations (inverse problem) or the prediction of solute concentrations (direct problem), using the one-dimensional ADE.

The ADE for one-dimensional flow in a homogeneous, isotropic porous media is shown in Eq. 5-4 (Fetter, 1999):

𝜕𝐶

𝜕𝑡= 𝐷𝐿

𝜕2𝐶

𝜕𝑥2− 𝑣𝑥

𝜕𝐶

𝜕𝑥 Eq. 5-4

where, 𝜕𝐶 𝜕𝑡⁄ is the change in concentration with time [MT-1/L3], 𝜕𝐶 𝜕𝑥⁄ is the concentration gradient [M/L2], 𝑣𝑥 is the average linear velocity or pore-water velocity [L/T], and DL is the hydrodynamic dispersion coefficient, or just dispersion coefficient, parallel to the principal direction of flow (longitudinal) [L2/T].

The previous equation can be modified to include sorption and decay, as in Eq. 5-5 (Miller and Webber, 1984):

𝜕𝐶

𝜕𝑡= 𝐷𝐿

𝜕2𝐶

𝜕𝑥2− 𝑣𝑥

𝜕𝐶

𝜕𝑥−

𝐵𝑑

𝜃

𝜕𝐶∗

𝜕𝑡+ (

𝜕𝐶

𝜕𝑡)

𝑟𝑥𝑛 Eq. 5-5

Evaluation of the Potential for Riverbank Filtration in Colombia

100

where, Bd is the bulk density of the aquifer [M/L3], Ө is the volumetric moisture content or porosity of the saturated media [L3/L3], C* is the amount of solute sorbed per unit weight of solid [M/M], and rxn is the subscript indicating biological or chemical reaction of the solute other than sorption.

Furthermore, the ADE for one-dimensional transport of reactive solutes, subject to adsorption, fist-order degradation and zero-order production, in a homogeneous soil can be written as Eq. 5-6 (Fetter, 1999; Toride et al., 1995):

𝜕

𝜕𝑡(𝜃𝐶 + 𝐵𝑑𝐶∗) =

𝜕

𝜕𝑥(𝜃𝐷𝐿

𝜕𝐶

𝜕𝑥− 𝑣𝑥𝐶) − 𝜇𝑙𝜃𝐶 − 𝜇𝑠𝐵𝑑𝐶∗ + 𝜉𝑙𝜃 + 𝜉𝑠𝐵𝑑 Eq. 5-6

where, C is the concentration of solute in liquid phase [M/L3], µl and µs are the first-order decay coefficients for degradation of the solute in the liquid and absorbed phases, respectively [1/T], and ξl [M/L3T] and ξs [1/T] are zero-order production terms for the liquid and absorbed phases, respectively.

Considering a linear sorption isotherm (C* = KdC, where Kd is the distribution coefficient [M/L]), and assuming steady-state flow, Eq. 5-6 can be written as:

𝑅𝜕𝐶

𝜕𝑡= 𝐷𝐿

𝜕2𝐶

𝜕𝑥2− 𝑣𝑥

𝜕𝐶

𝜕𝑥− 𝜇𝐶 + 𝜉 Eq. 5-7

where, R is the retardation factor, and µ and ξ are the combined first- and zero-order rate coefficients for the liquid and sorbed phases. Furthermore,

𝑅 = 1 +𝐵𝑑𝐾𝑑

𝜃 𝜇 = 𝜇𝑙 +

𝜇𝑠𝐵𝑑𝐾𝑑

𝜃 𝜉 = 𝜉𝑙 +

𝜉𝑠𝐵𝑑

𝜃 Eq. 5-8

With the software, the inverse problem is solved by minimizing an objective function, which consists of the sum of the squared differences between observed and fitted concentrations, using a nonlinear least-squares inversion approach based on the Lavenberg-Marquardt method (Marquardt, 1963).

The program may be used to estimate the pore-water velocity (v), the dispersion coefficient (D), the retardation factor (R), the first-order degradation coefficient (µ), and/or the zero-order production coefficient (ξ), from observed concentration distributions versus time and/or distance.

In this particular case, the deterministic equilibrium ADE (Eq. 5-7) was selected as the transport model to solve the inverse problem. The software also offers the possibility of using a chemical and physical non-equilibrium ADE, and a stochastic stream tube model based on the local-scale equilibrium or non-equilibrium ADE (Toride et al., 1995).

The general initial condition included in the CXTFIT is given by:

𝐶(𝑧, 0) = 𝐶𝑖(𝑧) Eq. 5-9

Chapter 5 – Pesticides Removal During RBF

101

where, Ci is the initial concentration as a function of z, i.e. at time equals zero and at some place in the soil column, z, the solute concentration is Ci.

For the inlet boundary condition, either a first- or third-type condition (fixed concentration or variable flux, respectively) can be used. The first-type condition states that the concentration introduced into the top of the soil column, where z = 0, at some time, t, is C0 (Eq. 5-10).

𝐶(0, 𝑡) = 𝐶0(𝑡) Eq. 5-10

On the other hand, the third-type condition states that, the rate at which a solute is introduced in the column, by both advection and diffusion, is equal to the pore-water velocity, v, times the initial inlet concentration, C0 (Eq. 5-11):

𝑣𝐶(0, 𝑡) − 𝐷𝜕𝐶(0, 𝑡)

𝜕𝑧= 𝑣𝐶0(𝑡) Eq. 5-11

Van Genuchten and Parker (1984, 1984a and 1994, in Toride et al., 1995) suggest that a third-type inlet condition is to be preferred for most transport scenarios because, assuming there is no dispersion outside the soil, mass is conserved. Thus, for the purpose of this work, a third-type inlet resident concentration condition was selected, as well as a zero initial concentration, Ci, and a zero production coefficient, ξ.

In order to determine the transport and reaction parameters, the code requires the input of the pore-water velocity, v, and the initial values for the dispersion coefficient, D, the retardation factor, R, and the first-order degradation coefficient, µ. The given pore-water velocity was the one calculated during each column experiment, and the initial retardation factor used was the one previously defined as R*, or the time (in pore volumes) at which the concentration of each component in the outlet sample was half of that in the inlet, as explained earlier and shown in Figure 5-4 to Figure 5-9. On the other hand, the initial values of dispersion and first-order dispersion coefficients were changed at the beginning of each run in order to obtain a better fit, based on the minimum square error (MSE).

In all cases the software calculated fitted values for D, R and µ. Since dimensionless time and dimensional position were chosen as time and space units, all parameters calculated, except D, are dimensionless.

5.2.2.2 Results and discussion

The main parameters of each experiment, i.e. temperature, discharge, pore-water velocity (v = Q/An), and time in minutes and in pore volumes are shown in Table 5-11. The results of the inverse modeling are shown in Table 5-12, where the calculated pore-water velocity (v), and the retardation factors (R*) obtained from the BTCs shown in Figure 5-4 to Figure 5-9, are referred to as “Observed” values. The values shown as “Fitted” are those found with CXTFIT and correspond to the best fitted values according to the MSE. As explained before, Carbofuran was not modeled for experiment # 4.

Evaluation of the Potential for Riverbank Filtration in Colombia

102

In general, the lowest value of dispersion coefficient (D) was 0.04 cm2/min, obtained for Atrazine during experiment 4-2, and the highest was 4.65 cm2/min for Diuron in experiment 4-2. Carbofuran registered the lowest retardation factor (R) with a value of 1.1 during experiment 3-2, and Diuron had the highest R of 5.6 in experiment 4-1. Finally, the values for the degradation coefficient (μ) fluctuated between 0 for Diuron during experiment 2-2 and 3 x 10-1 for Carbofuran in experiment 1-1.

Table 5-11 Main parameters of the column experiments used during inverse modeling

Experiment T oC Q (cm

3/min) v (cm/min) t (min) time (PV)

1-1 20 0.83 0.054 10560 11.3

2-2 10 0.56 0.038 12640 9.7

3-1 10 0.58 0.037 11750 8.6

3-2 10 0.57 0.039 11750 9.3

4-1 20 0.46 0.030 12350 7.4

4-2 20 0.55 0.034 12350 8.3

Table 5-12 Results of the modeling for each pesticide and each experiment

Pesticide Experiment Obs Fitted

MSE (10-3

) R2 v

(cm/min) R*

D (cm

2/min)

R μ

Ametryn

1-1 0.054 2.5 0.48 2.0 2.1E-01 3.358 0.9600

2-2 0.038 2.5 1.27 2.1 4.8E-02 0.872 0.9905

3-1 0.037 2.9 0.44 2.4 1.6E-01 0.230 0.9975

3-2 0.039 2.4 1.04 2.0 1.6E-01 0.375 0.9945

4-1 0.030 3.1 0.44 2.6 1.6E-01 0.328 0.9964

4-2 0.034 2.6 2.66 1.9 5.0E-05 0.689 0.9884

Atrazine

1-1 0.054 1.7 0.62 1.5 4.9E-02 0.518 0.9943

2-2 0.038 1.6 0.91 1.4 9.0E-09 0.496 0.9954

3-1 0.037 1.7 0.41 1.7 2.3E-02 0.281 0.9976

3-2 0.039 1.5 1.35 1.3 9.0E-09 0.789 0.9909

4-1 0.030 1.8 0.04 1.7 6.5E-02 0.302 0.9972

4-2 0.034 1.7 0.18 1.3 2.6E-04 0.409 0.9936

Carbofuran

1-1 0.054 1.5 0.25 1.1 3.0E-01 3.161 0.9171

2-2 0.038 1.5 1.00 1.3 1.6E-01 0.344 0.9942

3-1 0.037 1.6 0.34 1.4 2.9E-01 0.163 0.9972

3-2 0.039 1.6 1.57 1.1 2.2E-01 0.265 0.9930

Diuron

1-1 0.054 4.1 0.69 3.3 2.4E-01 0.358 0.9961

2-2 0.038 4.7 1.39 4.0 0 0.253 0.9968

3-1 0.037 5.3 0.44 4.2 2.7E-01 0.250 0.9963

3-2 0.039 4.2 1.46 3.4 1.1E-01 0.368 0.9939

4-1 0.030 5.5 1.00 5.6 8.6E-08 0.456 0.9896

4-2 0.034 5.4 4.65 2.9 1.9E-07 1.193 0.9685

Specifically, in the case of Ametryn, D ranged between 0.44 cm2/min (experiments 3-1 and 4-1) and 2.66 cm2/min (experiment 4-2), R fluctuated from 1.9 and 2.6 (experiments 4-2 and 4-1, respectively), and μ varied between 5 x 10-5 during experiment 4-2 and 2.1 x 10-1 for experiment 1-1. For Atrazine, the value of D was between 0.04 and 1.35 cm2/min

Chapter 5 – Pesticides Removal During RBF

103

(experiments 4-1 and 3-2, respectively), whereas the one for μ fluctuated between 9 x 10-9 (experiment 2-2) and 6.5 x 10-2 (experiment 4-1), and the value of R had a lowest of 1.3 during experiments 3-2 and 4-2 and a highest of 1.7 for experiments 3-1 and 4-1. These values are similar to retardation factors found for Atrazine in alluvial aquifers by Pang and Close, 1999 (R = 1) and Agertved et al., 1993 (R = 1.2), and for column experiments by Mao and Ren, 2005 (1.82 ≤ R ≤ 1.95). Carbofuran, on the other hand, presented its lowest value of dispersion coefficient during experiment 1.1 (0.25 cm2/min) and a highest during experiment 3-2 (1.57 cm2/min), the retardation factor ranged from 1.1 (experiments 1-1 and 3-2) and 1.4 (experiment 3-1), and the degradation coefficient varied between 1.6 x 10-1 and 3 x 10-1 (experiments 2-2 and 1-1, respectively). Finally, for Diuron, the values of D ranged from 0.44 cm2/min during experiment 3-1 and 4.65 cm2/min during experiment 4-2, R varied from 2.9 to 5.6 (experiments 4-2 and 4-9, respectively), and μ had a maximum value of 2.7 x 10-1 for experiment, whilst during experiment 2-2 there was no degradation at all (μ = 0).

Even though it is well known that sorption and degradation processes are affected by temperature (Paraiba and Spadotto, 2002), in this particular case it was not possible to find any clear trend relating the two different temperatures (10oC and 20oC) with the coefficients of dispersion and degradation, D and μ, or with the retardation factor, R.

As it can be seen in Table 5-13, the resulted values for Diuron present the highest standard deviation of all the parameters, whereas the lowest standard deviation for dispersion coefficient and degradation are those of Atrazine and, for the retardation factor, the correspondent to Carbofuran.

Table 5-13 Basic statistic parameters for the fitted values for each pesticide

Pesticide D (cm

2/min) R μ

Mean St.Dev. Mean St.Dev. Mean St.Dev.

Ametryn 1.06 0.86 2.16 0.28 0.12 0.08

Atrazine 0.58 0.49 1.49 0.19 0.02 0.03

Carbofuran 0.79 0.62 1.22 0.12 0.24 0.07

Diuron 1.60 1.54 3.89 0.96 0.10 0.12

The best fitted BTCs obtained from the inverse modeling (solid line), and the observed values from the column experiments (only markers) are depicted in Figure 5-11 to Figure 5-16. These graphs also include the BTC for the tracer test whose value of effective porosity, n, was used to calculate the time in pore volumes at which each sample was taken (see Table 5-6). Because the tracer used is a conservative contaminant, NaCl, the relative concentration, C/Co, at 1 PV is equal to 0.5 (center of mass), i.e. R = 1 and μ = 0.

The shapes of the BTCs are directly related to the dispersion coefficient and degradation factor (Fetter, 1999). Thus, the skewness of the curves reflects de degree of dispersion, i.e. the higher the dispersion coefficient, the higher the skewness of the BTC. Also, the higher the degradation, the lower is the total amount of contaminant that appears at the bottom of the column, which means that the relative concentration, C/Co, will be further away from the unit. Finally, since the retardation factor is a measure of adsorption behavior, or how tightly the pesticide binds or sticks to soil particles (Eckhard, 1999), the higher the

Evaluation of the Potential for Riverbank Filtration in Colombia

104

value of R the stronger the substance fixes onto soil material and the slower it appears at the bottom of the column.

Figure 5-11 BTC and observed values E 1-1

Figure 5-12 BTC and observed values E 2-2

Figure 5-13 BTC and observed values E 3-1

Figure 5-14 BTC and observed values E 3-2

Figure 5-15 BTC and observed values E 4-1

Figure 5-16 BTC and observed values E 4-2

Chapter 5 – Pesticides Removal During RBF

105

From the BTCs is clear that, in all cases, Diuron presents the highest skewness, which coincides with the higher values of dispersion coefficient for this pesticide in most of the experiments. In terms of degradation, the BTCs show that Atrazine is always the component whose final relative concentration is closer to 1, indicating that the degradation of such compound is the lowest of all the studied pesticides. However, the values of μ presented in Table 5-12 show that degradation is not always lowest for this particular pesticide, but vary from experiment to experiment between Atrazine and Diuron. These could be because these two pesticides are less or not degraded under aerobic conditions than they would be under anoxic conditions (Jekel and Grischek, 2003). One of the reasons why the low degradation of Diuron cannot be inferred from the BTCs could be the fact that the duration time of the experiments was not long enough as for the pesticide to leave the column due to its high retardation factor which translates into a slower travel velocity than the rest of the studied compounds.

Based on the results of the inverse modeling, it was decided to use the data from experiment 3-1 to feed the contaminant transport model, since those were the values with the lowest MSE for all the pesticides, according to the results shown in Table 5-12.

5.3 CONTAMINANT TRANSPORT MODEL

The values of retardation factor, dispersion coefficient and degradation coefficient correspondent to one of the column experiments were selected to feed a contaminant transport model. This model was run in a transient state with the code MT3DMS (Zheng and Wang, 1999) based on a steady-state groundwater flow model developed in Processing MODFLOW (PMWIN) Version 5.3.3 (Chiang and Kinzelbach, 1998). The software includes other codes such as MODFLOW (McDonald and Harbaugh 1988), MT3D (Zheng, 1990), MOC3D (Konikow et al., 1996), PEST (Doherty et al., 1994) and UCODE (Poeter and Hill, 1998).

5.3.1 Solute Transport Model

5.3.1.1 Methods

The site selected for the model domain corresponds to an RBF site in Lößnitztal, in the south of the state of Saxony in East Germany, where the sediments used to fill the columns were collected. In this location there are four (4) production wells (BR), three (3) observation wells (GWM), and four (4) points for measurements of surface water level (FWM), as shown in Figure 5-17.

The area was divided in 93 rows and 138 columns of 2m x 2m, and one layer set to be unconfined because it was an alluvial aquifer directly recharged by precipitation. The medium was considered to be homogeneous, with a hydraulic conductivity of 1.3 x 10-4 m/s (HTW-Dresden 2012), and an effective porosity of 0.35.

Both, the conductivity and the thickness of the aquifer were the parameters used for calibration and it was found that the model was very sensitive to the aquifer thickness and

Evaluation of the Potential for Riverbank Filtration in Colombia

106

not so much to the conductivity. For this reason, the hydraulic conductivity was equal throughout the domain but the thickness was changed according to field interpretations of the topography of the site as shown in Figure 5-18, where the elevation of the top was 345 m.a.s.l.

Figure 5-17 Study area in Lößnitztal, Germany

The boundary conditions were defined as shown in Figure 5-19. To the east and west of the active area a constant head boundary was selected, whereas in the south a RIVER boundary was used. The values used for the constant heads were the data of water level at points FWM1 for the east boundary and FWM4 for the west, measured on the 27th of April 2012.

Figure 5-18 Bottom elevations (in meters above sea level) of the aquifer

Chapter 5 – Pesticides Removal During RBF

107

Figure 5-19 Boundary conditions for the groundwater flow model

For the RIVER boundary, a conductance value of 0.0004 m2/s was used, calculated from an assumed thickness of the riverbed of 0.1 m and a permeability of the riverbed of 1 x 10-5 m/s (HTW Dresden 2012). The conductance was calculated with the Eq. 5-12:

𝐶𝑜𝑛𝑑𝑢𝑐𝑡𝑎𝑛𝑐𝑒 =𝐾𝑥𝐿𝑥𝑊

𝑀 Eq. 5-12

where K is the hydraulic conductivity of the riverbed material [L/T], L x W is the area of the cell [L2], and M is the thickness of the riverbed [L].

The elevation of the river bottom was calculated with the river head data that came from the interpolation of the data collected from the FWM points on the 27th of April of 2012, and a constant water level of 0.5 m.

Based on the knowledge of the area by some of the members of the teamwork at HTW-Dresden, the recharge over the main area of interest was set to 3 x 10-9 m/s, and for the northern part of the model, a recharge from run-off was assumed to be 5 x 10-7 m/s, considering a precipitation of 200 mm/yr and a run-off area of aproximately 275 m x 500 m. Figure 5-20 shows the map of the recharge over the entire modeled area.

The data for the initial head of the boreholes is shown in Table 5-14. This data is from the 27th of April, 2012 and was supplied by the people of the Lößnitztal waterworks. This date was selected because it lies within the period of time selected to calculate the pumping rate for the production well BR3.

Two scenarios of pumping of the production well BR3 were selected, one for calibration and the other for validation and particle tracking simulations, and also for the contaminant transport model. In both cases, in order to calculate the discharge of the well BR3, the period of time from the 26th of April at 8:15 to the 29th of May at 8:30 was selected. During that time the well pumped a total volume of 2322 m3 in 680 hours, approximately.

In scenario 1 (used for calibration) the discharge of the well was calculated dividing the total extracted volume by the total of pumping hours, i.e. assuming a continuous pumping

Evaluation of the Potential for Riverbank Filtration in Colombia

108

24 hours per day, for a value of 9.5 x 10-4 m3/s. For scenario 2, on the other hand, the pumping was assumed to occur only 10 hours per day (according to verbal communication of the people in charge of the waterworks). Since the model was run in steady-state, the pumping of scenario 1 was multiplied by 10/24, for a resulting value of 3.95 x 10-4 m3/s.

Figure 5-20 Recharge from precipitation (blue) and from run-off (black)

Table 5-14 Water heads at the boreholes used as the initial heads in the simulation

Borehole name Water head (m.a.s.l) BR1 339.41 BR2 340.33 BR3* 339.87 BR4 340.91 GWM1 340.43 GWM2 341.8 GWM3 340.76 *Production well

At the end, the software PMPATH (Chiang and Kinzelbach, 1994), contained in the PMWIN, was used to determine, with particle tracking, the flowlines both for water free of pesticides and for each one of the component Ametryn, Atrazine, Carbofuran and Diuron.

5.3.1.2 Results and discussion

The resulting piezometric head isolines for scenario 1 (pumping rate = 9.5 x 10-4 m3/s) are shown in Figure 5-21, and the calibration results are presented in Figure 5-22 as a graph of calculated vs observed heads. The normalized root-mean-square error (NRMSE) of this set of values was 0.026% and the correlation was 0.993. The error was normalized with the mean value of the observed heads, like this:

𝑅𝑀𝑆𝐸 = √∑ (ℎ𝑜𝑏𝑠,𝑖 − ℎ𝑐𝑎𝑙𝑐,𝑖)2𝑛

𝑖=1

𝑛 Eq. 5-13

where, RMSE is the root-mean square error, and hobs and hcalc are the observed and the calculated water heads in the wells, respectively.

Chapter 5 – Pesticides Removal During RBF

109

From Eq. 5-12, the NRMSE is calculates as followed:

𝑁𝑅𝑀𝑆𝐸 =𝑅𝑀𝑆𝐸

ℎ𝑜𝑏𝑠

Eq. 5-14

where ℎ𝑜𝑏𝑠 is the average of the observed water heads in the wells.

Figure 5-21 Isolines from scenario 1 of the GW flow model

Figure 5-22 Comparison of calculated and observed heads for scenario 1

The results of scenario 2 (pumping rate = 3.95 x 10-4 m3/s) are shown in Figure 5-23 and Figure 5-24. As it can be seen, the head isolines for both scenarios are very similar except around the production well BR3. In this case, the NRMSE was 0.031% and the correlation between observed and calculated heads was 0.996.

Figure 5-23 Isolines from scenario 2 of the GW flow model

Figure 5-24 Comparison of calculated and observed heads for scenario 2

With the use of backward particle tracking, the capture zone of well BR3 was found and the flowlines were extracted (Figure 5-25). The PMPATH allows changing, among others, the step length and the value of retardation factor. In this case, for illustration purposes, the step length was set to 60 days for every case and the retardation factor was changed

Evaluation of the Potential for Riverbank Filtration in Colombia

110

for each pesticide, according to the results of the CXTFIT model for experiment 3-1. The retardation factor, R, for clean water, i.e. water without pesticides, is assumed to be 1.

Figure 5-25 Flowlines and travel time marks (every 60 days) for clean water (R = 1)

The results of travel times (earliest arrival) in days and the values of R for each pesticide and for clean water are presented in Table 5-15. As expected, the higher the R, the longer the time it takes the fluid to travel from the river to the well. It is important to keep in mind that these results only take into account transport by advection, and it completely disregards the effect of dispersion and degradation in the contaminant.

Table 5-15 Retardation factor and travel time from the river to the well for each pesticide and for clean water

Component R Average travel time

(days)

Clean water 1 150

Ametryn 2.4 300

Atrazine 1.7 210

Carbofuran 1.4 180

Diuron 4.2 540

5.3.2 Contaminant Transport Model

MT3DMS is a modular 3-D multi-species transport model for simulation of advection, dispersion, and chemical reactions of contaminants in groundwater systems (Zheng and Wang, 1999). The model is based on the partial differential equation describing the fate and transport of contaminants of species 𝑘 in three-dimensional, transient groundwater flow systems as shown in Eq. 5-15.

Chapter 5 – Pesticides Removal During RBF

111

𝜕(𝜃𝐶𝑘)

𝜕𝑡=

𝜕

𝜕𝑥𝑖(𝜃𝐷𝑖𝑗

𝜕𝐶𝑘

𝜕𝑥𝑗) −

𝜕

𝜕𝑥𝑖

(𝜃𝑣𝑖𝐶𝑘) + 𝑞𝑠𝐶𝑘,𝑠 + ∑ 𝑅𝑛 Eq. 5-15

where 𝐶𝑘 is the dissolved concentration of species k [ML-3], 𝜃 is the porosity of the subsurface medium [dimensionless], 𝑡 is time [T], 𝑥𝑖 is the distance along the respective cartesian coordinate axis [L], 𝐷𝑖𝑗 is the hydrodynamic dispersion coefficient tensor [L2T-1],

𝑣𝑖 is the seepage or linear pore water velocity [LT-1] (𝑣𝑖 = 𝑞𝑖/𝜃), 𝑞𝑠 is the volumetric flow rate per unit volume of aquifer representing fluid sources (positive) and sinks (negative) [T-1], 𝐶𝑘,𝑠 is the concentration of the source or sink flux for species k [ML-3], ∑ 𝑅𝑛 is the

chemical reaction term [ML-3T-1] (general biochemical and geochemical reactions, i.e. aqueous-solid surface reaction (sorption) and first-order rate reaction).

5.3.2.1 Advection package

This package offers the possibility of using five different schemes for the solution of the advection term: Standard finite difference method, Method of characteristics (MOC), Modified method of characteristics (MMOC), Hybrid MOC/MMOC (HMOC), and 3rd order TVD scheme (ULTIMATE). Each solution has its own weighting scheme, particle tracking algorithm, and simulation parameters that have to be given to the model. More details on the mathematics behind the solution schemes can be found in Zheng and Wang (1999).

Before deciding which solution scheme to use, Chiang (2005) suggests to find out whether or not the problem to solve is advection-dominated. In order to do that the author recommends to calculate the Peclet number, Pe, which is defined in Eq. 5-16.

𝑃𝑒 =|𝑣|𝐿

𝐷 Eq. 5-16

where |𝑣| is the magnitude of the seepage velocity vector [LT-1], L is a characteristic length commonly taken as the grid cell width [L], and D is the Dispersion coefficient [L2T-1].

The Pe was then calculated for this particular case in two ways: first, using the seepage velocity from experiment 3-1 and the length of the columns; second, calculating the seepage velocity from the groundwater flow model and the width of the grid cell. In both cases the dispersion coefficient used was the average of the dispersion coefficients of the four pesticides. The magnitude of the seepage velocity for the second case was calculated like this:

𝑣 =𝑞

𝑛𝑒=

𝑄

𝐴𝑛𝑒=

𝑘

𝑛𝑒

𝑑ℎ

𝑑𝑙= 5.86𝑥10−6𝑚/𝑠 Eq. 5-17

In this case 𝑘 = 1.3 x 10-4 m/s, 𝑛𝑒 = 0.35, and 𝑑ℎ/𝑑𝑙 = 0.0158 (gradient from the river to the well, extracted from the graphs of the groundwater flow model).

Finally, the results of the Peclet numbers are shown in Eq. 5-18 and Eq. 5-19.

Evaluation of the Potential for Riverbank Filtration in Colombia

112

𝑃𝑒1 =0.037 𝑐𝑚 𝑚𝑖𝑛⁄ ∗ 50𝑐𝑚

0.4075𝑐𝑚2/𝑚𝑖𝑛= 4.5 Eq. 5-18

𝑃𝑒2 =5.85𝑥10−6 𝑚 𝑠⁄ ∗ 2𝑚

6.8𝑥10−7𝑚2/𝑠= 17 Eq. 5-19

With these values of Peclet number it is safe to say that the problem is not advection-dominated, and therefore it can be expected that numerical dispersion during the solution of the transport equation will not be a significant problem (Zheng and Wang, 1999). In a case like this, the authors suggest that the standard finite difference method can be used for greater computation efficiency and for obtaining first approximations in the initial stages of modeling. For all these reasons, the Upstream Finite Difference solution scheme was selected for this study.

5.3.2.2 Dispersion package

The following values must be specified for each layer in the Dispersion Package:

TRPT: the ratio of horizontal transverse dispersivity and longitudinal dispersivity

TRPV: the ratio of vertical transverse dispersivity and longitudinal dispersivity

DMCOEF: effective molecular diffusion coefficient, L2T-1

Considering that one of the assumptions in this study to solve the ADE is that the porous medium is homogeneous and isotropic, both for groundwater flow and contaminant transport, the horizontal and the vertical transverse dispersivities were therefore also assumed to be the constant during the simulations.

The definition of dispersion coefficient, D, can be used to calculate the value of the longitudinal dispersion coefficient using the values of pore-water velocity and dispersion coefficient found with CXTFIT. Thus:

𝐷𝐿 = 𝛼𝐿𝑣𝑖 + 𝐷∗ → 𝛼𝐿 =𝐷𝐿 − 𝐷∗

𝑣𝑖 Eq. 5-20

𝐷∗ = 𝜔𝐷𝑑 Eq. 5-21

where 𝐷𝐿 is the hydrodynamic dispersion coefficient parallel to the principal direction of flow (longitudinal) [L2T-1] (D found through CXTFIT), 𝛼𝐿 is the longitudinal dynamic dispersivity [L], 𝑣𝑖 is the average linear velocity in the principal direction of flow [LT-1] (pore-water velocity from column experiment 3-1), 𝐷∗ is the effective diffusion coefficient [L2T-1], Dd is the molecular diffusion coefficient [L2T-1], and ω is a coefficient related to tortuosity [dimensionless].

ω is a measure of the effect of the shape of the flowpath of a water molecule in a porous media, and for laboratory experiments its value ranges between 0.01 and 0.7, according to Perkins and Johnson (1963) and Freeze and Cherry (1979), respectively.

Chapter 5 – Pesticides Removal During RBF

113

Table 5-16 presents the results for longitudinal dispersivity, L, calculated using Eq. 5-20. Two effective diffusion coefficients were calculated, one with ω equal to 0.7 (D1

*), and the other with ω equal to 0.01 (D2

*), which are the two ends of the range suggested by Perkins

and Johnson (1963) and Freeze and Cherry (1979). Therefore, two different values of L

were also found, one for each value of ω ( and 2). The last column presents the results

when the effective diffusion coefficient is not taken into account into the calculation of L.

Table 5-16 Values of αL found from the equation for hydrodynamic dispersion coefficient (DL)

Pesticides DL

(cm2/min)

Dd (cm

2/min)

D1* (=0.7) (cm

2/min)

D2* (=0.01) (cm

2/min)

1

(cm) 2

(cm) 3 (D*=0)

(cm)

Ametryn 0.44 3.4E-04 2.4E-04 3.4E-06 11.885 11.892 11.892

Atrazine 0.41 3.4E-04 2.4E-04 3.4E-06 11.075 11.081 11.081

Carbofuran 0.34 3.2E-04 2.3E-04 3.2E-06 9.183 9.189 9.189

Diuron 0.44 3.2E-04 2.2E-04 3.2E-06 11.886 11.892 11.892

Average 0.41 3.3E-04 2.3E-04 3.3E-06 11.007 11.013 11.014

In view of the fact that the value of the effective diffusion coefficient, D*, is so small that it does not really affect the results for the longitudinal dispersivity (𝛼1 ≈ 𝛼2 ≈ 𝛼3), it was decided to set that parameter to zero in the Dispersion Package (DMCOEF = 0).

Fired (1975) defined the so-called scale effect of dispersion as the influence of the flow

length on the value of L, explaining that the greater the flow length, the larger the value of longitudinal dispersivity needed to fit the data to the advection-dispersion equation. Afterwards, , different authors came up with empirical equations based on measurements of dispersivity in the field, as shown in equations Eq. 5-22 to Eq. 5-27.

In order to consider such effect for this particular exercise, and considering that the length of the flow path between the river and the BR3 is approximately 95 m, the value of αL was calculated using those equations derived for flow paths shorter than 100 or with no restrictions in terms of length. The results are shown in Table 5-17.

𝛼𝐿 = 0.1𝐿(Lallemand-Barres and Peaudecerf, 1978) Eq. 5-22

𝛼𝐿 = 0.0175𝐿1.46, for 𝐿 ≤ 3500 m (Neuman, 1990) Eq. 5-23

𝛼𝐿 = 0.0169𝐿1.53, for 𝐿 ≤ 100 m (Neuman, 1990) Eq. 5-24

𝛼𝐿 = 0.32𝐿0.83, for 𝐿 > 100 m (Neuman, 1990) Eq. 5-25

𝛼𝐿 = 0.017𝐿1.5, for 𝐿 ≤ 3460 m (Neuman, 1990) Eq. 5-26

𝛼𝐿 = 0.83(𝑙𝑜𝑔𝐿)2.414 (Xu and Eckstein, 1995) Eq. 5-27

When comparing the results presented in Table 5-16 with those in Table 5-17, the scale effect of dispersion became clear, as the values found with data from column tests are one or two order of magnitude lower than those found with the path length from the numerical model, similar to what was previously reported by Fetter (1999).

Evaluation of the Potential for Riverbank Filtration in Colombia

114

Table 5-17 Values for longitudinal dispersivity calculated with different empirical equations

Equation L (m) L (m)

L = 0.1L 95 9.5

L = 0.0175L1.46

95 13.5

L = 0.0169L1.53

95 17.9

L = 0.017L1.5

95 15.7

L = 0.83(logL)2.414

95 4.3

Average 12.2

Taking into account that mechanical dispersion is caused by differences in the fluid velocities within a single pore, or between pores of slightly different size, or because different flow paths have slightly different lengths, Fetter (1999) defines the longitudinal dispersivity at field-scales flow lengths as macrodispersion. The author explains how if the former happens in a flow domain that encompasses a few pore lengths (such as in columns experiments), then the mechanical dispersion in an aquifer can be largely affected by the differences in hydraulic conductivities, and associated velocities, in layers or zones that could be present even in aquifers considered homogeneous.

Acknowledging the importance of field-scale in mechanical dispersion, it was decided to use for this exercise the values of 𝛼𝐿 found with the empirical equations presented in Table 5-17. At first, the average value of 12.2 m was used in order to have an initial idea of the behavior and fate of the compounds in the model domain, but later on such value was changed according to each of the empirical equations as to test the sensitivity of the model to this parameter.

Something similar was done with the ratio αT/αL. At the beginning of the simulation, it was decided to assume that DT was 10% of DL, as suggested by Fetter (1999). Therefore, TRPT was set equal to TRPV, equal to 0.1. Other scenarios where the ratio was changed were simulated as a sensitivity analysis of the solute-transport model with respect to transverse dispersivity.

5.3.2.3 Chemical reaction package

MT3DMS offers the possibility to use one of the following types of sorption: No sorption, linear equilibrium sorption isotherm, Freundlich nonlinear equilibrium isotherm, Langmuir nonlinear equilibrium isotherm, first-order kinetic sorption (non-equilibrium), and dual-domain mass transfer (with or without sorption). It also allows choosing to simulate the radioactive decay or biodegradation.

For this particular example, a linear equilibrium sorption isotherm was chosen, as well as the option to simulate radioactive decay or biodegradation. The parameters the software requires for these two options, for each pesticide, are bulk density of the porous medium, 𝜌𝑏 [ML-3], distribution coefficient, Kd [L3M-1], 1st-order reaction rate for the dissolved phase [T-1], and 1st-order reaction rate for the sorbed phase [T-1].

The bulk density was calculated from the equation presented in Fetter (2004):

Chapter 5 – Pesticides Removal During RBF

115

𝑛 = 100 [1 −𝜌𝑏

𝜌𝑑] Eq. 5-28

Where n is the porosity of the porous medium as a percentage and 𝜌𝑑 is the particle density [ML-3]. According to the same author, for most rocks and soils the particle density is about 2.65 g/cm3, which is the density of quartz.

The distribution coefficient, on the other hand, was found using Eq. 5-29:

𝑅 = 1 +𝜌𝑏𝐾𝑑

𝜃 Eq. 5-29

Where R is the retardation factor found during the column tests and 𝜃 is the porosity of the aquifer.

The results of the bulk density of the porous medium, 𝜌𝑏 , and the distribution coefficient, 𝐾𝑑, are shown in Table 5-18.

Table 5-18 Values of bulk density and distribution coefficient for each pesticide

Pesticide Porosity R b ( kg/m3) Kd (m3

/kg)

Ametryn 0.35 2.4 1723 0.0002845

Atrazine 0.35 1.7 1723 0.0001422

Carbofuran 0.35 1.4 1723 0.0000813

Diuron 0.35 4.2 1723 0.0006502

Finally, since it was not possible to determine individual degradation parameters for the reaction rates of the liquid and sorbed phases, it was decided to assign to the 1st-order reaction rate for the dissolved medium the value of the degradation coefficient µ obtained with the CXTFIT, and a value of 0 for the 1st-order reaction rate for the sorbed phase. This simplification can be seen as appropriate for some pesticide-soil combinations according to Toride et al. (1995).

The parameter μ found through inverse modeling is in fact a combination of the first-order degradation coefficient in the liquid and in the adsorbed phases, i.e., 𝜇𝑙 and 𝜇𝑠, respectively (Toride et al. 1995), as shown in Eq. 5-30:

𝜇 = 𝜇𝑙 +𝜌𝑏𝐾𝑑𝜇𝑠

𝜃 Eq. 5-30

Since we considered that there was only degradation in the liquid phase, the equation above is reduced to 𝜇 = 𝜇𝑙. As explained before 𝜇 is a dimensionless value that has to be converted into a dimensional one in order to use it in MT3DMS. This can be done by applying Eq. 5-31, as follow:

𝜇 =𝜇𝐸𝑣

𝐿 Eq. 5-31

Evaluation of the Potential for Riverbank Filtration in Colombia

116

where 𝜇 is the value of the degradation [T-1], 𝜇𝐸is the value found through inverse modeling [dimensionless], 𝑣 is the seepage velocity from the column experiment 3-1 [LT-1] (0.037 cm/min), and 𝐿 is the length of the column [L], i.e. 50 cm.

Table 5-19 presents the values of the 1-st order degradation coefficient of the liquid phase for each pesticide that were inserted into the chemical reaction package of MT3DMS.

Table 5-19 Value of the first-order degradation coefficient for each pesticide

Pesticide Edimensionless s

-1

Ametryn 0.159 1.96E-06

Atrazine 0.023 2.84E-07

Carbofuran 0.288 3.55E-06

Diuron 0.267 3.29E-06

5.3.2.4 Other parameters required by MT3DMS

ICBOUND: Chiang (2005) suggests using the value of 1 for an active concentration cell, a value of -1 for a constant-concentration cell, and a value of 0 for an inactive concentration cell. In this particular case a value of 1 was applied to the entire model domain.

Time: For the contaminant transport model a transient simulation flow type was selected, in order to simulate an accidental-spill scenario. The simulation time was divided into three (periods), each one of different duration, as shown in Table 5-20. Period number 2 was when the substances were introduced to the system.

Table 5-20 Time periods defined for the transient simulation

Period Length (s) Length (days) Time steps

1 3.2E+07 369.2 1

2 100000 1.16 1

3 6.3E+07 729.2 1

Initial Concentration: At the beginning of a transport simulation, MT3DMS requires the initial concentration of each active species at each active cell, i.e. ICBUND>0 (Chiang, 2005). It was assumed that at the beginning of the transport simulation the aquifer was free of any of the contaminants of interest and, therefore, the initial concentration value was 0 for each one of the pesticides.

Sink/Source Concentration: This menu offers the possibility to choose from point sources (wells, general head boundary cells, fixed-head cells, rivers and streams), spatially distributed sources (recharge), or sinks (evapotranspiration) of contaminants (Chiang 2005). In this case, a river source was chosen and it was supposed an accidental spill on the river with a concentration of 30 grams of contaminant per liter of water. Such concentration was applied during the second period of the simulation time, along the section of the river where, according to Figure 5-25, the water that entered production well BR3 came from.

Chapter 5 – Pesticides Removal During RBF

117

5.3.2.5 Results and analysis

Three (3) different scenarios were considered during the contaminant transport simulation in order to test the sensitivity of the model to longitudinal and transverse dispersivity, as explained in Table 5-21.

At first, the transport model for all four pesticides was run using αL= 12.2 m, i.e. the average of the values for longitudinal dispersivity calculated by different empirical (Table 5-17), and a ratio αT/αL equal to 0.1, in order to have an initial idea of the behavior and fate of the compounds in the model domain.

Table 5-21 Scenarios considered during contaminant transport simulation

Scenario αL (m) αT/αL Decay

1 12.2 0.1 Yes

2 Varying 0.1 Yes

3 12.2 Varying Yes

The plume of contaminant and the BTC of each pesticide at borehole BR3 at the end of the simulation period are shown in Figure 5-26. Table 5-22 presents the arrival time and concentration of each pesticide to well BR3 and the time it took for the maximum concentration to reach the borehole, as well as the value of such concentration. The time shown in the table was measured from the moment the spill occurred.

Table 5-22 Results of simulation scenario 1

Pesticide Arrival time (days) Arrival C (mg/l) Time of max. C (days) Max. C (mg/l)

Ametryn 5.8 1.5E-18 93 1.6.E-03

Atrazine 5.8 7.0E-13 139 4.0

Carbofuran 5.8 6.6E-13 41 3.5.E-05

Diuron 11.6 2.9E-18 122 2.4.E-05

From the contaminant plumes can be seen that the center of mass of Atrazine has almost completely left the active domain at the end of the simulation time, whereas those of Diuron and Carbofuran have not even reached the well, indicating that Atrazine is the contaminant that moves fastest than the rest. Diuron, on the contrary, is the slowest of the four, which presents correlation with the retardation factor, R, found with the column experiments, being the one that takes longer to reach BR3 from the time of spilling (almost 12 days). No matter how long it takes for the contaminant to arrive at the well, there is a great decrease in the maximum concentration compared to the initial in the river (30 g/l).

Also, even though the ratio 𝛼𝑇/𝛼𝐿 was the same, the plume of Carbofuran presents the smallest dispersion in the direction perpendicular to the flow, whereas Atrazine displays the highest of all the studied pesticides. On the other hand, the shape of the BTC is showing the spreading of the solute slug in the subsoil, i.e. the thinner the BTC, the narrower the contaminant plume. Under that consideration, the pesticide that shows a greater spreading in the direction perpendicular to the flow is Atrazine, and the one with the lowest spreading is Carbofuran.

Evaluation of the Potential for Riverbank Filtration in Colombia

118

Figure 5

-26 B

TCs an

d co

ntam

inan

t plu

mes fo

r each co

ntam

inan

t (αL = 1

2.2

m an

d a α

T = 0.1

αL )

Chapter 5 – Pesticides Removal During RBF

119

When taking a look at the maximum concentrations in the BTCs, the effect of the degradation coefficient, μ, is made clearer. Thus, Atrazine, which was the pesticide with the lowest value of μ (0.023), is the one with the highest concentration at the center of the plume, whereas Carbofuran and Diuron, with the highest degradation coefficients (0.288 and 0.267, respectively) are the components with the lowest concentrations.

The second scenario consisted on keeping constant the relationship between the transverse and the longitudinal dispersivities (αT = 0.1αL), and changing the value of αL according to the equations shown in Table 5-16. Figure 5-27 to Figure 5-30 show the results for each pesticide, with αL values of 4.3, 9.5, 13.5 and 15.7 m. In this case, as expected, the spreading in the flow direction becomes bigger as the value of αL increases, as well as the maximum concentration in the center of the plume, this behavior being more marked for Atrazine which is the contaminant that moves fastest.

The differences in BTCs, when changing the values of αL, is clearer in Figure 5-31, where the all the BTCs for each pesticide are shown in the same graph. In most cases the time from spilling associated to the maximum concentration remains relatively constant, except for Atrazine, where it becomes evident that lower values of longitudinal dispersivity represent, not only lower concentrations, but also longer travel times.

Finally, Figure 5-32 presents, for each pesticide, the graph of maximum concentration against time from the spilling event (in days) for the different values of longitudinal dispersivity. The values respond to an exponential behavior which permits to conclude that towards the lower values of αL, there will be a threshold where the maximum concentration asymptotically reaches a minimum value, whereas the time from spilling associated to such concentration will continue to increase. On the other side of the spectrum, i.e. for high values of αL, the maximum value of concentration will keep on increasing but the time from spilling will remain somehow constant.

The third scenario that was evaluated consisted on leaving constant the longitudinal dispersivity, while varying the ratio 𝛼𝑇/𝛼𝐿. This procedure was done for each pesticide, for the lowest and highest values of 𝛼𝐿 (4.3 and 17.9 m), and the ratios used were 1, 0.5, 0.1, 0.01, and 0.001. Figure 5-33 and Figure 5-34 show the BTCs for the four pesticides at different ratios of transverse and longitudinal dispersivity.

As it can be seen, the behavior is pretty much the same for either value of 𝛼𝐿 when it comes to the maximum concentration reached at the extraction well, i.e. the concentration increases with lower values of 𝛼𝑇/𝛼𝐿 until certain threshold is reached where the increment in concentration is insignificant.

This behavior is clear for all pesticides for the ratio values of 0.01 and 0.001, so it could be assumed that it will continue happening for even lower ratios. On the other hand, the difference in concentrations between the two graphs of the same pesticide is related to the values of longitudinal dispersivity as explained before.

Evaluation of the Potential for Riverbank Filtration in Colombia

120

Figure 5

-27 B

TCs an

d co

ntam

inan

t plu

mes fo

r Am

etryn fo

r differen

t values o

f αL , an

d α

T = 0.1

αL

Chapter 5 – Pesticides Removal During RBF

121

Figu

re 5

-28

BTC

s an

d c

on

tam

inan

t p

lum

es f

or

Atr

azin

e fo

r d

iffe

ren

t va

lues

of

αL,

an

d α

T =

0.1

αL

Evaluation of the Potential for Riverbank Filtration in Colombia

122

Figure 5

-29 B

TCs an

d co

ntam

inan

t plu

mes fo

r Carb

ofu

ran fo

r differen

t values o

f αL , an

d α

T = 0.1

αL

Chapter 5 – Pesticides Removal During RBF

123

Figu

re 5

-30

BTC

s an

d c

on

tam

inan

t p

lum

es f

or

Diu

ron

fo

r d

iffe

ren

t va

lues

of

αL,

an

d α

T =

0.1

αL

Evaluation of the Potential for Riverbank Filtration in Colombia

124

Figure 5

-31 B

TCs fo

r each p

esticide fo

r differen

t values o

f αL , an

d α

T = 0.1

αL

Chapter 5 – Pesticides Removal During RBF

125

Figu

re 5

-32

Max

imu

m c

on

cen

trat

ion

of

con

tam

inan

ts v

s. t

ime

fro

m s

pill

ing

for

dif

fere

nt

valu

es o

f α

L

Evaluation of the Potential for Riverbank Filtration in Colombia

126

Figure 5

-33 M

aximu

m co

ncen

tration

vs. time fro

m sp

illing fo

r Am

etryn an

d A

trazine at d

ifferent valu

es of α

T /αL

Chapter 5 – Pesticides Removal During RBF

127

Figu

re 5

-34

Max

imu

m c

on

cen

trat

ion

vs.

tim

e fr

om

sp

illin

g fo

r C

arb

ofu

ran

an

d D

iuro

n a

t d

iffe

ren

t va

lues

of

αT/

αL

The effects of the different ratios 𝛼𝑇/𝛼𝐿 on the contaminant plume are shown in Figure 5-35 (in this figure only the graphs for the extreme values of the ratio transverse/longitudinal dispersivity are presented, i.e. 0.001 and 1).

Evaluation of the Potential for Riverbank Filtration in Colombia

128

Figure 5

-35 C

on

tamin

ant p

lum

es for A

metryn

at αL =4

.3 an

d 1

7.9 m

, and

αT/ α

L = 0.00

1 and

0.1

Chapter 5 – Pesticides Removal During RBF

129

The results for Ametryn were selected as an example of what appears to be a steady behavior for all other pesticides, i.e. the higher the ratio, the broader the contaminant plume in the direction perpendicular to the flow.

After analyzing scenarios two and three, it is clear that the fate of contaminants in a porous media is very sensitive to the values of longitudinal and transverse dispersivity. Therefore, in order to obtain more precise results in a study of this type, it is recommendable to conduct experiments that would help determine reliable values of those two parameters.

The last scenario considered in the contaminant transport model consisted on evaluating the effect of degradation (μ) in the fate of the pesticides. This was achieved by not considering radioactive decay or biodegradation in the simulation, and comparing the results with those when degradation was, in fact, taken into consideration. In this case, an average value of longitudinal dispersivity was chosen for the simulation (𝛼𝐿 = 12.2 𝑚), as well as a ratio 𝛼𝑇/𝛼𝐿 = 0.1.

5.4 CONCLUSIONS AND RECOMMENDATIONS

5.4.1 COLUMN EXPERIMENTS AND INVERSE MODELING

The fate of five active ingredients of some common pesticides used in Colombia was determined through a series of columns experiments performed at the Geohydraulics laboratory at the University of Applied Sciences in Dresden, Germany (HTW-Dresden). The compounds Ametryn, Atrazine, Carbofuran, Diuron and Propanil were chosen based on its availability and possibility of being analyzed at the Institute of Water Chemistry at the Technical University in Dresden (TU-Dresden).

The experiments were carried out in two stainless steel columns filled with sediments from the bed of the river Lößnitztal in East Germany, which were classified as poorly graded sand based on the results of the sieve analysis. The basic parameters such as porosity, pore volume, and residence time were determined through a series of tracer tests performed at two different temperatures (10 and 20oC), which were also the temperatures used for the contaminant column experiments. Those values were chosen as they are commonly found temperatures in Colombian rivers. The higher increment in pore volumes in one of the columns, related to the different temperatures, was in concordance to the bigger increase in porosity compared to the other column. This could be due to a rearrangement of the sediment inside the column and the progressive removal of the finer sediment with time which, in turn, represented an increment in the porosity.

In total six column experiments were preformed, three in each column, with a mix of Elbe water and all the pesticides, prepared at the Institute of Water Chemistry at TU-Dresden. This mixture was later dissolved in a container with Elbe water at the Geohydraulics laboratory at HTW-Dresden and slowly added to the columns using a membrane pump. Samples of the inlet and outlet water were taken at different time intervals. One important outcome from the experiments is the fact that Propanil rapidly transformed

Evaluation of the Potential for Riverbank Filtration in Colombia

130

into the metabolite 3,4 Dichloroanaline and was almost completely missing in the water samples, both from the inlet and the outlet.

In order to determine the transport parameters (retardation factor, first-order degradation coefficient, and dispersion coefficient) for each pesticide, the results from the column experiments were inputted into the CXTFIT code included in the freeware STANMOD. This was done by solving the inverse problem and selecting the conventional equilibrium Advection Dispersion Equation (ADE) as the transport model. Also, a third-type inlet resident concentration was chosen, as well as zero initial concentration, and a zero production coefficient.

The highest values of dispersion coefficient and retardation factor were those of Diuron, whereas the lowest were the ones for Carbofuran. Therefore, Diuron was the contaminant that remained the longest in the soil and Carbofuran was the one that traveled fastest of them all. When it came to the first-order degradation coefficient, the highest values were for Carbofuran, and the lowest for Atrazine.

Based on the shapes of the breakthrough curves (BTCs), which are directly related to the dispersion coefficient and the degradation factor, Diuron presented the highest skewness, which coincided with its high values of dispersion coefficient, while Atrazine was the component whose final relative concentration (C/Co) wss closer to 1, indicating a very low degree of degradation. Despite this last observation, degradation is not easily interpreted from the BTCs and its values are not consistent from experiment to experiment, making it the parameter that shows a more unpredictable behavior of the three calculated with CXTFIT.

5.4.2 GROUNDWATER FLOW AND CONTAMINANT TRANSPORT MODELING

With the results from one of the runs with CXTFIT, a transient-state contaminant transport modeling was implemented using the MT3DMS code contained in the freeware Processing MODFLOW for Windows (PMWIN). This model was based on a steady-state groundwater flow model developed in MODFLOW, also part of the PMWIN software, for the Riverbank Filtration site in Lößnitztal, Germany, where the sediments for the column experiments were collected.

The groundwater flow model consisted of an unconfined, one-layer aquifer with a hydraulic conductivity of 1.3 x 10-4 m/s and a porosity of 0.35, which was the value obtained from the tracer test associated to the column experiment E 3-1, whose parameters were the ones used in the contaminant transport model. As boundary conditions, constant head was selected for the east and west limits of the active area, whereas a RIVER boundary was used in the south. The northern frontier was set as no-flow, and recharge was applied to the entire model domain.

Two pumping scenarios were chosen, one for calibration and one for the simulation that was finally going to be used as the basis for the contaminant transport model. The data used to estimate the pumping rates, as well as the values for the initial heads at the boreholes was supplied by the Lößnitztal waterworks.

Chapter 5 – Pesticides Removal During RBF

131

After the steady-state flow simulation, the flowlines for the groundwater were determined with the use of backward particle tracking through the software PMPATH (contained in the PMWIN). A retardation factor of 1 was used for clean water (without pesticides), whereas for each contaminant the value of retardation found with CXTFIT was introduced into the software. As expected, the higher the retardation factor, the longer it took the water to reach the pumping well. Thus, water free from pesticides was the first one to appear at the extraction well, whereas Diuron –the compound with the highest retardation factor- was the last one to arrive, preceded by Ametryn, Atrazine, and Carbofuran. These results completely disregard the effect of dispersion and degradation because PMPATH only takes into account transport by advection.

In order to simulate the effect of dispersion and degradation, as well as advection, on the fate and transport of the four contaminants, a simulation with MT3DMS was implemented considering the hypothetical case of an accidental spill on the river. The solution scheme selected for the advection package was the Upstream Finite Difference, selected after checking with the Peclet number that the problem was not advection-dominated. The longitudinal dispersivity (𝛼𝐿) used in the dispersion package was calculated by different empirical equations, and the effect of chemical diffusion was not considered (effective molecular diffusion coefficient, D*, was set equal to zero). Lastly, for the chemical reaction package, a linear equilibrium sorption isotherm was chosen, as well as the option to simulate degradation. In this case, the values of retardation factor (R) and degradation coefficient (μ) found through CXTFIT were used to calculate the distribution coefficient (Kd) and the 1st-order degradation coefficient in the liquid phase (since the one for the solid phase was assumed to be zero), respectively.

Three different scenarios where simulated. The first one consisted on using an average value of 𝛼𝐿 and a constant ratio of transverse-longitudinal dispersivity (𝛼𝑇/𝛼𝐿). For the second scenario, the value of 𝛼𝑇/𝛼𝐿 was kept constant but the value of αL changed according to the different empirical equations. Finally, the third scenario was simulated by leaving 𝛼𝐿constant and varying the ratio 𝛼𝑇/𝛼𝐿. The results of the numerical model showed a great reduction in the highest concentration reaching the well at the end of the simulation period compared to the initial concentration in the river.

In general, the results from the column experiments and the CXTFIT modeling coincide with those of the contaminant transport simulation becoming clear that the fate of contaminants in porous media is very sensitive to the values of dispersivity, as well as degradation. Therefore, in order to obtain better and more precise results on a study of this type it is recommendable to conduct field experiments that would help determine more reliable values for dispersivity both, longitudinal and transverse. Also, it is necessary to determine the fraction of the first-order decay coefficient that belongs to the dissolved and to the sorbed phases.

The results obtained in this work strengthen the hypothesis that riverbank filtration is efficient, if not on the complete removal of pesticides present in the river, on the reduction of the concentration of, at least, some pesticides. However, more research need

Evaluation of the Potential for Riverbank Filtration in Colombia

132

to be done, especially in what has to do with metabolites, which can be more persistent and, therefore, represent a higher threat to the ecosystems than their parent compounds.

5.5 ACKNOWLEDGMENTS

This chapter would not have been possible without the guidance of Prof. Thomas Grisheck from the University of Applied Sciences (HTW) in Dresden, Germany, who also allowed the author the use of the Geohydraulics Laboratory at this Institution to run the column tests. Also, a special thanks to Dr. Hilmar Boernick from the Institute of Water Chemistry at TU-Dresden, who generously performed the analysis of the water samples from the column tests and offered his support during the time of the experiments and result analysis.

133

6 FINAL CONCLUSIONS AND RECOMMENDATIONS

6.1 CONCLUSIONS

Riverbank filtration is a water treatment technology that involves extracting water from a river through a well in the adjacent alluvial plain and utilizes the aquifer as a natural filter for the removal or reduction of contaminants through a series of physical, chemical and biological processes. This technique, whose efficiency degree depends on specific site conditions, has been successfully used for over a century in European countries and more recently in the United States, where it has proved to be equally, if not more, efficient than conventional water treatment techniques, and economically more attractive than the latter.

Despite of all the benefits that RBF provides, its application in Colombia has not been reported even though there are apparently appropriate conditions for its implementation, and could significate an alternative in water treatment for that half of the population who currently receives water that is not safe for human consumption. Therefore, this doctoral thesis was carried out with the main objective of evaluating the potential of RBF in Colombia, and the most important conclusions and recommendations of this work are presented below.

From the literature review, it became clear that there is still a lot about the processes involved during RBF that is unknown, especially because the technique was used in an empirical way for so many years before it started being studied systematically. This is particularly true in tropical, sub-tropical and desert countries where the potential of RBF is just being explored, but the research has yielded so positive results that has encouraged the hypothesis that Colombia also has the necessary characteristics as for RBF to be applied with success. This thesis then focused on three main subjects: potential RBF sites in Colombia, water quality and pesticides removal.

6.1.1 Potential RBF Sites in Colombia – A Geomorphological Analysis

As a first approach to assess the potential of RBF in Colombia, a geomorphological analysis involving downstream hydraulic geometry was done to some stream-reaches of pre-selected sites in Colombia and of currently operating RBF wellfields in Europe and the United States. The goal was to find easy-to-measure elements that could be used in prefeasibility studies for new RBF sites, and to determine whether the Colombian rivers had the necessary characteristics as to be considered suitable for the application of the technique.

The analysis and comparison of the planform features like drainage area, reach slope, average width, curvature radius, sinuosity and bed wavelength allowed drawing conclusions on the most adequate locations for RBF sites and how some variables

Evaluation of the Potential for Riverbank Filtration in Colombia

134

appeared to prevail for the existing RBF sites, and should be considered when evaluating a new location. Moreover, the results showed that most of the selected Colombian reaches had geomorphological similitudes with the streams from abroad, supporting once more the hypothesis of the potential for the implementation of RBF in the country.

On the other hand, the higher sediment load of the Colombian rivers compared to those in Europe and the United States could point at mechanical clogging as a possible limitation to the efficiency of RBF in the removal of contaminants. However, this study showed that the self-cleaning capacity of the Colombian rivers, determined using specific stream power, could be enough to erode the clogged layers and restore the hydraulic connection between the river and the aquifer.

6.1.2 RBF and Water Quality at Two Sites in Antioquia

At two sites in the department of Antioquia, where the direct interaction between the rivers and the aquifers had been previously established through numerical modeling, the efficiency of RBF in the removal of the most common substances in the water was assessed from the point of reference of the Colombian legislation on drinking water standards. In order to accomplished that, several water samples from the rivers and the wells were analyzed for different physical, chemical and microbiological parameters, and the results were compared to the legal maximum accepted values for water for human consumption. At the end, the portion of bank filtrate in the wells water was estimated using chloride as a natural tracer.

Even though both sites have different climatological, hydrological, and hydrogeological characteristics, in general terms the quality of the water extracted from the wells complied with the Colombian standards, except for the microbiological component, and significant improvement in parameters such as turbidity, total suspended solids, total coliforms and E.coli was observed at both sites. One of the locations also exhibited an important decrease in the concentration of iron, chloride and sulphate. These results show the potential that RBF has in the pretreatment of drinking water, and encourage the idea that processes like coagulation/flocculation, sedimentation and physical filtration, which are usually employed at the type of water treatment plants operating in the country, could be replaced by this technology, thus reducing costs of treatment and, hence, the final price of the cubic meter of water for the user.

However the good results obtained here, it is clear that not all the parameters included in the legislation were measured and their presence in the bank filtrate could represent the need for further treatment before disinfection. It is necessary then knowing in advance the type of particular contaminants expected to be found in the river, so their presence in the bank filtrate is also measured.

The results of the hydrochemical analysis also showed a very close and direct relationship between the quality of the water both in the river and the aquifer with the local geology and the climate of the area. Specifically, the strong influence of the precipitation events related to La Niña conditions, dominant during the second half of the sampling campaigns,

Chapter 6 – Final Conclusions and Recommendations

135

was observed at both sites. The effects of this climatological condition were also made clear in the portion of river-borne water in the bank filtrate which was estimated as the degree of mixing between the water from the river and the well using chloride as a tracer.

6.1.3 Pesticides Removal

Finally, the potential of RBF in the reduction of organic compounds, most specifically pesticides, was approached from the stand point of laboratory column experiments and the numerical modeling of contaminant transport.

Five common pesticides used in Colombia were chosen to be studied using column tests with sediment collected from the riverbed of an RBF site in Germany. Afterward, the equilibrium advection dispersion equation was solved using an inverse approach and variables such as retardation factor, first-order degradation coefficient, and dispersion coefficient were assessed for all the pesticides but one, which rapidly and almost completely transformed into its metabolite and whose final concentration in the outlet water was practically negligible.

The results from the inverse modeling were then introduced into a steady-state groundwater flow model to perform a contaminant transport simulation of an accidental spill under three different scenarios, where the values for longitudinal and transverse dispersivity, as well as their ratio, were varied. Parameters such as time of travel of the contaminants from the river to the well, as well as final concentration of the compounds in the borehole, were determined.

At the end, it was found that the fate of contaminants in porous media is very sensitive to the values of dispersivity and degradation, and therefore it is important to determine beforehand the decay coefficient of both the dissolved and sorbed phases of the compounds, and the type of degradation (biotic or abiotic) to be expected. Also, field experiments should be carried out in order to determine more precisely the longitudinal and transverse dispersivity.

From the column experiments and numerical modeling of contaminant transport can be concluded that, for the same hydrogeological setting, the persistence of pesticides is not constant but varies depending on the properties of each compound. Therefore, RBF could be a potential solution in the removal of some pesticides from surface water, or at least for the reduction of the initial concentrations in the river. Nevertheless, in locations where pesticides are present in the river water, bank filtrate will more probably require additional treatment before consumption.

6.2 RECOMMENDATIONS FOR FURTHER RESEARCH

The results obtained in this thesis are very encouraging about the potential to effectively implement the technology of riverbank filtration in the pretreatment of water for human consumption in Colombia. However, these are just first approximations to the subject and more research needs to be done in order to better understand the processes involved in the removal of contaminants, the identification of possible limitations of the technique,

Evaluation of the Potential for Riverbank Filtration in Colombia

136

and determination of the geomorphological settings more appropriate to establish new RBF sites.

The recommendations for further research listed below summarize some of the challenges identified during the development of this work and several interesting topics that are believed to deserve more investigation, with the purpose of increasing the knowledge around RBF, not only locally but also to a global level. Nonetheless, given the fact that serious studies on RBF are relatively recent compared to the time the technique has been used and, therefore, the research needs are almost innumerable, this recommendations are just what the author believes should be the next step to be taken in Colombia with the aiming of improving the evidences on the efficiency of RBF and finally getting it implemented as an alternative in the treatment of drinking water.

Hydraulic geometry, through the use of geo-processing tools, proved in Chapter 3 to have the potential to be used when evaluating a new site for the implementation of RBF. However, in order to obtain more conclusive results, further research on this area needs to be done, as indicated below.

- It is necessary to broaden the database on existing RBF sites and perform a hydraulic geometry analysis similar to the one carried out in this thesis, with especial attention to parameters like bankfull width, curvature radius and bankfull discharge. The more data, the better the definition of general guidelines on the geomorphological conditions that need to be considered during site selection.

- The geomorphological classification scheme used in this thesis, despite its facility of use, has the limitation that it only takes into account the stream-reaches planforms but not the bedforms that can be present inside the channel. Therefore, an interesting research would be to study the bedforms in stream-reaches where RBF is currently being used and determine if they can be an indicator on the feasibility of a site for the technique to be implemented with success.

- The use of specific stream power demonstrated to be a promising alternative in the indirect determination the erosion potential of a river, with the advantage that the involved parameters are very easy to estimate. Nonetheless, it has the disadvantage that it does not consider the resisting forces of the riverbed and banks, or the availability of the sediments, focusing only on the river driving forces. For this reason, it would be very constructive to conduct field experiments in different streams to determine the shear stress of the riverbed and compare the results with the specific stream power of the same reaches, in order to find relationships of the two variables, aiming at finding a solution for the evaluation of the scouring potential of a river to counteract for possible clogging, in places where there is no information on parameters like flow depth, sediment size, etc.

- Scouring of the bed and banks of the river can also become a limitation of RBF due to the constant removal of the clogging layer which can represent a reduction in the efficiency of the technique. Therefore, it would be advisable to study the self-cleaning potential of the rivers in terms of recurrent floods, their magnitude, duration, and

Chapter 6 – Final Conclusions and Recommendations

137

intensity at places where RBF is being used successfully. The results obtained would be very helpful when evaluating new sites for the implementation of the technology.

- At the end, the goal would be to develop a step-by-step methodology that allows an accurate selection of future RBF sites, based on geomorphological characteristics of the streams easily to assess through the use of geospatial data, aerial images, satellite images and, in general, readily available geo-processing tools.

One of the hypotheses of this thesis was that RBF was being used in Colombia in an empirical or unintentional way, and just as this was proved to be right with the two studied sites in Antioquia, certainly there are other places where something similar is happening. It would be interesting then to search for some well, or wells, where RBF is suspected to be happening and carry out a similar research as the one presented here in Chapter 4, with the following characteristics:

- The well should have been adequately constructed, i.e. not a dug-well, built with the necessary technical standards, proper documentation on the well characteristics (length, diameter, water level at the moment of construction, etc.), the stratigraphic column of the alluvial plain, and preferably with the data from pumping tests and well logs.

- The selected river should be one for which there is available information on discharge, depth, water level, sediment load (both bed and suspended), and water quality, as well as records of precipitation and temperature in the area.

- If there is not an available numerical flow model, this should be done to establish the degree of interaction between the river and the well, and to indirectly determined the portion of river-borne water that can be expected in the bank filtrate. This model should preferably be in a transient-state to be able to incorporate the variations in recharge by precipitation and the changes in the river water level associated to different climatological conditions.

- Observation wells should be built between the river and the production well in order to better monitor the changes in water quality and level.

- The sampling campaigns should last minimum one hydrological year, at least on a bimonthly basis, and comprise not only the parameters included in the Colombian legislation for drinking water standards but any important contaminants expected to be present in the river or the aquifer as a product of either anthropogenic activities or natural processes according to the geology of the site.

- Water parameters like temperature, pH, electrical conductivity and dissolved oxygen should always be measured on-site in the water from the river and from the monitoring and extraction wells.

- In order to determine the portion of river water in the well using chloride as a tracer, it cannot be forgotten to also sample a well that is far enough from the river that can be considered to be extracting only groundwater (this can be achieved with the numerical modeling) as to establish the background concentration of the ion.

Evaluation of the Potential for Riverbank Filtration in Colombia

138

- It would also be interesting to examine other environmental tracers that could be used instead of chloride to determine the degree of mixing between the river water and the groundwater, or to complement the investigation with the use of Radon-222 to establish residence times and flow water velocities and compare them with the results of the numerical modeling.

- It is advised to install some kind of devise that allows measuring the river stage every time that water samples are taken, to be able to make comparisons with the changes in the piezometric heads in the well.

- An analysis of precipitation and temperature, preferably on a daily basis, should provide to be useful when interpreting the results of the water quality analysis and relate them to the climatological conditions.

- Finally, it would be very beneficial if the well is located near a water treatment plant that directly takes the water from the same river, in order to compare both treatment techniques and draw some more definite conclusions about the efficiency of RBF in the removal of contaminants.

The potential of RBF in the removal of common pesticides used in Colombia was also explored through column experiments carried out with sediments and water from two different rivers in Germany, and numerical solute transport modeling of a German RBF site. Therefore it is suggested to conduct a similar exercise but under local conditions, with the following recommendations:

- The sediments and water for the column tests should come from the same river. This will allow a better interpretation of the results based on the actual conditions that influence the fate of the contaminants.

- The water from the river should be tested for pesticides in order to determine if there is any background concentration that needs to be considered. Some knowledge of the type of agricultural activities in the watershed area might give some insight on what organic compounds to expect.

- Tracer tests should be carried out before the actual experiments start and also every time a test is performed, with the objective of monitoring possible changes in the effective porosity as a result of the progressive removal of the finer sediments.

- The pumping rate should be adjusted until the flow velocity in the column is close to the one expected on the field, in order to simulate an equivalent residence time. Another possibility is to use different flow velocities (simulating different pumping rates in the extraction well) and evaluate the impact of this variable in the removal of the pesticides.

- It would be interesting if the concentration of the contaminants in the input water could be close to that expected to find in the river, either under normal hydrological conditions or during extreme precipitation events (like those that occurring when La Niña conditions are being experimented) when run-off from the agricultural fields is expected to be higher, so the influence of climate in the process can be determined.

Chapter 6 – Final Conclusions and Recommendations

139

This however, would depend on the available analytical methods and their detection limits.

- In cases where the pesticide being analyzed is expected to transform into one or more metabolites, these should also be considered because most of the time such products remain for a longer time in the aquifer, representing a higher risk to the environment, and some of them can be even more toxic than their parenting compounds.

- Given the sensitivity of the contaminant transport model to longitudinal and transverse dispersivity, it would be very valuable to conduct some field tests to determine more precise values of these parameters and the ratio αT/αL to obtain a more real simulation of the mechanical dispersion on the site.

- Sorption and degradation are other essential variables intervening in the fate and transport of contaminants in the subsurface and, consequently, a correct estimation of their numerical values is important. Therefore, it is recommended that laboratory tests are run in order to determine parameters such as the bulk density of the porous media, concentration of the contaminants in the sorbed and dissolved phases, and the type of degradation (chemical or biological) and reaction rates to be expected to occur in both phases. In this line of research, interdisciplinary work with chemists and microbiologists would be extremely useful.

- Lastly, considering that in Colombia the studies on contaminant transport in the subsoil are still very seldom, and aiming at improving the knowledge on the numerical modeling of this kind of processes, an interesting topic of investigation would be to conduct a more detailed exploration on the existing codes, perform simulations using different solution schemes for the advection, dispersion and reaction terms in the ADE equation, and comparing the results in order to find the advantages or disadvantages among them. Furthermore, the implications of the heterogeneity of the porous media should be considered.

The recommendations presented above are just some examples of work that can be done in order to improve the knowledge behind riverbank filtration. Furthermore, given the fact that the study field of RBF is relatively recent, almost any investigation around it will be more than welcome. However, the ultimate wish of the author is that one day a pilot test RBF site can be set up in Colombia from the very beginning, starting with the selection of the location based on geomorphological information, and where interdisciplinary work is carried out with the objective of answering questions related to the processes that govern the removal of contaminants through riverbank filtration.

141

7 BIBLIOGRAPHIC REFERENCES

Abdalla, F.A. and Shamrukh, M. 2011. Riverbank filtration as an alternative treatment technology: AbuTieg case study, Egypt. In: Ray C and Shamrukh M (eds.) Riverbank Filtration for Water Security in Desert Countries. Springer, NATO Science for Peace and Security Series, pp. 255 – 268.

Agertved, J., Rugge, K. and Barker, J.F. 1993. Transformation of the herbicides MCPP and Atrazine under natural aquifer conditions. Groundwater, 30 (4): 500 – 506.

AGUA Ltda. 2000. Construcción de pozo municipio de Guarne, Antioquia – OMYA Colombia. Memoria Técnica. Asesorías Geológicas AGUA Ltda. Medellín.

Alvarez, E. and Gonzalez, H. 1978. Geología y geoquímica del cuadrángulo I-7 Urrao. Informe 1761. INGEOMINAS. Bogota. 347 p.

Appelo, C.A.J. and Postma, D. 2005. Geochemistry, groundwater and pollution. 2nd Edition. CRC Press. 649 p.

Aronino, R., Dlugy, C., Arkhangelsky, E., Shandalov, S., Oron, G., Brenner, A. and Gitis, V. 2009. Removal of viruses from surface water and secondary effluents by sand filtration. Water Research. 43: 87–96.

Aseltyne, T.A., Rowe, H.D. and Fryar, A.E. 2006. Stable isotopic fingerprint of a hyporheic-hypolentic boundary in a reservoir. Hydrogeology Journal, 14: 1688 – 1695.

ASTM. 2006. Standard Practice for Classification of Soils for Engineering Purposes (Unified Soil Classification System). Designation: D 2487-06. ASTM International. United States. 12 p.

Baluch, H.U., Somasundaram, L., Kanwar, R.S. and Coats, J.R. 1993. Fate of major degradation products of atrazine in Iowa soils. Journal of Environmental Sciences and Health B28 (2), 127–149.

Baskaran, S., Ransley, T., Brodie, R.S. and Baker, P. 2009. Investigating groundwater-river interactions using environmental tracers. Australian Journal of Earth Sciences, 56 (1): 13 – 19.

Baveye, P., Vandevivere, P., Hoyle, B.L., Deleo, P.C. and De Lozada, D.S. 1998. Environmental impact and mechanisms of the biological clogging of saturated soils and aquifer materials. Critical Reviews in Environmental Science and Technology, 28 (2): 123–191. 1998.

Bertin, C. and Bourg, A.C.M. 1994. Radon-222 and chloride as natural tracers of the infiltration of river water into an alluvial aquifer in which there is significant river/groundwater mixing. Environmental Science and Technology, 28: 794 – 798.

Evaluation of the Potential for Riverbank Filtration in Colombia

142

Bourg, A.C.M. and Bertin, C. 1993. Biogeochemical processes during the infiltration of river water into an alluvial aquifer. Environmental Science and Technology, 27: 661 – 666.

Bourg, A.C.M. and Bertin, C. 1993. Biogeochemical processes during the infiltration of river water into an alluvial aquifer. Environmental Science and Technology, 27: 661 – 666.

Buchanan, I., Liang, H.C., Liu, Z., Razaviarani, V. and Rahman, Md.Z. 2010. Pesticides and Herbicides. Water Environment Research, 82 (10): 1594 – 1693.

Caldwell, T.G. 2006. Presentation of data for factors significant to yield from several riverbank filtration systems in the U.S. and Europe. In: Hubbs SA (ed.) Riverbank Filtration Hydrology – Impacts on System Capacity and Water Quality. Springer, Nato Science Series IV, Vol. 60, pp. 299–344.

Chen, X.H. 2001. Migration of induced-infiltrated stream water into nearby aquifer due to seasonal ground water withdrawal. Ground Water, 39 (5): 721 – 728.

Chiang, W-H. 2005. 3D-Groundwater Modeling with PMWIN - A Simulation System for Modeling Groundwater Flow and Transport Processes. 2nd Edition. Springer. 414 p.

Chiang, W-H. and Kinzelbach, W. 1994. PMPATH – An advective transport model for Processing Modflow and Modflow. Geological Survey of Hamburg, Germany.

Chiang, W-H. and Kinzelbach, W. 1998. Processing Modflow – A simulation system for modeling groundwater flow and pollution. User’s guide. 334 p.

Constantz, J. 1998. Interaction between stream temperature, streamflow, and groundwater exchanges in alpine streams. Water Resources Research, 34 (7): 1609 – 1615.

Constantz, J., Cox, M.H. and Su, G.W. 2003. Comparison of heat and bromide as ground water tracers near streams. Ground Water, 41 (5): 647 – 656.

Constantz, J., Su, G.W. and Hatch, C. 2006. Heat as a ground-water tracer at the Russian River RBF facility, Sonoma County, California. In: Hubbs, S.A. (eds.) Riverbank Filtration Hydrology – Impacts on System Capacity and Water Quality. Springer, Dordrecht, pp.

Cox, M.H., Su, G.W. and Constantz, J. 2007. Heat, chloride, and specific conductance as ground water tracers near streams. Ground Water, 45 (2): 187 – 195.

De Roda Husman, A.M., Lodder, W.J., Rutjes, S.A., Schijven, J.F. and Teunis, P.F.M. 2009. Long-term inactivation study of three enteroviruses in artificial surface and groundwaters using PCR and cell culture. Applied and Environmental Microbiology, Vol. 75, 4: 1050–1057.

Defensoría del Pueblo. 2007. Tercer diagnóstico sobre calidad de agua para consumo humano. 168 p.

Defensoría del Pueblo. 2010. Diagnóstico de la calidad de agua para consumo humano año 2009, 66 p.

Bibliographic References

143

Deutsches Institute fur Normung. 2009. DIN EN ISO 11885:2009-09 - Wasserbeschaffenheit - Bestimmung von ausgewählten Elementen durch induktiv gekoppelte Plasma-Atom-Emissionsspektrometrie (ICP-OES) (ISO 11885:2007).

Deutsches Institute fur Normung. 2011. DIN 18123:2011-04 - Baugrund, Untersuchung von Bodenproben - Bestimmung der Korngrößenverteilung.

Deutsches Institute fur Normung. 2012. DIN EN 15936:2012-11 – Schlamm, behandelter Bioabfall, Boden und Abfall - Bestimmung des gesamten organischen Kohlenstoffs (TOC) mittels trockener Verbrennung.

DNP. 2011. Estadísticas del sector Agua Potable y Saneamiento Básico. Colombia. Departamento Nacional de Planeación, DNP. http://www.dnp.gov.co/Programas/ViviendaAguayDesarrolloUrbano/AguaySaneamiento.aspx (Accessed on December 5th, 2012).

Doherty, J., Brebber, L. and Whyte, P. 1994. PEST - Model-independent parameter estimation. User’s manual. Watermark Computing. Australia.

Doussan, C., Ledoux, E. and Detay, M. 1998. River-groundwater exchanges, bank filtration, and groundwater quality: Ammonium Behavior. Journal of Environmental Quality, 27 (6): 1418–1427.

Doussan, C., Poitevin, G., Ledoux, E. and Detay, M. 1997. River bank filtration: modelling of the changes in water chemistry with emphasis in nitrogen species. Journal of Contaminant Hydrogeology, 25: 129 – 156.

Eckhard, W. 1999. Laboratory tests for simulation of Riverbank Filtration processes. Abstracts International Riverbank Filtration Conference, Louisville (Kentucky), November 4th to 6th 1999, p.11.

Engesgaard, P., Seifert, D. and Herrera, P. 2006. Bioclogging in porous media: tracer studies. In: Hubbs SA (eds.) Riverbank Filtration Hydrology – Impacts on System Capacity and Water Quality. Springer, Nato Science Series IV, pp. 93–118.

Escobar, R. 2011. Simulación numérica de Riverbank Filtration o Filtración Ribereña para su aplicación en tres municipios de Antioquia: Santa Fe de Antioquia, Guarne y Nechí. Trabajo de investigación presentado como requisito para optar al título de Magíster en Ingeniería – Recursos Hidráulicos, Universidad Nacional de Colombia, Medellín.

Fetter, C.W. 1999. Contaminant Hydrogeology. Second Edition. Prentice Hall. 500 p.

Fetter, C.W. 2004. Applied Hydrogeology. Fourth Edition. Prentice Hall. 598 p.

Freeze, R.A. and Cherry, J.A. 1979. Groundwater. Englewook Cliffs, NJ: Prentice Hall, 604 p.

Gerba, C.P. 1984. Applied and theoretical aspects of virus adsorption to surfaces. Advances in Applied Microbiology, 30: 133–168.

Gibert, J., Fournier, F. and Mathieu, J. 1997. The groundwater/surface water ecotone perspective: state of the art. In: Gibert J, Mathieu J, Fournier F (eds.)

Evaluation of the Potential for Riverbank Filtration in Colombia

144

Groundwater/Surface Water Ecotones: Biological and Hydrological Interactions and Management Options. Cambridge University Press: New York, pp. 3–6. 1997.

Gollnitz, W.D., Clancy, J.L. and Garner, S.C. 1997. Reduction of microscopic particulates by aquifer. Journal of American Works Association (AWWA), 89 (11): 84 – 93.

Gollnitz, W.D., Clancy, J.L., McWen, J.B. and Garner, S.C. 2005. Riverbank Filtration for IESWTR compliance. Journal of American Works Association (AWWA), 97 (12): 64–76.

Gollnitz, W.D., Clancy, J.L., Whitteberry, B.L. and Vogt, J.A. 2003. RBF as a microbial treatment process. Journal of American Water Works Association (AWWA), 95 (12): 56–66.

Gollnitz, W.D., Whitteberry, B.L. and Vogt, J.A. 2004. Riverbank filtration: induced infiltration and groundwater quality. Journal of American Works Association (AWWA), 96 (12): 98–110.

Grischek, T., Schoenheinz, D. and Ray, C. 2002. Siting and design issues for riverbank filtration schemes. In: Ray C, Melin G, Linsky RB (eds.) Riverbank Filtration Improving Source-Water Quality. Springer, Water Science and Technology Library, Vol. 43, pp. 291–302.

Heberer, T., Massmann, G., Fanck, B., Taute, T. and Dünnbier, U. 2008. Behavior and redox sensitivity of antimicrobial residues during bank filtration. Chemosphere, 73: 451–460.

Hiscock, K.M. and Grischek, T. 2002. Attenuation of groundwater pollution by bank filtration. Journal of Hydrology, 266 (3-4): 139–144.

Hoehn, E., von Gunten, H.R., Stauffer, F. and Dracos, T. 1992. Radon-222 as a groundwater tracer: a laboratory study. Environmental Science and Technology, 26: 734 – 738.

HTW. 2012. Uferfiltration im Lößnitztal - Modellierung der Grundwasserströmung, HTW-Dresden (Hochschule für Technik und Wirtschaft Dresden). 27 p.

Hubbs, S.A. 2006a. Evaluating streambed forces impacting the capacity of riverbed filtration systems. In: Hubbs SA (ed.) Riverbank Filtration Hydrology – Impacts on System Capacity and Water Quality. Springer, Nato Science Series IV, Vol. 60, pp. 21–42.

Hubbs, S.A. 2006b. Changes in riverbed hydraulic conductivity and specific capacity at Louisville. In: Hubbs SA (ed.) Riverbank Filtration Hydrology – Impacts on System Capacity and Water Quality. Springer, Nato Science Series IV, Vol. 60, pp. 199–220.

IAvH, IDEAM, IIAP, INVEMAR, SINCHI. 2011. Informe del Estado del Medio Ambiente y de los Recursos Naturales Renovables 2010. Instituto de Hidrología, Meteorología y Estudios Ambientales – IDEAM. Bogotá D.C., Colombia. 384 p.

ICA. 2010. Comercialización de plaguicidas 2009. Instituto Colombiano Agropecuario, Bogotá, 72 p.

Bibliographic References

145

ICA. 2011. Estadísticas de comercialización de plaguicidas químicos de uso agrícola 2010. Instituto Colombiano Agropecuario. Bogotá D.C., Colombia. 96 p.

IDEAM. 2010. Estudio Nacional del Agua 2010. Instituto de Hidrología, Meteorología y Estudios Ambientales. Bogotá D.C., Colombia. 420 p.

INGEOMINAS. 2002. Memoria explicativa mapa de recursos minerales de Colombia “Minerales Industriales”, Escala 1:500.000.

INGEOMINAS. 2005. Complementación geológica, geoquímica y geofísica de la parte occidental de las Planchas 130 Santa Fe de Antioquia y 146 Medellín Occidental, escala 1:000.000 - Mapa geológico y Memoria Explicativa.

Jekel, M. and Grischek, T. 2003. Riverbank filtration: the European experience. In: G. Melin (ed) Riverbank Filtration: the future is now. Program and abstracts of the 2nd international Riverbank Filtration conference, Cincinnati, Ohio, USA. September 16th to 19th, 2003.

Jüttner, F. 1995. Elimination of terpenoid odorous compounds by slow sand and river bank filtration of the Ruhr River, Germany. Water Science and Technology, 31 (11): 211–217.

Jüttner, F. 1999. Efficacy of bank filtration for the removal of fragrance compounds and aromatic hydrocarbons. Water Science and Technology, 40 (6): 123–128.

Keswick, B.H., Gerba, C.P., Secor, S.L. and Cech, I. 1982. Survival of enteric viruses and indicator bacteria in groundwater. Journal of Environmental Science and Health, A12: 903 – 912.

Kim, S.B. and Corapcioglu, M.Y. 2002. Contaminant transport in riverbank filtration in the presence of dissolved organic matter and bacteria: a kinetic approach. Journal of Hydrology, 266: 269–283.

Kim, S.B., Corapcioglu, M.Y. and Kim, D.J. 2003. Effect of dissolved organic matter and bacteria on contaminant transport in riverbank filtration. Journal of Contaminant Hydrology, 66: 1–23.

Konikow, L.F., Goode, D.J. and Hornberger, G.Z. 1996. A three-dimensional method-of-characteristics solute-transport model (MOC3D): U.S. Geological Survey Water-Resources Investigations Report 96-4267, 87 p.

Kuehn, W. and Mueller, U. 2000. Riverbank filtration: an overview. Journal of American Water Works Association (AWWA), 92 (12): 60–69.

Lallemand-Barres, P. and Peaudecerf, P. 1978. Recherche des relations entre la veleur de la dispersivite macroscopique d’un milieu aquifere, ses autres caracteristiques et les conditions de mesure, etude bibliographique. Bulletin, Bureau de Recherches Geologiques et Miniéres. Sec. 3/4: 277 – 287.

Evaluation of the Potential for Riverbank Filtration in Colombia

146

Lewis, J. and Sjöstrom, J. 2010. Optimizing the experimental design of soil columns in saturated and unsaturated transport experiments. Journal of Contaminant Hydrology, 115: 1 – 13.

Mao, M. and Ren, L. 2005. Simulating nonequilibrium transport of Atrazine through saturated soil. Groundwater, 42(4): 500 – 508.

Marquardt, D.W. 1963. An algorithm for least-squares estimation of nonlinear parameters. Journal of the Society for Industrial and Applied Mathematics, II: 431 – 441.

McDonald, M.G. and Harbaugh, A.W. 1988. A modular three-dimensional finite-difference ground-water flow model. U.S. Geological Survey, Techniques of Water-Resources Investigations, Book 6, Chapter A1, 586 p.

McDowell-Boyer, L.M., Hunt, J. R. and Sitar, N. 1986. Particle transport through porous media. Water Resources Research. 22: 1901–1921.

McFeters, G.A., Bissonnette, G.K, Jewelski, J.J., Thomson, C.A. and Stuart, D.G. 1974. Comparative survival of indicator bacteria and enteric pathogens in well water. Applied and Environmental Microbiology, 27(5): 823 – 829.

Miller, C.T. and Webber, W.J. 1984. Modeling organic contamination partitioning in ground-water systems. Ground Water, 22(5): 584 – 592.

Mucha, I., Banský, L., Hlavatý, Z. and Rodák, D. 2006. Impact of riverbed clogging – colmatation – on ground water. In: Hubbs SA (ed.) Riverbank Filtration Hydrology – Impacts on System Capacity and Water Quality. Springer, Nato Science Series IV, pp. 43–72.

Ndabigengesere, A. and Narasiah, K.S. 1998. Quality of water treated by coagulation using Moringa oleifera seeds. Water Resources, 32 (3): 781–791.

Neuman, S.P. 1990. Universal scaling of hydraulic conductivities and dispersivities in geologic media. Water Resources Research, 26 (8): 1749 – 1758.

Ormad, M.P., Miguel, N., Claver, J., Matesanz, J.M. and Ovelleiro, J.L. 2008. Pesticides removal in the process of drinking water production. Chemosphere, 71: 97 – 106.

Pang, L. and Close, M.E. 1999. Attenuation and transport of atrazine and picloram in an alluvial gravel aquifer: A tracer test and batch study. New Zealand Journal of Marine and Freshwater Research, 33 (2): 279-291

Paraiba, L.C. and Spadotto, C.A. 2002. Soil temperature effect in calculating attenuation and retardation factors. Chemosphere 48: 905 – 912.

Perkins, T.K. and Johnson, O.C. 1963. A review of diffusion and dispersion in porous media. Society of Petroleum Engineers Journal, 3: 70 – 84.

Poeter, E.P. and Hill, M.C. 1998. Documentation of UCODE, a computer code for universal inverse modeling, U.S. Geological Survey, Water-Resources Investigations Report 98-4080

Bibliographic References

147

Rashid, B., Husnain, T. and Riazuddin, S. 2010. Herbicides and pesticides as potential pollutants: a global problem. In: M. Ashraf et al. (eds.) Plant Adaptation and Phytoremediation. Springer Science + Business Media. pp. 427 – 447.

Ray, C. 2008. Worldwide potential of riverbank filtration. Clean Technologies and Environmental Policy, 10: 223–225.

Ray, C., Grischek, T., Schubert, J., Wang, J.Z. and Speth, T.F. 2002. A perspective of riverbank filtration. Journal of American Water Works Association (AWWA), 94 (4): 149–160.

Ray, C., Soong, T.W.D., Roadcap, G.S. and Borah, D.K. 1998. Agricultural chemicals: Effects on wells during floods. Journal of American Water Work Association, 90: 90-100.

Reed, B.E., Matsumoto, M.R., Viadero, R.Jr. and Segar, R.L.Jr. 1999. Physicochemical processes. Water Environment Research, 71 (5): 584–618.

República de Colombia. 2007. Resolución No. 2115 del 22 de Junio de 2007 por la cual se señalan características, instrumentos básicos y frecuencias del sistema de control y vigilancia para la calidad del agua para consumo humano. Ministerio de la Protección Social y Ministerio de Ambiente, Vivienda y Desarrollo Territorial. 23 p.

Rice, E.W., Johnson, C.H., Wid, D.K. and Reasoner, D.J. 1992. Survival of Escherichia coli O157:H7 in drinking water associated with a waterborne disease outbreak of hemorrhagic colitis. Letters in Applied Microbiology, 13: 38 – 40.

Richardson, S.D. 2003. Disinfection by-products and other emerging contaminants in drinking water. TrAC Trends in Analytical Chemistry, 22 (10): 666–684.

SAN, 2011. List of prohibited pesticides. Sustainable Agriculture Network. 8 p.

Sandhu, C., Grischek, T., Kumar, P. and Ray, C. 2010. Potential for riverbank filtration in India. Clean Techn Environ Policy, pp. 1-22 (DOI 10.1007/s10098-010-0298-0).

Schafer, F. 2006. Use of aquifer testing and groundwater modeling to evaluate aquifer/river hydraulics at Louisville Water Company, Luisville, Kentucky, USA. In: Hubbs SA (ed.) Riverbank Filtration Hydrology – Impacts on System Capacity and Water Quality. Springer, Nato Science Series IV, pp. 179–198.

Schaffner, C., Ahel, M. and Giger, W. 1987. Field studies on the behaviour of organic micropollutants during infiltration of river water to groundwater. Water Science and Technology, 19: 1195-1196.

Schijven, J.F. and Hassanizadeh, S.M. 2000. Removal of viruses by soil passage: overview of modelling, processes, and parameters. Critical Reviews in Environmental Science and Technology, 30 (1): 49–127.

Schijven, J.F., Berger, P. and Miettinen, I. 2002. Removal of pathogens, surrogates, indicators, and toxins using Riverbank Filtration. In: Ray C, Melin G, Linsky RB (eds.) Riverbank Filtration Improving Source-Water Quality. Springer, Water Science and Technology Library, Vol. 43, pp. 73–116.

Evaluation of the Potential for Riverbank Filtration in Colombia

148

Schön, M. 2006. Systematic comparison of riverbank filtration sites in Austria and India. Diplomarbeit, Leopold Franzens Universität. Innsbruck.

Schubert, J. 2000. How does it work? Field studies on riverbank filtration. In: Julich W, Schubert J (eds.) Proceedings of the International Riverbank Filtration Conference. IAWR, Dusseldorf, Germany, pp. 41 – 55.

Schubert, J. 2002a. Hydraulic aspects of riverbank filtration – field studies. Journal of Hydrology, 266: 145 – 161.

Schubert, J. 2002b. German experience with riverbank filtration systems. In: Ray C, Melin G, Linsky R. (eds.) Riverbank Filtration Improving Source-Water Quality. Springer, Water Science and Technology Library, Vol. 43, pp. 35 – 48.

Schubert, J. 2002c. Water-quality improvements with riverbank filtration at Düsseldorf waterworks in Germany. In: Ray C, Melin G, Linsky RB (eds.) Riverbank Filtration Improving Source-Water Quality. Springer, Water Science and Technology Library, Vol. 43, pp. 267 – 277.

Schubert, J. 2006a. Changes in riverbed hydraulic conductivity and specific capacity at Louisville. In: Hubbs, S.A. (eds) Riverbank Filtration Hydrology – Impacts on System Capacity and Water Quality. Springer, Dordrecht, pp. 199 – 220.

Schubert, J. 2006b. Experience with riverbed clogging along the Rhine River. In: Hubbs SA (ed.) Riverbank Filtration Hydrology – Impacts on System Capacity and Water Quality. Springer, Nato Science Series IV, pp. 221 – 242.

Schubert, J. 2006c. Sifgnificane of hydrologic aspects on RBF performance. In: Hubbs SA (ed.) Riverbank Filtration Hydrology – Impacts on System Capacity and Water Quality. Springer, Nato Science Series IV, pp. 1 – 20.

Scribner, E.A., Thurman, E.M. and Zimmerman, L.R. 2000. Analysis of selected metabolites in surface and ground water in the United States. The Science of the Total Environment: 248, 157–167.

Seifert, D. and Engesgaard, P. 2007. Use of tracer tests to investigate changes in flow and transport properties due to bioclogging in porous media. Journal of Contaminant Hydrology, 93: 58 – 71.

Simunek, J., van Genuchten, M.Th., Sejna, M., Toride, N. and Leij, F.J. 1999. The STANMOD computer software for evaluating solute transport in porous media using analytical solutions of convection-dispersion equation. Versions 1.0 and 2.0, IGWMC - TPS - 71, International Ground Water Modeling Center, Colorado School of Mines, Golden, Colorado. 32p.

Sinclair, C.J. and Boxall, A.B.A. 2003. Assessing the ecotoxicity of pesticide transformation products. Environmental Science and Technology, 37(20): 4617–4625.

Son, B.T. 2010. Role of riverbank filtration in the attenuation of herbicides. PhD. Dissertation, School of Environmental Sciences, University of East Anglia. 255 p.

Bibliographic References

149

Stuyfzand, P.J., Juhasz-Holterman, M.H.A. and De Lange, W.J. 2006. Riverbad filtration in the Netherlands: well fields, clogging and geochemical reactions. In: Hubbs SA (ed.) Riverbank Filtration Hydrology – Impacts on System Capacity and Water Quality. Springer, Nato Science Series IV, pp. 119 – 153.

Sutherland, I.W. 2001. Biofilms exopolysaccharides: a strong and sticky framework. Microbioloty, 147: 3 – 9.

Toride, N., Leij, F.J. and van Genuchten, M.Th. 1995. The CXTFIT Code for Estimating Transport Parameters from Laboratory or Field Tracer Experiments, Version 2.0, Research Report No. 137, U. S. Salinity Laboratory, USDA, ARS, Riverside, CA. 131 p.

Trettin, R., Grischek, T., Strauch, G., Mallean, G. and Nestler, W. 1999. The suitability and usage of 18O and chlorides as natural tracers for bank filtrate at the Middle River Elbe. Isotopes in Environmental and Health Studies, 35 (4): 331 – 350.

Tufenkji, N., Ryan, J.N. and Elimelech, M. 2002. The promise of bank filtration. Environmental Science and Technology, 1: 423 – 428.

UNAL - CORANTIOQUIA. 2004. Evaluación del potencial acuífero en los municipios de Santa Fe de Antioquia, San Jerónimo, Sopetrán, Olaya y Liborina. Medellín.

UNAL - CORNARE. 2000. Investigación de aguas subterráneas región Valle de San Nicolás. Fase II. Informe Final. Medellín.

Vanek, V. 1997. Heterogeneity of groundwater-surface water ecotones. In: Gibert J, Mathieu J, Fournier F (eds.) Groundwater/Surface Water Ecotones: Biological and Hydrological Interactions and Management Options. Cambridge University Press: New York; pp 151–161.

Verstraeten, I.M. and Heberer, T. 2002. Organic chemical removal issues. In: Ray C, Melin G, Linsky R. (eds.) Riverbank Filtration Improving Source-Water Quality. Springer, Water Science and Technology Library, Vol. 43, pp. 321 – 330.

Verstraeten, I.M., Carr, J.D., Steele, G.V., Thurman, E.M. and Dormedy, D.F. 1999. Surface-water/ground-water interaction: Herbicide transport into municipal collector wells. Journal of Environmental Quality, 28(5): 1396-1405.

Verstraeten, I.M., Heberer, T. and Scheytt, T. 2002a. Occurrence, characteristics, transport, and fate of pesticides, pharmaceuticals, industrial products, and personal care products at riverbank filtration sites. In: Ray C, Melin G, Linsky R. (eds.) Riverbank Filtration Improving Source-Water Quality. Springer, Water Science and Technology Library, Vol. 43, pp. 175 – 227.

Verstraeten, I.M., Thurman, E.M., Lindsey, M.E., Lee, E.C. and Smith, R.E. 2002b. Changes in concentrations of triazine and acetamide herbicides by bank filtration, ozonation, and chlorination in a public water supply. Journal of Hydrology (266): 190 – 208.

Evaluation of the Potential for Riverbank Filtration in Colombia

150

Wang, J. 2002. Riverbank filtration case study at Luisville, Kentucky. In: Ray C, Melin G, Linsky RB (eds.) Riverbank Filtration Improving Source-Water Quality. Springer, Water Science and Technology Library, Vol. 43, pp. 117–145.

Weiss, W.J., Bouwer, E.J., Ball, W.P., O’Melia, C.R., Arora, H. and Speth, T.F. 2002. Reduction in disinfection byproduct precursors and pathogens during riverbank filtration at three Midwestern United States drinking-water utilities. In: Ray C, Melin G, Linsky RB (eds.) Riverbank Filtration Improving Source-Water Quality. Springer, Water Science and Technology Library, Vol. 43, pp. 147–173.

Weiss, W.J., Bouwer, E.J., Ball, W.P., O’Melia, C.R., Lechevallier, M.W., Arora, H. and Speth, T. F. 2003. Riverbank filtration – fate of DBP precursors and selected microorganisms. Journal of American Water Works Association (AWWA), 95 (10): 68–81.

WHO – UNICEF. 2008. Progress on drinking water and sanitation: special focus on sanitation. World Health Organization and UNICEF Joint Monitoring Programme for Water Supply and Sanitation (JMP). 58 p.

WHO – UNICEF. 2010. Progress on sanitation and drinking water, 2010 update. World Health Organization and UNICEF Joint Monitoring Programme for Water Supply and Sanitation (JMP). 60 p.

WHO. 2009. 10 facts about water scarcity. World Health Organization. http://www.who.int/features/factfiles/water/en/index.html (Last accessed on November 30th, 2011).

WHO. 2011. World health statistics. World Health Organization, 171 p.

Xu, M. and Eckstein, Y. 1995. Use of weighted least-squares method in evaluation of the relationship between dispersivity and field scale. Ground Water, 16 (6): 905 – 908.

Yates, M.V., Gerba, C.P. and Kelley, L.M. 1985. Virus persistence in Groundwater. Applied and Environmental Microbiology, Vol. 49 (4): 778–781.

Zheng, C. 1990. MT3D A modular three-dimensional transport model for simulation of advection, dispersion and chemical reactions of contaminants in groundwater systems. S.S. Papadopulos & Associates, Inc., Rockville, Maryland, 163 p.

Zheng, C. and Wang, P. 1999. MT3DMS A modular three-dimensional multispecies transport model for simulation of advection, dispersion and chemical reactions of contaminants in groundwater systems – Documentation and User’s guide. US Army Corps of Engineers, Washington, D.C. 239 p.

Zhou, N., Matsumoto, T., Hosokawa, T. and Suekane, T. 2010. Pore-scale visualization of gas trapping in porous media by X-Ray CT Scanning. Flow Measurement and Instrumentation, 21 (3): 262–267.

151

APPENDIX A Multiannual monthly values of discharge in m3/s (blue bars) and total suspended solids in kg/m3 (orange line) for the selected stations in Colombia

Evaluation of the Potential for Riverbank Filtration in Colombia

152

153

APPENDIX B Resulting diagrams for the estimation of mean curvature radius based on maximum and minimum local variances

Evaluation of the Potential for Riverbank Filtration in Colombia

154

Appendix B

155

Evaluation of the Potential for Riverbank Filtration in Colombia

156

157

APPENDIX C Analytical Methods used at the laboratories of CORANTIOQUIA (left) and CORNARE (right)

CORANTIOQUIA CORANARE

Parameter Analytical Method

Parameter Analytical Method

Bicarbonates (mg/L HCO3) S.M.2320 Bicarbonates (mg/L HCO3) Total Alkalinity x 1.22

Calcium (mg/L Ca) S.M.3111D Calcium (mg/L Ca) S.M.3111D

Carbonates (mg/L CO3) S.M.2320 Chloride (mg/L Cl) S.M.4500Cl-B

Chloride (mg/L Cl) S.M.4500Cl-B Conductivity (uS/cm) S.M.2510B

Conductivity (uS/cm) S.M.2510B Dissolved Oxygen (mg/L O2) S.M.4500O-C

Disolved Organic Carbon (mg/L DOC) S.M.5310B E. Coli (CFU/100mL) S.M.9222H

Dissolved Iron (mg/L Fe) S.M.3500Fe-B Magnesium (mg/L Mg) S.M.3111B

Dissolved Oxygen (mg/L O2) S.M.4500O-B Manganese (ug/L Mn) S.M.3111B-Mn

Dissolved Silica (mg/L Si) S.M.3111D Nitrites + Nitrates (mg/L N-) S.M.4500NO3-E and NO2-B

E. Coli (NMP/100mL) S.M.9223 pH (units of pH) S.M.4500H+B

Magnesium (mg/L Mg) S.M.3111B Potassium (mg/L K) S.M.3111B

Nitrates (mg/L NO3-N) S.M.4500NO3-D Sodium (mg/L Na) S.M. 3111B

Nitrites (mg/L NO2-N) S.M.4500NO2-D Soluble Phosphorous (mg/ P) S.M.4500P-B,E

pH (units of pH) S.M.4500H+B Sulphates (mg/L (SO4)2) S.M.4500SO4=E

Potassium (mg/L K) S.M.3111B Total Alcalinity (mg/L CaCO3) S.M.2320B

Sodium (mg/L Na) S.M.3111B Total BOD (mg/L O2) S.M.5210B

Soluble Phosphorous (mg/ P) S.M.4500P-E Total COD (mg/L O2) S.M.5220C

Sulphates (mg/L (SO4)2) S.M.4500SO4=D Total Coliforms (CFU/100mL) S.M.9222B

Total Alcalinity (mg/L CaCO3) S.M.2320 Total Dissolved Solids TS-TSS

Total BOD (mg/L O2) S.M.5210B Total Hardness (mg/L CaCO3) S.M.2340C

Total COD (mg/L O2) S.M.5220D Total Iron (mg/L Fe) S.M.3500Fe-B

Total Coliforms (NMP/100mL) S.M.9223 Total Phosphates (mg/L (PO4)3) S.M.4500P-B,E

Total Cyanide (mg/L CN) S.M.4500CN-D Total Phosphorous (mg/L P) S.M.4500P-B,E

Total Dissolved Solids (mg/L) S.M.2540B,E Total Solids (mg/L) S.M.2540B

Total Hardness (mg/L CaCO3) S.M.2340C Total Suspended Solids (m/-L) S.M.2540D

Total Mercury (mg/L Hg) S.M.3112B Total volatile solids (mg/L) S.M.2540E

Total Organic Carbon (mg/L TOC) S.M.5310B True Color (UPC) S.M.2120B

Total Phosphates (mg/L (PO4)3) S.M.4500P-E Turbidity (NTU) S.M.2130B

Total Phosphorous (mg/L P) S.M.4500P-E Volatile Suspended Solids (mg-L) S.M.2540E

Total Solids (mg/L) S.M.2540B,E Total Suspended Solids (m/-L) S.M.2540D,E

Total Volatile Solids (mg/L) S.M.2540B,E True Color (UPC) S.M.2120B Turbidity (NTU) S.M.2130B Volatile Suspended Soids (mg-L) S.M.2540D,E Volitile Dissolved Solids (mg/L) S.M.2540B,E

S.M.: Standard Methods

159

APPENDIX D Results of the water analysis in Santa Fe de Antioquia, performed at the lab in CORANTIOQUIA

Corantioquia ID 745WIC1 745WIC2 787WIC1 787WIC2

Sampling Date 2-Jun-10 2-Jun-10 23-Jun-10 23-Jun-10

Parameter River Well River Well

Bicarbonates (mg/L HCO3) 51.1 139.0 64.3 115.0

Carbonates (mg/L CO3) 0.00 0.00 0.00 0.00

Chloride (mg/L Cl) 4.13 6.33 3.73 5.62

Conductivity (μS/cm) 124 346 144 287

Dissolved Organic Carbon (mg/L DOC) 2.25 1.97 2.84 1.51

Dissolved Sodium (mg/L Na) 7.05 9.90 8.07 12.77

Dissolved Calcium (mg/L Ca) 18.6 37.8 15.2 32.5

Dissolved Iron (mg/L Fe) 0.257 0.047 0.344 0.050

Dissolved Magnesium (mg/L Mg) 4.60 13.90 4.97 11.10

Dissolved Oxygen (mg/L O2) 4.00 5.49 5.21 6.24

Dissolved Potassium (mg/L K) 1.36 2.51 2.05 2.12

Dissolved Silica (mg/L Si) 8.93 13.40 9.58 13.60

E. Coli (NMP/100mL) 4.0E+04 1.0E+02 2.0E+04 1.1E+02

Nitrates (mg/L NO3-N) 3.91 5.05 3.30 5.42

Nitrites (mg/L NO2-N) 0.007 <0.003 0.019 0.004

pH (units of pH) 7.52 7.15 7.86 7.60

Soluble Phosphorous (mg/ P) 0.069 0.141 0.512 0.135

Sulphates (mg/L (SO4)2) 6.60 54.10 14.70 42.50

Total Alcalinity (mg/L CaCO3) 41.9 114.0 52.7 93.9

Total BOD (mg/L O2) 4.79 <2 10.30 7.59

Total COD (mg/L O2) 41.9 16.9 28.2 <12

Total Coliforms (NMP/100mL) 1.6E+08 9.4E+03 1.4E+06 9.2E+05

Total Cyanide (mg/L CN) Not measured Not measured Not measured Not measured

Total Dissolved Solids (mg/L) 215 317 33 194

Total Hardness (mg/L CaCO3) 71.1 167.0 60.2 128.0

Total Mercury (mg/L Hg) Not measured Not measured Not measured Not measured

Total Organic Carbon (mg/L TOC) 4.26 2.37 5.23 3.60

Total Phosphates (mg/L (PO4)3) <0.153 0.399 0.160 0.342

Total Phosphorous (mg/L P) 1.070 0.149 0.858 0.142

Total Solids (mg/L) 1118 320 1716 197

Total Suspended Solids (m/-L) 904 <7 1683 <7

True Color (UPC) >150 <5 >150 10

Turbidity (NTU) 639 1.2 400 0.7

Volatile Solids (mg/L) 176 146 116 44

Volatile Suspended Solids (mg-L) 99 3 104 3

Volatile Dissolved Solids (mg/L) 78 140 12 41

Evaluation of the Potential for Riverbank Filtration in Colombia

160

Corantioquia ID 865WIC1 865WIC2 907WIC1 907WIC2

Sampling Date 7-Jul-10 7-Jul-10 28-Jul-10 28-Jul-10

Parameter River Well River Well

Bicarbonates (mg/L HCO3) 64.3 217.0 62.0 184.0

Carbonates (mg/L CO3) 0.00 0.00 0.00 0.00

Chloride (mg/L Cl) 3.92 6.78 3.85 5.78

Conductivity (μS/cm) 135 406 129 351

Dissolved Organic Carbon (mg/L DOC) 2.12 1.92 3.65 1.54

Dissolved Sodium (mg/L Na) 6.68 11.20 6.94 8.83

Dissolved Calcium (mg/L Ca) 14.1 63.8 14.7 39.2

Dissolved Iron (mg/L Fe) 0.245 0.050 0.369 0.050

Dissolved Magnesium (mg/L Mg) 4.99 19.70 5.27 18.00

Dissolved Oxygen (mg/L O2) 5.76 4.42 5.61 6.08

Dissolved Potassium (mg/L K) 1.81 3.40 1.72 2.93

Dissolved Silica (mg/L Si) 9.03 14.10 7.23 9.90

E. Coli (NMP/100mL) 2.1E+04 6.8E+02 1.4E+05 2.0E+01

Nitrates (mg/L NO3-N) <1.5 <1.5 <1.5 <1.5

Nitrites (mg/L NO2-N) 0.015 <0.003 0.013 <0.003

pH (units of pH) 7.90 7.51 7.80 7.82

Soluble Phosphorous (mg/ P) 0.662 0.143 0.092 0.064

Sulphates (mg/L (SO4)2) 10.10 36.50 11.70 29.00

Total Alcalinity (mg/L CaCO3) 52.7 178.0 50.8 151.0

Total BOD (mg/L O2) 7.56 7.26 2.35 <2

Total COD (mg/L O2) 59.1 25.6 142.0 <12

Total Coliforms (NMP/100mL) 2.8E+05 9.4E+03 1.6E+07 1.3E+04

Total Cyanide (mg/L CN) Not measured Not measured Not measured Not measured

Total Dissolved Solids (mg/L) 180 336 125 246

Total Hardness (mg/L CaCO3) 56.8 242.0 59.1 173.0

Total Mercury (mg/L Hg) Not measured Not measured Not measured Not measured

Total Organic Carbon (mg/L TOC) 6.92 2.49 7.67 1.54

Total Phosphates (mg/L (PO4)3) <0.153 0.394 0.189 0.197

Total Phosphorous (mg/L P) 0.991 0.147 0.619 0.064

Total Solids (mg/L) 1404 340 1408 246

Total Suspended Solids (m/-L) 1225 <7 1283 <7

True Color (UPC) >150 15 >150 <5

Turbidity (NTU) 767 2.8 720 0.6

Volatile Solids (mg/L) 174 152 170 54

Volatile Suspended Solids (mg-L) 116 4 118 7

Volatile Dissolved Solids (mg/L) 58 148 52 54

APPENDIX D

161

Corantioquia ID 1140WIC1 1140WIC2 1182WIC1 1182WIC2

Sampling Date 8-Sep-10 8-Sep-10 22-Sep-10 22-Sep-10

Parameter River Well River Well

Bicarbonates (mg/L HCO3) 66.7 224.0 67.0 257.4

Carbonates (mg/L CO3) 0.00 0.00 0.00 0.00

Chloride (mg/L Cl) 3.91 12.80 4.52 19.20

Conductivity (μS/cm) 148 541 142 789

Dissolved Organic Carbon (mg/L DOC) 3.50 1.82 5.53 1.75

Dissolved Sodium (mg/L Na) 7.21 12.20 7.28 20.80

Dissolved Calcium (mg/L Ca) 14.1 70.0 16.7 102.0

Dissolved Iron (mg/L Fe) 0.153 0.012 0.363 0.002

Dissolved Magnesium (mg/L Mg) 5.78 27.60 5.62 36.10

Dissolved Oxygen (mg/L O2) 5.82 3.86 4.25 2.50

Dissolved Potassium (mg/L K) 1.85 2.63 2.02 3.49

Dissolved Silica (mg/L Si) 9.06 12.00 10.70 14.80

E. Coli (NMP/100mL) 1.4E+03 2.0E+02 1.6E+05 2.3E+02

Nitrates (mg/L NO3-N) <1.5 <1.5 1.71 1.97

Nitrites (mg/L NO2-N) 0.014 <0.003 0.007 0.005

pH (units of pH) 7.96 7.50 7.84 7.40

Soluble Phosphorous (mg/ P) 0.080 0.040 0.602 0.089

Sulphates (mg/L (SO4)2) 9.42 70.00 8.62 195.00

Total Alcalinity (mg/L CaCO3) 54.7 184.0 54.9 211.0

Total BOD (mg/L O2) <2 2.58 2.74 2.76

Total COD (mg/L O2) 167.0 <12 160.0 12.2

Total Coliforms (NMP/100mL) >1.6+E08 7.0E+03 2.2E+05 9.2E+04

Total Cyanide (mg/L CN) Not measured Not measured Not measured Not measured

Total Dissolved Solids (mg/L) <10 406 164 742

Total Hardness (mg/L CaCO3) 60.0 288.0 40.2 404.0

Total Mercury (mg/L Hg) Not measured Not measured Not measured Not measured

Total Organic Carbon (mg/L TOC) 5.65 2.06 9.14 2.13

Total Phosphates (mg/L (PO4)3) <0.153 <0.153 <0.153 0.248

Total Phosphorous (mg/L P) 0.431 0.047 1.110 0.090

Total Solids (mg/L) 750 416 1738 744

Total Suspended Solids (m/-L) 743 10 1574 <7

True Color (UPC) >150 10 >150 <5

Turbidity (NTU) 400 2.3 >1000 1.5

Volatile Solids (mg/L) 68 112 226 290

Volatile Suspended Solids (mg-L) 55 9 169 2

Volatile Dissolved Solids (mg/L) 7 103 57 288

Evaluation of the Potential for Riverbank Filtration in Colombia

162

Corantioquia ID 1268WIC1 1268WIC2 1324WIC1 1324WIC2

Sampling Date 6-Oct-10 6-Oct-10 20-Oct-10 20-Oct-10

Parameter River Well River Well

Bicarbonates (mg/L HCO3) 67.8 218.0 74.2 246.0

Carbonates (mg/L CO3) 0.00 0.00 0.00 0.00

Chloride (mg/L Cl) 3.49 14.60 5.00 18.20

Conductivity (μS/cm) 138 625 156 747

Dissolved Organic Carbon (mg/L DOC) 3.81 1.00 2.57 1.00

Dissolved Sodium (mg/L Na) 7.13 19.00 8.2 20.2

Dissolved Calcium (mg/L Ca) 12.3 107.0 16.2 98.5

Dissolved Iron (mg/L Fe) 0.267 0.025 0.162 0.011

Dissolved Magnesium (mg/L Mg) 4.86 30.20 5.99 35.40

Dissolved Oxygen (mg/L O2) 5.28 2.88 4.99 2.84

Dissolved Potassium (mg/L K) 1.73 2.84 1.93 3.47

Dissolved Silica (mg/L Si) 7.23 9.95 5.70 7.96

E. Coli (NMP/100mL) 3.9E+04 2.0E+02 8.4E+04 1.8E+01

Nitrates (mg/L NO3-N) <1.5 <1.5 1.88 <1.5

Nitrites (mg/L NO2-N) 0.008 <0.003 0.012 <0.003

pH (units of pH) 7.68 7.37 7.86 7.32

Soluble Phosphorous (mg/ P) 0.348 0.070 0.730 0.094

Sulphates (mg/L (SO4)2) 8.61 119.00 13.20 154.00

Total Alcalinity (mg/L CaCO3) 55.6 179.0 60.8 202.0

Total BOD (mg/L O2) <2 <2 2.48 2.66

Total COD (mg/L O2) 119.0 <12 77.5 <12

Total Coliforms (NMP/100mL) 1.6E+07 6.3E+03 >1.6+E08 1.6E+06

Total Cyanide (mg/L CN) 0.099 0.050 0.050 0.050

Total Dissolved Solids (mg/L) 92 532 138 635

Total Hardness (mg/L CaCO3) 51.0 393.0 66.0 393.0

Total Mercury (mg/L Hg) 0.003 <0.001 0.004 <0.001

Total Organic Carbon (mg/L TOC) 7.38 1.95 5.27 1.33

Total Phosphates (mg/L (PO4)3) <0.153 0.206 0.210 0.263

Total Phosphorous (mg/L P) 0.873 0.070 0.762 0.096

Total Solids (mg/L) 2266 534 1528 638

Total Suspended Solids (m/-L) 2175 <7 1390 <7

True Color (UPC) >150 8 >150 10

Turbidity (NTU) >1000 0.3 502.00 0.4

Volatile Solids (mg/L) 208 150 155 173

Volatile Suspended Solids (mg-L) 194 2 116 2

Volatile Dissolved Solids (mg/L) 14 148 39 170

163

APPENDIX E

Universidad Nacional de Colombia – Sede Medellín

Facultad de Minas

Laboratorio de preparación de muestras

Informe:

Ensayos y correlación de resultados de las muestras N1, N2, N3, B1, B2 y L tomadas en inmediaciones del río Cauca

Presentado a:

Marcela Jaramillo

Medellín

Marzo de 2012

Evaluation of the Potential for Riverbank Filtration in Colombia

164

INTRODUCCIÓN

Este informe se elaboró con base en análisis realizados a 6 muestras tomadas en inmediaciones del río Cauca entre las poblaciones de Santa Fé de Antioquia y Anzá. El objetivo principal del muestreo era determinar si en las muestras recolectadas había presencia de algún tipo de sulfato.

Para cumplir con el objetivo propuesto, las muestras se analizaron por tres medios diferentes. Primero se realizó un análisis granulométrico, donde se determinó la distribución granulométrica, esfericidad, redondez y selección de los clastos; después, se realizaron dos ensayos químicos, el primero de porcentaje de materia orgánica y el segundo de carbonatos; finalmente, se emplearon medios ópticos como lupa para analizar los clastos tamaño grava y microscopio en secciones delgadas, ambos con el fin de determinar la presencia de sulfatos en las muestras.

Las muestras fueron tomadas de 2 lugares diferentes: Nacho1, Nacho2 y Nacho3 corresponden a tres muestras tomadas de un pozo ubicado sobre una terraza aluvial del río Cauca; mientras que las muestras Banca1, Banca2 y Lecho, corresponden a tres muestras tomadas en una quebrada cercana al punto de muestreo las dos primeras de cada banca de la quebrada y la tercera del lecho de esta. Para facilidad en el manejo dentro del laboratorio y en la presentación de los resultados, las muestras fueron abreviadas de la siguiente manera:

Nacho1: N1

Nacho2: N2

Nacho3: N3

Banca1: B1

Banca2: B2

Lecho: L

Aunque a todas las muestras se les practicaron los mismos ensayos, hubo una excepción en el procedimiento de las secciones delgadas, pues solo se realizaron secciones de las muestras N1, B1, B2 y L, las razones de esta decisión se explicarán en el capítulo correspondiente.

APPENDIX E

165

ANÁLISIS GRANULOMÉTRICO

En este apartado se explica el procedimiento seguido desde la recepción de las muestras, pasando por la preparación de ellas y finalmente presentando los resultados obtenidos de selección, redondez y esfericidad en las muestras.

Las muestras se recibieron húmedas, pues provenían como se dijo previamente, de una terraza aluvial y una quebrada. El primer paso que se realizó, fue el secado de las muestras a 30° centígrados para no afectar el posible contenido de materia orgánica (m.o.) en las muestras; este secado se realizó durante 5 horas en promedio para cada muestra.

Posteriormente, las muestras fueron pesadas en una balanza de precisión correctamente calibrada; los pesos iniciales obtenidos para cada una de las muestras se consignan en la tabla 1.

Tabla 1. Pesos iniciales de las muestras y pasos seguidos para la preparación de las muestras.

Con el fin de realizar el tamizaje de las muestras, se realizó un cuarteo de cada una para obtener la mayor homogeneidad en la distribución de los clastos. El tamizaje fue realizado con la ayuda de un Ro-Tap 6 tamices y un fondo, donde cada muestra permaneció por 15 minutos, al cabo de este tiempo, cada fracción fue pesada y posteriormente empacada en una bolsa. La tabla 2 muestra el peso obtenido de cada una de esas fracciones para cada muestra y en la figura 1, se observa la distribución granulométrica de cada muestra.

Como se observa en la figura 1 y después de haber realizado una inspección visual con ayuda de una lupa a cada muestra, todas las muestras tienden a un alto porcentaje de clastos tamaño grava e incluso mayores, aunque presentan una gran variación en sus tamaños de grano, pues ninguna de las otras fracciones está ausente; debido a esto, se dice que las muestras tienen una selección muy mala o que su selección es muy baja.

MuestraPeso inicial

(g)Secado Tamizado Macerado

N1 821,6 x x x

N2 823,2 x x x

N3 871,7 x x x

B1 1190,7 x x x

B2 841,7 x x x

L 844,3 x x x

Evaluation of the Potential for Riverbank Filtration in Colombia

166

Tabla 2. Fracciones obtenidas para cada muestra después del tamizaje.

Figura 1. Distribución granulométrica para cada muestra.

De igual manera y con la ayuda de la lupa, se analizó la redondez y la esfericidad de los clastos tamaño grava y arena, los cuales arrojaron que todas las muestras presentaban una redondez buena, en especial las muestras B1, B2 y L; en cuanto a la esfericidad, esta variaba entre buena e intermedia, siendo mejor nuevamente en las muestras B1, B2 y L. Lo dicho anteriormente no implica que no existan en las muestras clastos angulosos o irregulares, simplemente son escasos y poco representativos para la observación general de la muestra.

Tamaños de

muestra>2000μm

2000μm -

840 μm

840 μm -

500 μm

500 μm -

250 μm

250 μm -

125 μm

125 μm -

63 μm<63 μm

Nacho 1 (N1) 49,94 11,95 6,76 8,61 8,24 7,40 6,78

Nacho 2 (N2) 47,88 12,92 7,77 9,74 8,54 7,32 5,27

Nacho 3 (N3) 40,42 18,54 11,14 11,91 8,59 5,99 3,32

Banca 1 (B1) 32,37 18,89 16,24 18,61 8,72 3,50 1,48

Banca 2 (B2) 34,90 14,52 11,24 10,37 15,47 11,85 1,50

Lecho (L) 38,26 33,43 16,22 7,59 2,15 1,20 0,69

Tamizado wt%

0

10

20

30

40

50

60

>2000μm 2000μm - 840 μm

840 μm - 500 μm

500 μm - 250 μm

250 μm - 125 μm

125 μm - 63 μm

<63 μm

N1

N2

N3

B1

B2

L

APPENDIX E

167

La determinación de la esfericidad y redondez en la fracción limo y arcilla no se realizó, debido a que ambos conceptos según varios autores desaparecen en esta fracción o en este tamaño de clasto.

ANÁLISIS QUÍMICOS

En esta sección se explicarán los procedimientos que se siguieron para determinar los porcentajes de materia orgánica y carbonato de calcio (CaCO3) para cada muestra.

Antes de iniciar los respectivos procedimientos determinativos, se requiere tomar parte de la muestra inicial, con distintos tamaños de grano y macerar hasta llevar a un tamaño arcilla, para facilitar la determinación de materia orgánica y carbonatos, según los métodos empleados.

Determinación de Materia Orgánica

El porcentaje de materia orgánica para cada muestra fue determinado por el laboratorio de carbones de la Universidad Nacional de Colombia – Sede Medellín. El proceso seguido por dicho laboratorio consiste en tomar el peso inicial de la muestra entregada y someterlo a una calcinación en un horno a _____ grados centígrados, durante _______ tiempo; después de esto, se retira la muestra del horno se deja a temperatura ambiente durante unos instantes y se toma el peso final. La resta entre ambos pesos, el inicial y el final, será la cantidad de materia orgánica que había en la muestra. La tabla 3 muestra los resultados obtenidos de materia orgánica en porcentaje para cada muestra.

Tabla 3. Porcentaje de m.o. para cada muestra.

Se puede observar a partir de los resultados, una similitud en los valores de materia orgánica para las tres muestras provenientes de la quebrada (B1, B2 y L), mientras que para las otras tres muestras (N1, N2 y N3), se observan diferentes valores, lo cual tiene sentido pues estas tres muestras fueron tomadas de un pozo a diferentes profundidades en una terraza maluvial del río Cauca.

Muestra Porcentaje de m.o. (%)

N1 7,88

N2 8,36

N3 5,13

B1 4,30

B2 4,79

L 4,21

Evaluation of the Potential for Riverbank Filtration in Colombia

168

Determinación de Carbonato de Calcio

La determinación del porcentaje de carbonatos presentes en la muestra se realiza mediante un equipo llamado calcímetro de Bernard (Figura 2), el cual requiere de una calibración manual. Dicha calibración se debe de realizar cada vez que se inicie un proyecto nuevo, debido a que se pueden generar imprecisiones por evaporación de la columna de agua o por cambios en la presión manométrica del calcímetro.

Esta calibración consiste en tomar muestras de CaCO3 de alta pureza, con pesos que van desde 0.1 gramos hasta 0.01 gramos, en intervalos de 0.01 gramos; cada una de estas fracciones se pone a reaccionar con ácido clorhídrico (HCl), el cual tiene una concentración del 10% en volumen. El objetivo es, que la columna de gas generado durante la reacción del CaCO3 con el HCl desplace la columna de agua que se encuentra al interior de la bureta. Este procedimiento se realiza por duplicado para las 10 muestras de CaCO3, dando un total de 20 puntos para la calibración.

Figura 2. Calcímetro de Bernard.

APPENDIX E

169

La tabla 4 muestra los datos obtenidos durante la calibración, del desplazamiento de la columna de agua, para cada muestra de CaCO3.

Tabla 4. Tabla de calibración del calcímetro.

Con los datos obtenidos, se realizó una gráfica de porcentaje de CaCO3 (% de CaCO3) versus distancia desplazada en el calcímetro (Dist.), y de esta manera obtener la recta que mejor se ajustara a la serie de puntos, de igual manera se obtuvo la ecuación de dicha recta y el valor de R2 (Figura 3).

Figura 3. Recta de calibración para los valores obtenidos.

Muestra 1 Muestra 2

0,1 50 19,1 19,1

0,09 45 16,9 16,8

0,08 40 14,7 14,7

0,07 35 13,3 13,2

0,06 30 11,1 10,9

0,05 25 10,5 10,4

0,04 20 9,3 9,4

0,03 15 7,3 7,3

0,02 10 5,1 5,0

0,01 5 3,1 3,0

Distancia desplazada (cm)Peso CaCO3(g) Peso CaCO3 (%)

Tabla de calibración del calcímetro con CaCO3

y = 2,5767xR² = 0,969

0

10

20

30

40

50

60

0 5 10 15 20 25

Muestra 1

Muestra 2

CaCO3 (%)

Dist. (cm)

Evaluation of the Potential for Riverbank Filtration in Colombia

170

Una vez realizada la calibración, se procedió a determinar el porcentaje de carbonatos para cada una de las muestras objeto de este estudio, utilizando el procedimiento previamente descrito. La tabla 5 muestra la distancia desplazada en el calcímetro en ambos ensayos, la distancia promedio de ambos y los valores en porcentaje de CO3 obtenidos para las muestras N1, N2, N3, B1, B2 y L. El análisis de estas muestras y la correlación con otras variables se dejará para el final del texto.

Tabla 5. Valores de CO3 para las muestras analizadas.

PROCEDIMIENTOS ÓPTICOS

Después de haber realizado el tamizaje, se decidió que la búsqueda de sulfatos se realizaría en tres escalas diferentes, cada una con un método propio. Las escalas o tamaños de análisis escogidos fueron gravas, arenas y finos; es decir, que si utilizamos las mismas fracciones obtenidas en el tamizaje, la primera (> 2000 µm) serían las gravas, las 5 siguientes (2000 µm – 63 µm) corresponderían a las arenas, y la última fracción (< 63 µm) serían los finos (limos y arcillas).

Para buscar los sulfatos en los clastos tamaño grava, se utilizó una lupa electrónica, binocular, con luz dirigida; mientras que para los tamaños arena se decidió elaborar secciones delgadas par ser observadas bajo un microscopio petrográfico. La fracción fina sería vista en el microscopio electrónico de barrido; sin embargo, este último ítem no se logró realizar por falta de recursos, por lo que no se presentará en este informe.

Análisis de los clastos tamaño grava

Como se mencionó antes, la inspección de las gravas se realizó con la ayuda de una lupa electrónica. A continuación se describe lo encontrado para cada muestra.

Nacho 1 (N1):

Fragmentos de cuarzo lechoso, y rocas ígneas (basaltos) y metamórficas (esquistos y gneises) bien redondeados y con esfericidad buena. No se observan otro tipo de materiales, en particular para el caso de los sulfatos, no se observa ningún ejemplar de esta familia mineral.

Muestra Dist. desplazada 1 (cm) Dist. desplazada 2 (cm) Dist. prom. Porcentaje de CO3 (%)

N1 4,5 4,6 4,55 11,72

N2 2,7 2,6 2,65 6,83

N3 1,5 1,5 1,50 3,87

B1 1,4 1,4 1,40 3,61

B2 1,5 1,4 1,45 3,74

L 1,4 1,3 1,35 3,48

APPENDIX E

171

Nacho 2 (N2):

Fragmentos de cuarzo lechoso, y rocas ígneas (basaltos) y metamórficas (esquistos y gneises) bien redondeados y con esfericidad buena. No se observan otro tipo de materiales, en particular para el caso de los sulfatos, no se observa ningún ejemplar de esta familia mineral.

Nacho 3 (N3):

Fragmentos de cuarzo lechoso, y rocas ígneas (basaltos) y metamórficas (esquistos y gneises) bien redondeados y con esfericidad buena. No se observan otro tipo de materiales, en particular para el caso de los sulfatos, no se observa ningún ejemplar de esta familia mineral.

Banca 1 (B1):

Fragmentos de cuarzo lechoso, muy bien redondeados y con muy buena esfericidad; fragmentos de rocas sedimentarias y metamórficas con redondez media y esfericidad baja. No se observan otro tipo de materiales, en el caso particular de los sulfatos, no se observan ejemplares de esta familia mineral.

Banca 2 (B2):

Fragmentos de cuarzo lechoso, muy bien redondeados y con muy buena esfericidad; fragmentos de rocas sedimentarias y metamórficas con redondez media y esfericidad baja. No se observan otro tipo de materiales, en el caso particular de los sulfatos, no se observan ejemplares de esta familia mineral.

Lecho (L):

Fragmentos de cuarzo lechoso, muy bien redondeados y con muy buena esfericidad; fragmentos de rocas metamórficas en mayor abundancia que en las muestras anteriores con redondez y esfericidad bajas. No se observan otro tipo de materiales, en el caso particular de los sulfatos, no se observan ejemplares de esta familia mineral.

Como se puede leer en las descripciones de las muestras, ningún tipo de sulfato se encontró en esta fracción, por lo que se procedió a continuar con la siguiente fracción.

Análisis de los clastos tamaño arena

Para esta fracción se realizaron cuatro secciones delgadas de las muestras N1, B1, B2 y L, debido a falta de recursos no se pudieron elaborar las secciones correspondientes a las muestras N2 y N3. Las secciones fueron analizadas en un microscopio petrográfico perteneciente al Laboratorio de Petrografía de la Universidad Nacional de Colombia – Sede Medellín; a continuación se presenta la descripción de cada una de ellas.

Evaluation of the Potential for Riverbank Filtration in Colombia

172

Nacho 1 (N1):

El tamaño de los sedimentos es arena fina. Los sedimentos son muy variados, su esfericidad es muy baja, al igual que su redondez. La selección de los granos es muy mala.

Los granos son principalmente de cuarzo, plagioclasa, feldespato, piroxenos, biotita y fragmentos de rocas ígneas, plutónicas y volcánicas, sedimentarias y en menor proporción aparecen fragmentos de rocas metamórficas.

Todos los granos de minerales se presentan aislados en la resina, a excepción del cuarzo y el piroxeno, los cuales además de presentarse aislados se observan también en agregados de cristales.

Los fragmentos de rocas ígneas plutónicas son de composición intermedia, al igual que las volcánicas; los de rocas sedimentarias por su parte corresponden a areniscas muy finas o limolitas, mientras los de rocas metamórficas que se observan corresponden a esquistos.

No se encontraron granos de yeso (primer barrido), o de algún otro tipo de sulfato en la muestra (segundo barrido).

Banca 1 (B1):

El tamaño de los sedimentos es arena fina. Los sedimentos son de composición muy variada, con esfericidad media en los granos más pequeños y menores en los granos más grandes; la redondez de los granos es baja y la selección es muy mala.

Los granos son principalmente de cuarzo, plagioclasa, feldespato, piroxenos, moscovita, biotita y fragmentos de rocas sedimentarias y metamórficas en mayor proporción.

Todos los granos de minerales están aislados a excepción del cuarzo y el piroxeno, los cuales se observan en agregados cristalinos.

Los fragmentos de roca son un poco más redondeados. Ambos tipos de rocas sedimentarias y metamórficas son muy ricas en cuarzo, y corresponden a areniscas cuarzosas de grano fino, y gneises con intercalaciones de moscovita, respectivamente.

No se encontraron granos de yeso (primer barrido), o de algún otro tipo de sulfato en la muestra (segundo barrido).

Banca 2 (B2):

El tamaño de los sedimentos es arena fina. Los granos se observan de composición muy variada, la redondez y la esfericidad de los granos de minerales es muy baja, mientras que en los granos de fragmentos de roca, que son de un tamaño un poco mayor, la redondez es alta y la esfericidad es media.

Los granos son principalmente de cuarzo, feldespato, plagioclasa, glauconita, piroxenos, moscovita, biotita y fragmentos de rocas sedimentarias y metamórficas.

APPENDIX E

173

Como se mencionó anteriormente los granos de minerales son más pequeños que los de fragmentos de roca. No se observan agregados de piroxenos como en las muestras anteriormente descritas, pero sí de cuarzo.

Los fragmentos de rocas sedimentarias corresponden a cuarzo arenitas de grano fino, mientras que los de rocas metamórficas corresponden a gneises cuarzo-sericíticos.

No se encontraron granos de yeso (primer barrido), o de algún otro tipo de sulfato en la muestra (segundo barrido).

Lecho (L):

El tamaño de los sedimentos es arena fina. Los granos son de composición muy variada y se presentan en diferentes tamaños dentro de este rango. La redondez y esfericidad de los granos es baja, la selección de los granos es muy mala.

Los granos son principalmente de cuarzo, feldespato, plagioclasa, piroxeno, moscovita, biotita y fragmentos de roca, siendo estos últimos más abundantes que los minerales.

Los granos de minerales se observan aislados en la resina, a excepción del cuarzo y el piroxeno que se observan en agregados cristalinos.

Los fragmentos de roca corresponden a rocas ígneas plutónicas y volcánicas, rocas sedimentarias y rocas metamórficas. Las rocas ígneas son ambas de composición intermedia, las rocas sedimentarias son cuarzoarenitas de grano fino y las rocas metamórficas son esquistos moscovíticos y gneises cuarzo-sericíticos.

No se encontraron granos de yeso (primer barrido), o de algún otro tipo de sulfato en la muestra (segundo barrido).

Al igual que para la fracción grava, en esta fracción arena tampoco se encontraron sulfatos con lo cual establece que por lo menos en estas dos fracciones no hay presencia de sulfatos.

ANÁLISIS DE RESULTADOS

Las muestras estudiadas presentan algunas pequeñas correlaciones que aunque no sean muy importantes para el objetivo con el cual se inició este estudio, de encontrar sulfatos, pues son importantes de mencionar, para futuros análisis.

Lo primero que hay que anotar es que en el análisis granulométrico las muestras tienen una gran similitud, pues todas comparten una naturaleza de transporte caótica, que les da esa característica de ser pobremente seleccionadas y tener variedad de materiales. Los medios que depositan estos sedimentos son de alta energía, pues solo un afluente con suficiente capacidad de carga puede transportar tal variedad de materiales y tamaños, aunque al mismo tiempo la

Evaluation of the Potential for Riverbank Filtration in Colombia

174

redondez y esfericidad de los clastos reflejan que sufrieron procesos de transporte largos.

La figura 1 muestra que las distribuciones granulométricas de N1, N2 y N3 son muy similares, lo que confirma que provienen de la misma fuente; mientras que B1 y B2 tienen una tendencia a tener más finos que las demás muestras, probablemente porque correspondan a las paredes de la quebrada, lo que puede significar que en ellas se deposita mayor cantidad de material que viaja en suspensión, mientras que el lecho de la quebrada muestra una mayor cantidad de arenas, lo que también concuerda con la idea de que por el pasa mayor material de arrastre.

En los análisis químicos realizados a las muestras, se puede notar cierta similitud entre el contenido de materia orgánica y el de carbonatos, como parámetros que confirman la proveniencia de las muestras. Para las muestras N1, N2 y N3, los porcentajes de materia orgánica son muy diferentes, sin embargo esto puede corresponder a que las muestras fueron tomadas a distintas profundidades en la terraza aluvial, lo cual puede significar una variación en el aporte de materia orgánica por parte del río mismo. Algo similar ocurre con el contenido de carbonatos, pues este parámetro disminuye a medida que las muestras son de mayor profundidad, lo que también indica una variación en el aporte de iones carbonato por parte del río.

Para las otras muestras, B1, B2 y L, se observa una alta similitud entre los valores de las tres, en ambas variables; es decir que los valores de materia orgánica estas muestras son muy similares entre sí, al igual que los valores de carbonatos, todas estas similitudes solo se pueden interpretar como una confirmación de que las muestras pertenecen a la misma fuente.

En la parte correspondiente a los análisis ópticos, no hay mucho que decir aparte de lo que ya se enunció en el capítulo correspondiente. Las muestras de gravas y arenas analizadas cada una con su respectivo método, no mostraron presencia de sulfatos en ningún tamaño; y aunque solo faltó analizar la fracción de finos por los motivos expuestos, se considera poco probable la existencia de estos minerales en ese tamaño.

Las conclusiones obtenidas después de elaborar este informe son:

Los análisis demostraron que la proveniencia de las muestras era de la misma fuente y que además las muestras correspondientes a ambientes similares tenían características que las agrupaban entre si y permitían diferenciarlas de las demás.

No se encontraron sulfatos en la muestra en las dos fracciones analizadas (gravas y arenas), sin tener certeza de lo que pueda haber en la fracción fina.

Se recomienda buscar los sulfatos en la fracción fina para tener total certeza de su ausencia en las muestras estudiadas, o de comprobación de existencia en esta fracción de tamaño.

175

APPENDIX F Results of the water analysis in Guarne, performed at the lab in CORNARE

Cornare ID 2010-08-1807 2010-08-1808 2010-09-2056 2010-09-2057

Sampling Date 11-Aug-10 11-Aug-10 14-Sep-10 14-Sep-10

Parameter River Well River Well

Bicarbonates (mg/L HCO3) 29.15 94.92 28.02 92.72

Calcium (mg/L Ca) 3.80 14.20 2.69 28.30

Chloride (mg/L Cl) 7.50 0.75 5.00 <0.68

Conductivity (μS/cm) 75.8 147.0 67.5 154.0

Dissolved Oxygen (mg/L O2) 6.60 3.10 6.50 1.50

E. Coli (CFU/100mL) 5,200 100 5,200 50

Magnesium (mg/L Mg) 1.93 3.36 1.60 3.13

Manganese (μg/L Mn) Not measured Not measured Not measured Not measured

Nitrites + Nitrates (mg/L N-) 0.317 0.138 0.291 0.158

pH (on-site) 6.48 5.56 6.42 5.59

Potassium (mg/L K) 0.842 1.850 0.715 1.090

Sodium (mg/L Na) 8.51 6.83 6.46 4.79

Soluble Phosphorous (mg/ P) Not measured Not measured Not measured Not measured

Sulphates (mg/L (SO4)2) 7.280 1.450 3.910 <0.391

T°C (on-site) 18.0 21.0 17.0 19.0

Total Alcalinity (mg/L CaCO3) 23.89 77.80 22.97 76.00

Total BOD (mg/L O2) 1.80 <1.43 1.80 <1.43

Total COD (mg/L O2) 11.10 <10.82 17.10 <10.82

Total Coliforms (CFU/100mL) 9,900 200 9,950 1,450

Total Dissolved Solids (m/-L) 62.90 128.80 51.60 119.28

Total Hardness (mg/L CaCO3) 26 69 33 67

Total Iron (mg/L Fe) Not measured Not measured Not measured Not measured

Total Phosphates (mg/L (PO4)3) Not measured Not measured Not measured Not measured

Total Phosphorous (mg/L P) Not measured Not measured Not measured Not measured

Total Solids (mg/L) 87.7 132.0 110.0 120.0

Total Suspended Solids (mg/L) 24.80 3.20 58.40 0.72

Total volatile solids (mg/L) 17.0 15.0 37.3 32.0

True Color (UPC) 100 <2.5 350 <2.5

Turbidity (NTU) 10.10 0.72 26.80 0.68

Volatile Suspended Solids (mg-L) 6.2 1.3 12.9 0.7

Evaluation of the Potential for Riverbank Filtration in Colombia

176

Cornare ID 2010-09-2201 2010-09-2202 2010-10-2341 2010-10-2342

Sampling Date 28-Sep-10 28-Sep-10 13-Oct-10 13-Oct-10

Parameter River Well River Well

Bicarbonates (mg/L HCO3) 23.67 89.16 15.53 86.68

Calcium (mg/L Ca) 2.58 33.40 3.24 6.43

Chloride (mg/L Cl) 3.74 <0.68 4.49 <0.68

Conductivity (μS/cm) 56.3 141.1 63.2 147.8

Dissolved Oxygen (mg/L O2) 6.00 2.70 6.30 1.90

E. Coli (CFU/100mL) 4,850 50 5,000 50

Magnesium (mg/L Mg) 2.08 3.05 0.29 2.72

Manganese (μg/L Mn) 37.0 36.5 59.8 26.5

Nitrites + Nitrates (mg/L N-) 0.280 0.120 0.308 0.131

pH (on-site) 6.12 5.58 7.1 5.57

Potassium (mg/L K) 1.010 0.845 0.271 1.590

Sodium (mg/L Na) 5.16 4.94 5.70 5.10

Soluble Phosphorous (mg/ P) 0.105 0.105 0.590 0.105

Sulphates (mg/L (SO4)2) 2.550 0.991 3.700 0.749

T°C (on-site) 17.0 18.8 17.2 21.3

Total Alcalinity (mg/L CaCO3) 19.40 73.08 12.73 71.05

Total BOD (mg/L O2) 2.10 <1.43 1.70 <1.43

Total COD (mg/L O2) <10.82 <10.82 <10.82 <10.82

Total Coliforms (CFU/100mL) 9,800 600 10,000 4,800

Total Dissolved Solids (m/-L) 50.90 128.08 53.00 129.28

Total Hardness (mg/L CaCO3) 24 64 23 65

Total Iron (mg/L Fe) 6.370 0.179 2.390 0.142

Total Phosphates (mg/L (PO4)3) 0.026 0.031 0.013 0.074

Total Phosphorous (mg/L P) 0.105 0.105 Not measured Not measured

Total Solids (mg/L) 121.0 129.0 77.1 130.0

Total Suspended Solids (mg/L) 70.10 0.92 24.10 0.72

Total volatile solids (mg/L) 39.4 30.6 19.7 29.8

True Color (UPC) 100 5 50 5

Turbidity (NTU) 37.40 0.95 20.50 1.12

Volatile Suspended Solids (mg-L) 13.7 0.9 5.8 0.7

APPENDIX F

177

Cornare ID 2010-10-2503 2010-10-2504

Sampling Date 27-Oct-10 27-Oct-10

Parameter River Well

Bicarbonates (mg/L HCO3) 28.65 99.65

Calcium (mg/L Ca) 4.38 6.42

Chloride (mg/L Cl) 5.24 1.25

Conductivity (μS/cm) 70.9 154.0

Dissolved Oxygen (mg/L O2) 6.43 1.54

E. Coli (CFU/100mL) 3,000 50

Magnesium (mg/L Mg) 2.07 1.58

Manganese (μg/L Mn) 79.7 39.0

Nitrites + Nitrates (mg/L N-) 0.180 0.135

pH (on-site) 6.4 5.53

Potassium (mg/L K) 0.505 0.099

Sodium (mg/L Na) 4.97 4.65

Soluble Phosphorous (mg/ P) 0.105 0.105

Sulphates (mg/L (SO4)2) 2.820 0.660

T°C (on-site) 16.6 19.0

Total Alcalinity (mg/L CaCO3) 23.48 81.68

Total BOD (mg/L O2) 1.73 <1.43

Total COD (mg/L O2) <10.82 <10.82

Total Coliforms (CFU/100mL) 8,000 2,550

Total Dissolved Solids (m/-L) 69.64 137.28

Total Hardness (mg/L CaCO3) 17.8 65

Total Iron (mg/L Fe) 2.860 0.171

Total Phosphates (mg/L (PO4)3) 0.028 0.106

Total Phosphorous (mg/L P) Not measured Not measured

Total Solids (mg/L) 88.2 138.0

Total Suspended Solids (mg/L) 18.56 0.72

Total volatile solids (mg/L) 29.1 23.2

True Color (UPC) 100 <2.5

Turbidity (NTU) 12.90 0.37

Volatile Suspended Solids (mg-L) 5.5 0.7