Assessment of the riparian condition of two tributary ...

21
Assessment of the riparian condition of two tributary rivers in an agricultural- dominated landscape in the middle reaches of the Magdalena River, Santander, Colombia Maria Alejandra Ariza Rojas Department of Biological Sciences Universidad de los Andes, Bogotá, Colombia ABSTRACT The degradation of riparian zones in response to a growing demand for livestock and agriculture areas in the middle reaches of the Magdalena River in Colombia is a current pressing conservational issue. Thus, there is an urgent need for a rapid method of assessment and monitoring of riparian ecological condition that can be easily replicated and that can provide reliable information to conservational practitioners. The aim of this study was to assess the applicability of the Australian Tropical Rapid Appraisal of Riparian Condition (TRARC) method to characterize the riparian zones of two tributary rivers (San Juan and Carare) and to explore the main causes of ecological degradation. Results show a median-to- good overall riparian condition and moderate pressure levels at both tributaries. The San Juan River was characterized by greater regeneration and plant coverage, while the Carare River was mostly defined by human-derived pressure variables. Multivariate analysis further showed that only 15 out of the original 36 TRARC variables were needed to fully characterize the total variation of the riparian condition in the evaluated area. This study highlights the suitability of applying a modified TRARC method to assess lowland riparian condition in the Magdalena tributaries of Colombia. Key Words: Riparian zones, Colombian Magdalena River, ecological condition assessment, TRARC methodology, PCA. INTRODUCTION The riparian zone is the ecotone between terrestrial and freshwater ecosystems (Richardson et al. 2017). It encompasses the space between flowing water at low levels and the highest watermark where vegetation is influenced by floods, elevated water tables, and soil type (González et al. 2017). The riparian zones offer multiple ecological and environmental functions that are key to ecosystem, economic and social services. For example, the riparian ecotone promotes high plant productivity through the entrapment of organic matter and nutrients derived from overland water runoff and groundwater inputs, and through the entrapment of viable flood-dispersing seeds (Dixon & Douglas 2006). Similarly, the riparian zone provides water flow regulation, help stabilize riverbanks, reduce soil erosion, control water and terrestrial temperature, and purify water (Dixon & Douglas 2006). It also provides habitat structure, refuge and movement corridors for aquatic and terrestrial animals (Dixon &

Transcript of Assessment of the riparian condition of two tributary ...

Page 1: Assessment of the riparian condition of two tributary ...

Assessment of the riparian condition of two tributary rivers in an agricultural-dominated landscape in the middle reaches of the Magdalena River, Santander,

Colombia

Maria Alejandra Ariza Rojas

Department of Biological Sciences Universidad de los Andes, Bogotá, Colombia

ABSTRACT The degradation of riparian zones in response to a growing demand for livestock and agriculture areas in the middle reaches of the Magdalena River in Colombia is a current pressing conservational issue. Thus, there is an urgent need for a rapid method of assessment and monitoring of riparian ecological condition that can be easily replicated and that can provide reliable information to conservational practitioners. The aim of this study was to assess the applicability of the Australian Tropical Rapid Appraisal of Riparian Condition (TRARC) method to characterize the riparian zones of two tributary rivers (San Juan and Carare) and to explore the main causes of ecological degradation. Results show a median-to-good overall riparian condition and moderate pressure levels at both tributaries. The San Juan River was characterized by greater regeneration and plant coverage, while the Carare River was mostly defined by human-derived pressure variables. Multivariate analysis further showed that only 15 out of the original 36 TRARC variables were needed to fully characterize the total variation of the riparian condition in the evaluated area. This study highlights the suitability of applying a modified TRARC method to assess lowland riparian condition in the Magdalena tributaries of Colombia. Key Words: Riparian zones, Colombian Magdalena River, ecological condition assessment, TRARC methodology, PCA. INTRODUCTION The riparian zone is the ecotone between terrestrial and freshwater ecosystems (Richardson et al. 2017). It encompasses the space between flowing water at low levels and the highest watermark where vegetation is influenced by floods, elevated water tables, and soil type (González et al. 2017). The riparian zones offer multiple ecological and environmental functions that are key to ecosystem, economic and social services. For example, the riparian ecotone promotes high plant productivity through the entrapment of organic matter and nutrients derived from overland water runoff and groundwater inputs, and through the entrapment of viable flood-dispersing seeds (Dixon & Douglas 2006). Similarly, the riparian zone provides water flow regulation, help stabilize riverbanks, reduce soil erosion, control water and terrestrial temperature, and purify water (Dixon & Douglas 2006). It also provides habitat structure, refuge and movement corridors for aquatic and terrestrial animals (Dixon &

Page 2: Assessment of the riparian condition of two tributary ...

Douglas 2006). Furthermore, it provides many services to society such as soil fertility, water purification, goods supply and recreation among others (González et al. 2017). Unfortunately, and despite all these welfares, riparian ecosystems are some of the most altered and degraded environments by human actions owing to their particularly vulnerable position in the landscape (Richardson et al. 2017). In fact, humans have shaped riparian landscapes since the beginning of human settlement in river valleys, making the riparian ecosystems hotspots of intensive human activity (Capon 2019). Anthropogenic pressures such as urban expansion, agriculture, mining, grazing, erosion and pollution are some of the many stress factors besieging the riparian zones and associated vegetation (Fu et al. 2017). For this reason, it is essential that riparian zones are correctly conserved and managed so that it is possible to prevent, or in some cases stop, degradation and damage that have become more frequent and evident in the past decades (Tabacchi et al. 1998).

Fig. 1. Examples of riparian zones along the tributaries of the San Juan River (left) and the Carare River (right) showcasing both ends of the human-derived pressures’ continuum observed by the median reaches of the Magdalena River. Patches along both rivers can show primary forest mostly untouched and unaltered (left) as well as areas completely annihilated for human landuse, primarily, cattle. The median reaches of the Magdalena River, Santander, Colombia, are naturally characterized by rich riparian cover and highly productive floodplains (Rincón 2009). Nevertheless, there is current evidence of various anthropogenic pressures like intense logging for livestock, agriculture and human housing, direct contamination of both aquatic and terrestrial ecosystems, and disturbance of the river fluent through human-derived infrastructures; all of which form a continuum of environmentally harmful pressures along the tributaries of the San Juan and Carare Rivers. Through satellite observation of the complete riparian coverage by use of Google Earth, for example, as well as direct presence in the field, it is possible to observe areas completely obliterated from use by cattle and large extensions of palm oil crops, as well as zones of preserved primary forest (Figure 1). Such agricultural activities are likely to further expand in the forthcoming years (Gobernación de

Page 3: Assessment of the riparian condition of two tributary ...

Santander 2017) and thus the degradation of the associated riparian zones. In response, there is a growing need for a rapid method to measure riparian condition in order to reinforce sustainable management and conservation strategies. Unfortunately, for the case of Colombia, although there are different methods to characterize vegetation of riparian ecosystems or for restoration of the landscape; none of these is known to be a standard and unified methodology, with quantitative, simple applicability and replicable features. The Tropical Rapid Appraisal of Riparian Condition methodology (2006), or TRARC method, is a visual assessment and monitoring tool designed for northern Australia’s riparian zones directly adjacent to rivers. This method could be of potential applicability for the Magdalena River case as it evaluates the overall ecological condition and identifies pressures through the assessment of multiple indicators (Dixon & Douglas 2006). The present study aims to 1) Assess the suitability of the Tropical Rapid Appraisal of Riparian Condition methodology (TRARC: Dixon & Douglas 2006) to evaluate the ecological condition of the riparian zone of the San Juan and Carare tributary rivers, two tributaries of the Magdalena River in Santander, Colombia; and 2) Identify the main environmental and human-derived factors affecting the natural dynamics of the riparian zone of the San Juan and Carare tributary rivers to guide future conservation and management programs. METHODOLOGY Implementation of the TRARC methodology In order to assess the suitability of the TRARC methodology to evaluate riparian condition, I applied the TRARC method across a series of transects in the shorelines of the San Juan and Carare Rivers in Santander, Colombia, during a fieldwork campaign from the 10th and 16th of January 2019. The TRARC method focuses on 100 m shore-line transects, each constituted by three assessment points separated by approximately 50 m among them, starting from the tributary’s mouth and following along the main stem. For the San Juan River the transects were performed upstream as far as the river’s fluent allowed for mobilization by boat (GPS coordinates ranging from N 6˚40’27.2’’ W 74˚10’36.2’’ to N 6˚44’20.4’’ W 74˚6’51.6’’), considering that data collection was carried out during the dry season. For the Carare River the total area sampled ranged between the coordinates N 6˚40’50.1’’ W 74˚5’35’’ to N 6˚46’26.1’’ W 74˚6’29.3’’. The limit of the sampled area in the headwaters was determined by a thorough evaluation of the main characteristics of the river riparian zone; i.e. when human-derived stressors in the headwaters were highly similar to those already observed among the sampling area. Mapping of all transects’ location is shown in Figure 2.

Page 4: Assessment of the riparian condition of two tributary ...

Fig. 2. Location of study areas. C-named sites correspond to the Carare River, SJ-named sites

correspond to the San Juan River. A series of riparian vegetation characteristics (i.e. canopy cover, dominant tree regeneration, canopy continuity), geomorphological variables (i.e. exposed soil, slumping, gullying) and human-derived activities (i.e. animals managed, tree clearing, instream structures) were recorded at each sampling point (see Table 1 for details). All TRARC score evaluations were conducted using a small boat at a distance of approximately 4 m due to unstable or unreachable land areas. Forty-two transects were assessed in total for both tributary rivers, 19 for the San Juan River and 23 for the Carare River (Table 2). All riparian indicators were measured in the field and subsequently introduced in the TRARC score sheet for calculation of the overall index for every study area. All indicators, or variables, are grouped into categories: Plant Cover, Regeneration, Erosion and Weeds. The given scores are summed together within their respective category and converted to a new score between 0–25; these can then be summed to give an overall Condition index of 0–100, where a higher score implies better condition. Additionally, the indicators grouped in the Pressure category are summed and converted to an index of 0–100, where a higher score implies greater threats (Dixon & Douglas 2006). In this manner, the TRARC methodology allows for quantitative calculation of riparian ecosystem condition along the tributary rivers of San Juan and Carare. For riparian condition comparison between both tributary rivers the algebraical mean of the Condition and Pressure indexes was calculated and statistically compared through a two-tailed Student’s t test using the R software.

Page 5: Assessment of the riparian condition of two tributary ...

Table 1. TRARC variables and corresponding description, ordered by category.

Category Variable Definition

Plant Cover

Canopy cover % cover of trees and tall shrubs >5 m in height. Look directly above you (approx. 5 m radius). Midstorey cover % cover of shrubs and small trees 1.5–5 m in height (natives and weeds).

Understory cover % cover of shrubs, sedges, herbs and groundcovers <1.5 m in height (natives and weeds). Do not include grass.

Grass cover % cover of grass of any height (natives and weeds). Organic litter % cover of leaves and fallen branches <10 cm diameter, do not include ash.

Logs Number of logs ( >10 cm diameter, 1-3 m in length) and large logs ( >10 cm diameter, >3 m in legth).

Canopy continuity Proportion of transect length that has a canopy. Gaps between canopies must be >5 m and span the width of the transect (max 20 m).

Regeneration

Canopy health Canopy health of surrounding native trees and tall shrubs >5 m in height. Look around area.

Tree size classes Variation in trunk width/height of dominant native trees >3 m tall. Size groups: <10 cm, 10–20 cm, 20–30 cm, 30–40cm, >40 cm.

Dominant tree regeneration Number of juveniles 0.3–3 m tall of dominant tree species. Must be same species as in Tree size classes.

Other tree regeneration Number of juveniles present that are common riparian species, even though adult individuals of these species are not dominant within the transect.

Large trees Number of large trees (native and alive) with >30 cm trunk diameter when measured 1.3 m from base of trunk, do not include dead or fallen trees.

Erosion

Exposed soil % cover of exposed soil and ash. Exclude large natural rock formations, boulders, leaf litter and roots.

Exposed tree roots Proportion of trees or tall shrubs with exposed tree roots (thicker than 20 mm) due to erosion. Do not include species with natural aerial roots.

Slumping Combined slumping width along 100 m transect. Gullying Combined width of active, unstable gullies passing through 100 m transect.

Undercutting Combined length of undercutting along 100 m transect.

Weeds

Midstorey weeds Proportion of weeds. Understorey weeds Proportion of weeds.

Grass weeds Proportion of weeds. Organic litter weeds Proportion of weeds (litter from weed plants).

Maximum bank sediment size Maximum bank sediment size.

Dominant bank sediment size Dominant bank sediment size. Bank stability: bank slope Approximate bank slope.

High impact weeds Grass coverture. High impact weed distribution Grass coverage distribution pattern.

Canopy weeds Proportion of canopy plants (>5 m tall) that are weeds, including vines in the canopy.

Pressure

Animals: managed Extent of damage (tree ringbarking; vegetation trampling; grazing; wallowing; soil compaction; track formation; instream substrate disturbance) caused by managed animals (i.e. cattle).

Animals: unmanaged Extent of damage (tree ringbarking; vegetation trampling; grazing; wallowing; soil compaction; track formation; instream substrate disturbance) caused by wild unmanaged animals.

Fire Average fire impact - DOES NOT APPLY for Colombia.

Tree clearing Average buffer width (measured away from waterway) and average clearing width compared to riparian width.

Flow regime: large dams Large dam upstream and vegetation response to environmental flows.

Bank stability: instream structures Human-built instream structures within 200 m upstream or downstream of transect.

Others Proportion of transect impacted by human structures or activities that have not yet been recorded.

Page 6: Assessment of the riparian condition of two tributary ...

Data analysis

Principal component analysis of main TRARC indicators A Principal Component Analysis was performed on all the variables (indicators) evaluated through the TRARC score sheet. The analysis was run in the R software using the packages stats, FactoMineR and cluster. PCA is a dimensionality reduction method for multivariate data (STHDA, 2019); in this case, the TRARC score sheet, which evaluates a total of 36 inter-correlated variables associated with riparian vegetation ecological condition. In order to build a parsimonious model, analysis results were thoroughly evaluated to estimate the contribution of each study variable and to eliminate redundant and uninformative variables. In the PCA, each indicator is considered as a different dimension so that PCA works as a linear combination of all continuous variables; by extracting non-redundant information the analysis expresses all of the given data as a set of two new dimensions, or principal components, through minimization of the projection error and simultaneous maximization of the variance of the projected data (STHDA 2019). In this manner the PCA allows for dimensionality reduction, identification of correlated variables and hidden patterns in a data set. This analysis was used to reduce the number of TRARC indicators so as to evaluate the most important features characterizing each assessed transect; therefore, PCA can be applied to identify with greater precision the exact variables describing the riparian vegetation condition. The PCA results are expressed in various features; such as eigenvalues, quality of representation (cos2) and contributions of variables to each principal component, which allow for correct and thorough interpretation of the dimensionality reduction process. Since the eigenvalues measure the amount of variance retained by each dimension (and remembering that the purpose of the PCA is to maximize variance), by examining the eigenvalues for all variables it was possible to determine the number of principal components to be considered out of the original 36. With this in mind, a value > 0.3 was used as a cutoff point for which PCs were retained, associated to approximately 90% of cumulative percentage of the variance explained, so that greater information could be retained despite elimination of TRARC indicators This first step allows for reduction of an overall quantity of dimensions that are necessary for ecological condition assessment. The next step was to determine which TRARC variables highly characterized the principal components chosen; for that, the quality of representation for all variables to each dimension (cos2) and the contribution value were extracted from the PCA results. For both Dim1 and Dim2 components, a variable with a contribution larger than 3.6% was emphasized as important; considering that if the contribution of the variables were uniform the expected value would be the inverse of the total of considered variables (1/36). Similarly, the cos2 for each variable was extracted and compared to the contribution value; with high cos2 indicating a good representation of the variable on the given principal component. By careful evaluation of all these descriptive PCA features, selective elimination of redundant variables was conducted and only those variables with all three features (eigenvalue, cos2 and contribution value) supporting a significant variance contribution to the first and second principal components were retained.

Page 7: Assessment of the riparian condition of two tributary ...

Finally, in order to quantify the statistical efficiency of the selected PCA parsimonious model in comparison to the original full PCA model, a prediction test was ran using the same R Software packages on a subset data matrix extracted from the whole and unaltered sample data; with 75% of data for training and 25% for prediction.

Identification of clusters by common condition assessment After the TRARC variables were evaluated and reduced based on their contribution to variance explained, the next step was to group all assessed transects of both tributary rivers by their similarities in overall ecological condition using the R cluster package. Each cluster is assigned a number of characterizing variables from the parsimonious PCA model to identify the specific riparian condition and pressure features affecting the associated riparian zones. Furthermore, resulting clusters were geographically mapped so that observation of possible differentiating riparian condition between the San Juan River and the Carare River could be easily spotted. For better visualization of the summarized results and further support of the clusters formed and their respective characteristic variables, a Multiple Factor Analysis (MFA) was also conducted. In addition, as previously performed, a statistical comparison of the overall riparian condition scores was made through a Student’s t test; but this time, instead of comparing the San Juan River against the Carare River transects, the differentiation was evaluated among clusters. This was made in order to statistically estimate the accuracy of clustering differentiation. RESULTS Implementation of the TRARC methodology A recompilation of the categories’ scores and final Condition and Pressure indexes for each evaluated transect of the San Juan River and the Carare River are presented in Table 2. By comparing the overall riparian condition score between the tributary rivers of San Juan and Carare, there was no statistical difference found for the Condition index (C mean = 46.04 ±8.12, SJ mean = 52.16 ±8.10); even so, the Pressure index was statistically higher for the Carare tributary river relative to the San Juan (30.74 ±9.48 and 27.79 ±8.19 respectively, P-value < 0.05).

Page 8: Assessment of the riparian condition of two tributary ...

Table 2. Principal index scores of assessed sites.

Stream name

Site Number GPS Coordinates

Plant Cover Regeneration Erosion Weeds Condition Pressure

Wild Animal sittings

San Juan SJ1 N 6˚43'21.3'' W 74˚8'1.9'' 3 3 20 21 30 46

SJ2 N 6˚44'20.4'' W 74˚6'51.6'' 9 8 15 18 27 50

SJ3 N 6˚44'21.5'' W 74˚6'58.2'' 3 7 20 19 26 49

SJ4 N 6˚44'14.3'' W 74˚7'4.7'' 13 11 15 19 34 57

SJ5 N 6˚44'2.6'' W 74˚7'16.3'' 5 7 18 17 49 47

SJ6 N 6˚43'53.7'' W 74˚7'17.8'' 11 10 16 17 38 53

SJ7 N 6˚40'27.2'' W 74˚10'36.2'' 1 0 17 19 39 37

SJ8 N 6˚40'47.2'' W 74˚10'13.5'' 8 3 16 19 26 46

SJ9 N 6˚41'2.2'' W 74˚9'42.3'' 13 14 11 20 27 58

SJ10 N 6˚41'34.9'' W 74˚9'26.5'' 17 15 14 19 23 65 Ateles

SJ11 N 6˚41'51.8'' W 74˚9'17.1'' 15 15 15 20 17 65 Ateles

SJ12 N 6˚42'7.8'' W 74˚8'53.6'' 18 15 14 18 17 65

SJ13 N 6˚42'31.1'' W 74˚8'43.1'' 16 11 15 19 18 62

SJ14 N 6˚42'46.9'' W 74˚8'30.2'' 7 8 9 21 26 46

SJ15 N 6˚42'59.2'' W 74˚8'16.7'' 9 9 9 20 28 47

SJ16 N 6˚43'40.5'' W 74˚7'24.3'' 6 4 16 18 33 44

SJ17 N 6˚43'31.7'' W 74˚7'34.4'' 9 10 13 19 22 50

SJ18 N 6˚43'32.7'' W 74˚7'48.8'' 9 9 18 21 20 57 Howler monkey

SJ19 N 6˚43'21.3'' W 74˚8'1.9'' 6 4 18 19 28 47 Ateles, Howler

monkey Carare C1 N 6˚43'12.7'' W 74˚8'4.5'' 5 7 15 19 40 46

C2 N 6˚44'49.7'' W 74˚6'28'' 1 3 13 19 38 36

C3 N 6˚44'47.2'' W 74˚6'1.2'' 3 1 14 19 29 37

C4 N 6˚44'26.8'' W 74˚8'42.9'' 2 0 19 18 51 39

C5 N 6˚44'7.1'' W 74˚5'39.6'' 10 7 18 19 40 53

C6 N 6˚44'13.3'' W 74˚6'16.9'' 7 6 16 19 33 49

C7 N 6˚43'43.7'' W 74˚6'42.4'' 8 7 13 20 27 48

C8 N 6˚43'27.6'' W 74˚6'28.4'' 3 2 15 19 28 39

C9 N 6˚43'19.1'' W 74˚6'3.9'' 7 4 14 18 27 43

C10 N 6˚40'50.1'' W 74˚5'35'' 11 14 19 21 17 65 White-faced

capuchin

C11 N 6˚41'27.9'' W 74˚5'35.8'' 5 3 13 19 22 39

C12 N 6˚41'37.1'' W 74˚5'18.9'' 5 4 13 16 45 38

C13 N 6˚41'48.3'' W 74˚5'44.9'' 10 13 24 20 15 66

C14 N 6˚42'7'' W 74˚5'56.2'' 4 6 12 20 35 41

C15 N 6˚42'26.4'' W 74˚5'24.5'' 5 8 24 22 17 59

C16 N 6˚42'49.4'' W 74˚5'31.1'' 6 3 17 18 32 44

C17 N 6˚42'45.6'' W 74˚6'4.1'' 5 3 15 20 31 44

C18 N 6˚44'26.8'' W 74˚6'51.7'' 6 5 15 20 26 45

C19 N 6˚44'54.9'' W 74˚6'49.7'' 11 6 12 17 18 46

C20 N 6˚45'5.3'' W 74˚6'23.5'' 4 3 19 19 39 46

C21 N 6˚45'33.8'' W 74˚6'27.4'' 10 7 14 19 32 49

C22 N 6˚46'3.3'' W 74˚6'14.4'' 2 3 20 19 40 44 C23 N 6˚46'26.1'' W 74˚6'29.3'' 2 4 19 18 25 43

Page 9: Assessment of the riparian condition of two tributary ...

Principal component analysis of the TRARC indicators The dimensionality reduction of the original variables of the TRARC methodology was conducted by considering analysis features as described in the Methodology section. In this way, Figures 3c and 4c show the percentage of total variance explained by each considered dimension (Table 3 summarizes the eigenvalues for all principal components of the original data and the final, reduced score sheet). Additionally, for both Dim1 and Dim2 components, all variables with a contribution value larger than 3.6% and high cos2 value were emphasized as important (variable is positioned close to the circumference of the correlation circle (Figures 3b and 4b)). This procedure was carried out until reaching a minimum of 15 variables out of an original 36 riparian condition indicators. Figures 3 and 4 show the differences between the first PCA run with all variables considered and the final reduced analysis, respectively. Knowing that by the characterization of Dixon & Douglas’ Field guide it is possible to classify all variables into one of five categories: Plant Cover, Regeneration, Erosion, Weeds and Pressure (the first four representing the Condition index); the final chosen variables and their respective category classification are shown in Table 4. PCA showed that these 15 variables of the TRARC score sheet retained sufficient information to describe and assess riparian vegetation of the study areas, as supported by the prediction test. For the prediction test, the coordinates of each variable based on eigenvectors for each dimension obtained from the original PCA (all 36 variables included) were compared to the eigenvectors of the reduced PCA (15 variables considered), since the coordinates of each point on a PCA plot indicate the riparian condition of the corresponding point. In this way, the test results between the parsimonious and the originals PCAs showed non-significant difference of the predicted riparian condition of the sampling sites in comparison to the original coordinates, which indicates a good estimation of the ecological condition for each study area based solely on the 15 chosen variables.

Page 10: Assessment of the riparian condition of two tributary ...

Table 3. PCA eigenvalues for each dimension considered. The corrected PCA refers to the reduced number of dimensions after election of significantly predicting variables.

PCA PCA corrected

standard

dev cumulative

% explained Eigen value

standard dev

cumulative % explained

Eigen value

Dim1 3.128 0.368 9.781 2.886 0.486 8.328 Dim2 1.794 0.489 3.218 1.665 0.648 2.773 Dim3 1.511 0.575 2.284 1.285 0.744 1.652 Dim4 1.377 0.646 1.897 1.120 0.817 1.254 Dim5 1.182 0.699 1.396 0.853 0.860 0.728 Dim6 1.071 0.742 1.147 0.761 0.894 0.580 Dim7 0.983 0.778 0.965 0.668 0.920 0.446 Dim8 0.922 0.810 0.850 0.579 0.939 0.335 Dim9 0.844 0.837 0.713 0.537 0.956 0.288 Dim10 0.793 0.861 0.629 0.462 0.969 0.214 Dim11 0.755 0.882 0.569 0.437 0.980 0.191 Dim12 0.637 0.897 0.406 0.351 0.987 0.123 Dim13 0.614 0.911 0.377 0.308 0.993 0.095 Dim14 0.605 0.925 0.366 0.279 0.997 0.078 Dim15 0.567 0.937 0.321 0.224 1.000 0.050 Dim16 0.548 0.949 0.301 Dim17 0.494 0.958 0.244 Dim18 0.460 0.966 0.211 Dim19 0.436 0.973 0.190 Dim20 0.412 0.979 0.170 Dim21 0.376 0.985 0.141 Dim22 0.321 0.988 0.103 Dim23 0.286 0.991 0.082 Dim24 0.275 0.994 0.076 Dim25 0.230 0.996 0.053 Dim26 0.190 0.998 0.036 Dim27 0.175 0.999 0.031 Dim28 0.117 0.999 0.014 Dim29 0.098 1.000 0.010 Dim30 0.069 1.000 0.005 Dim31 0.046 1.000 0.002 Dim32 0.030 1.000 0.001 Dim33 0.000 1.000 0.000 Dim34 0.000 1.000 0.000 Dim35 0.000 1.000 0.000 Dim36 0.000 1.000 0.000

Page 11: Assessment of the riparian condition of two tributary ...

Figure 3. PCA results for all 36 variables evaluated. (a) Variables expressed by cos2. (b) Contribution of each variable to dimension 1 and 2. (c) Variance (eigenvalues) explained by dimension, including only the first 10 components. The TRARC indicators associated to each variable are listed in Appendix Table 2.

Page 12: Assessment of the riparian condition of two tributary ...

Figure 4. Corrected PCA results for the 15 chosen variables. (a) Variables expressed by cos2. (b) Contribution of each variable to dimension 1 and 2. (c) Variance (eigenvalues) explained by dimension, including only the first 10 components. The TRARC indicators associated to each variable are listed in Appendix Table 2. Identification of clusters by common condition assessment Three main groups were identified using the resulting parsimonious PCA model, each defined by a unique set of variables and a principal riparian category (Table 4).

Page 13: Assessment of the riparian condition of two tributary ...

Table 4. Identification of clusters and association of evaluated transects by common riparian condition. Each cluster is characterized by a specific set of TRARC variables, each described by a riparian condition category.

Cluster Variable TRARC Indicator Category Stream name Site number

V14 Exposed soil Erosion San Juan SJ5:6, SJ14, SJ16, SJ19

V17 Bank stability: bank slope Pressure V22 Canopy weeds Weeds 1 V28 Animals managed Pressure Carare

“red” V31 Tree clearing Pressure C1:6, C8:9, C11:12, C14, C16:18, C20:23

V35 Animals Pressure V36 Bank stability Pressure 2 V24 Exposed tree roots Erosion San Juan SJ1:4, SJ7:8

“green” V25 Slumping Erosion Carare C7, C15

V1 Canopy cover Plant cover San Juan SJ9:13, SJ15, SJ17:18

V2 Canopy health Regeneration 3 V3 Tree size classes Regeneration Carare C10, C13, C19

“blue” V18 Large trees Regeneration V21 High impact weed distribution Weeds V23 Canopy continuity Plant cover

As shown in Table 4, each group is represented by a set of variables (considering the reduced dimension election performed in the previous section), each of which is associated to a riparian condition category. Plant cover and Regeneration categories indicate a good overall ecological condition of the riparian areas, mostly represented by high score values of canopy state and large tree presence. These variables characterize Cluster 3, shown as blue circles in Figure 5, and constitute the greatest percentage of the studied area of the San Juan River. Cluster 2, on the other hand, is defined by Erosion-like variables (exposed tree roots and slumping coverage) and composes most of the other transects of the San Juan River (shown as green circles in Figure 5). Finally, Cluster 1 mostly compromises variables associated to the Pressure category, indicating high degree of human-derived impact and, consequently, poor quality of riparian condition. This cluster includes 18 out of 23 assessed transects of the Carare River, displayed as not-encircled in Figure 5. The groups were mapped in order to describe the distribution and determine a possible geographical grouping pattern (Figure 5). As explained before, a clear differentiation can be seen between the tributary rivers, with the San Juan River mostly described by Erosion and Regeneration processes and the Carare River by human pressure variables.

Page 14: Assessment of the riparian condition of two tributary ...

Figure 5. Map of assessed transects associated to a specific cluster defined by riparian condition. (a) Carare River transects (b) San Juan River transects. Cluster 1 = not-circled, Cluster 2 = green circles, Cluster 3 = blue circles (see Table 3 for variable characterization of each cluster).

Once the clusters were defined, as defined in Table 4, statistical comparison of the riparian condition indexes of each individual cluster was performed. There was statistical difference for the Condition index for all clusters (Cluster 1 mean = 44 ±4.66, Cluster 2 mean = 49 ±6.84, Cluster 3 mean = 58.73 ±7.75; P-value < 0.05); while the Pressure index was statistically higher for Cluster 3 (Cluster 1 mean = 34 ±7.74, Cluster 2 mean = 28.25 ±6.45, Cluster 3 mean = 20.18 ±4.31; P-value < 0.05) relative to the other two clusters, as would be expected given its defining variables. After running the MFA, Figure 6b shows all TRARC variables (the original 36 indicators assessed) distributed according to their representation in each principal component, the color scheme indicates the contribution value of each variable to the main dimensions.

Page 15: Assessment of the riparian condition of two tributary ...

Consistently, Figure 6c shows a clearer alignment of the variables by consideration of their riparian condition categories, with the factor map colors highlighting the quality of representation (cos2) for each riparian condition category relative to the principal components. Both of these graphs indicate that the first dimension (PC1) is defined by Regeneration, Plant cover, Weeds and Pressure related variables (mostly vegetation-associated variables), while the second dimension (PC2) is defined mostly by Erosion-like variables. Lastly, Figure 6a summarizes transect grouping according to descriptive variables, considering only the chosen variables after the dimension-reduction process. The MFA allows for identification of variance representation between all original TRARC variables, with those most important represented in Figure 6. Here it is possible to visualize a segregation between the main variables describing the riparian condition for the Carare River against those for the San Juan River, with the former being mostly characterized by indicators such as canopy weeds, tree clearing and animals; and the latter segregated into two sub-groups: one described by canopy and tree linked indicators and the other by erosion related indicators (slumping and exposed tree roots). These variable groupings correspond to those previously identified by the clusters, and the river differentiation shown in Figure 6a also overlaps to the transect distribution seen in Figure 5.

Canopy-cover

Canopy continuity

Canopy-healthTree size-classes

Large trees

Canopy-weeds

Bank-stability:slope

exposed.soil

Exposed tree-roots

Slumping

Bank-stability

Animals-managedTree-clearing

animals

-1.0

-0.5

0.0

0.5

1.0

-1.0 -0.5 0.0 0.5 1.0Dim1 (47.1%)

Dim

2 (1

5.6%

)

5

10

15

contrib

Quantitative variables - MFA

SJ1

SJ2

SJ3 SJ4

Sj5SJ6

SJ7 SJ8

SJ9

SJ10

SJ11

SJ12SJ13

SJ14SJ15

S16SJ17

SJ18

SJ19C1

C2

C3

C4C5

C6

C7

C8

C9

C10

C11C12C13

C14

C15

C16C17

C18

C19

C20

C21

C22

C23

Carare

San Juan

Page 16: Assessment of the riparian condition of two tributary ...

Figure 6. (a) Variables are colored based on contribution value, which accounts for percentage of total variance explained by each variable to a given PC. Individuals (points) on the plot mark the transect sites, ‘SJ’ for the San Juan River and ‘C’ for the Carare River transects. Three clusters are identified based on transects with similar riparian condition: 1) Red circle: defined by Pressure-like variables (exposed soil, animals, tree clearing and bank stability), encompasses most of the Carare River transects. 2) Blue circle: defined by Plant cover and Regeneration variables (canopy condition and tree occurrence) which indicates good overall riparian condition. 3) Green circle: defined by Erosion-like variables (slumping and exposed tree roots). The last two clusters form sub-groups of the San Juan River. The projection of a variable vector onto the component axis represents the correlation between the variable and said component, with vector size indicating the quality of the variable on the factor map and direction marking correlation between variables (opposite variables are negatively correlated while positively correlated variables are grouped together). (b) Categories (vectors) are colored based on quality of representation ‘cos2’ which indicates how much a variable is represented on a given dimension; (c) individual variables are positioned based on association to categories and colored by ‘cos2’ value.

DISCUSSION

Firstly, it was possible to assess the suitability of the Tropical Rapid Appraisal of Riparian Condition methodology (TRARC: Dixon & Douglas, 2006) to easily and rapidly evaluate the ecological condition of the riparian zones of the San Juan and Carare tributary rivers of the Magdalena River in Santander, Colombia. The TRARC methodology is a good resource for monitoring and assessment of the riparian condition in Colombia, although some modifications to the variables most be considered. For example, eliminating pressure variables like ‘fire’, which does not naturally apply to the study regions in Colombian; or changing weed species and adding other more appropriate human impacts. Given the results obtained in this study, a more parsimonious and shorter TRARC model with only 15 variables can be applied while still retaining all accuracy of the original model. Nonetheless, by observing the actual mean values for the Condition and Pressure TRARC scores for each

canopy.cover

midstorey.cover

understorey.cover

grass.cover

organic.litter

canopy.continuity

logs

canopy.healthtree.size.classesdominant.tree.regeneration

other.tree.regeneration

large.treesmidstorey.weeds

understorey.weeds

grass.weeds

organic.litter.weedshigh.impact.weed.distribution

canopy.weeds

maximum.bank.sediment.size

dominant.bank.sediment.size

bank.stability..bank.slope

exposed.soil

exposed.tree.roots

slumping

gullying

undercutting

bank.stability

animals.managedtree.clearing

bank.stability..instream.structures

other

animals

-1.0

-0.5

0.0

0.5

1.0

-1.0 -0.5 0.0 0.5 1.0Dim1 (27.8%)

Dim

2 (1

1.5%

)

2

4

6

contrib

Quantitative variables - MFA

Plant.Cover

Regeneration

Erosion

Weeds

Condition

Pressure

-1.0

-0.5

0.0

0.5

1.0

-1.0 -0.5 0.0 0.5 1.0Dim1 (54.4%)

Dim

2 (2

1.9%

)

0.6

0.7

0.8

0.9

cos2

Variables - PCA

Page 17: Assessment of the riparian condition of two tributary ...

river, is not feasible to assign one unique conservation indicator that will define whether the studied area is or is not in a good ecological condition. This is because the difference among indexes between the San Juan River and the Carare River is only an overall approximation to the true state of the riparian ecosystems, and it must be considered that the TRARC scores by themselves will only give a general idea of the riparian condition. In this manner, both tributary rivers could be categorized under a mid-to-good condition for the San Juan River and mid-to-bad condition for the Carare River. This means that any action, present and future, taken in regard to the riparian ecosystems of the median reaches of the Magdalena River will determine de ecological state of these ecosystems and whether we can recover the forestall areas that have been lost or if, in the other hand, finish abolishing what still remains.

Secondly, according to the Principal Component Analysis, it was possible to describe the specific ecological features characterizing each tributary river, in order to complement (or interpret) the TRARC scoring results. Despite the fact that both tributaries showed median riparian condition, the San Juan River cluster grouping shows a better vegetation coverture but higher levels of erosion relative to the Carare River, whose cluster is characterized by pressure type variables. In this regard, the TRARC score sheet allows for a description of principal features characterizing the riparian vegetation but overall judgment of the conservation state most consider all possible variables, including field observations and every indicator by itself. The San Juan River presents greater vegetation coverage and more and larger tree size classes; however, it also shows a greater erosion value (defined by slumping and tree root exposure). The Carare River, in the other hand, is characterized by a larger exposed soil percentage and managed animal impact, which directly contribute to a greater Pressure score; as well as a more affected bank stability score (defined by slope, sediment type and size). This information, plus the observations made in the field, can lead to conclude that the San Juan River, although more eroded, has more tree coverage and thus a greater resilience capacity than the Carare River which is highly characterized by large areas of continuously distributed grass.

These differential characteristics among the tributaries of the San Juan River and the Carare River could be explained by various parameters, principally considering the difference between the river width which is greater for the Carare tributary than for the San Juan tributary. The San Juan is more prone to streambank erosion given the smaller river width and, thus, faster stream flow which influents greatly in the construction of meanders (Twidale 2004). Since the San Juan River is a more sinuous channel, erosion is concentrated in specific zones regularly spaced and alternating on opposite banks (Twidale 2004); which could explain the erosion values characterizing Cluster 2 and that are more present along the San Juan tributary in comparison to the Carare tributary. Similarly, the San Juan River shows a wider greenbelt relative to its clearing width than does the Carare River; this feature directly influents multiple plant cover and regeneration features such as canopy cover, canopy health and canopy continuity (these three variables belong to the more parsimonious TRARC model with 15 chosen features) which will provide a greater TRARC value to the riparian zones of the San Juan River. Parallelly, the Carare River is highly characterized by extensive grass coverage, and thus, a broad tree clearing in comparison to the San Juan River. This variable alone will bias the TRARC score to more pressure-like variables. This same feature of the

Page 18: Assessment of the riparian condition of two tributary ...

Carare River also allows for more presence of managed animals, more specifically, cattle. Because the TRARC score is highly sensitive to minor differences, which holds for its practical use in monitoring riparian vegetation, and adding the variable reduction process, which will give more score power to each variable, a marked difference between the tributary rivers of just a couple of specific variables (such as those previously explained) will lead to the segregating patterns described by the clusters.

Is important to note that the studied riparian zones were assessed during the dry season and thus the river flow and, consequently, degree of erosion are different throughout the year; thus the TRARC index must be repeated over time to monitor changes in vegetation condition and have a better assessment of the conservation state of both rivers. Still, the application of the method is simple and can be easily implemented twice a year (once during the dry season and once during the rainy season) by an organized local group, lead by a professional environmental engineer and constituted by representative members of the local community, as a monitoring tool for the riparian ecosystems along the median reaches of the Magdalena River. In this manner it could be possible to effectively asses and monitor riparian zones through time, and so determine the effect of human impact as well as follow possible future ecological restoration programs.

CONCLUSION

By the implementation of Dixon & Douglas’ 2006 TRARC score sheet it is possible to assess the ecological condition of riparian vegetation areas in Colombia’s middle reaches of the Magdalena River with a high confidence of evaluation of conservation and pressure states. In the end, only 15 out of the original 36 indicators of Dixon & Douglas’ 2016 TRARC methodology are sufficient to evaluate the condition and functioning of riparian zones, with both tributary rivers showing a median-to-good overall riparian condition and moderate pressure levels.

The San Juan River is characterized by greater regeneration and plan coverage relative to the Carare River riparian zones; however, specifically located erosion-defined areas are also present on the San Juan River. In the other hand, the Carare River is mostly characterized by human-derived pressure variables. By identifying the main indicators that describe the studied data set it is possible to perform a more precise evaluation of the features affecting the riparian vegetation of the San Juan and Carare Rivers and, thus, contemplate a conservation program targeted towards these factors.

ACKNOLEDGMENTS Special thanks to Fundación Primates and student of Biology Laura Lopera for all collaboration in the development of this study. Also, to professors Jorge Salgado and Andres Link for guidance and support throughout the study.

Page 19: Assessment of the riparian condition of two tributary ...

REFERENCES • Abdi, H. (2010). Principal Component Analysis: How to reveal the most

importantvariablesinyourdata?-Rsoftwareanddatamining.Retrievedfromhttp://www.sthda.com/english/wiki/print.php?id=204#variances-of-the-principal-components

• Analytics Vidhya Content. (2016, March 21). Practical Guide to PrincipalComponent Analysis (PCA) in R & Python. Retrieved fromhttps://www.analyticsvidhya.com/blog/2016/03/practical-guide-principal-component-analysis-python/

• Arango,C.;Dorado,J;GuzmanD.;Ruiz,J.F.ClimatologíaTrimestraldeColombia.Grupo de Modelamiento de Tiempo, Clima y Escenarios de Cambio ClimaticoSubdirecciondeMeteorologıa–IDEAM.

• BolinFu,YingLi,YeqiaoWang,BaiZhang,ShubaiYin,HongleiZhu,ZefengXing,Evaluationofecosystemservicevalueofriparianzoneusinglandusedatafrom1986 to 2012, Ecological Indicators, Volume 69, 2016, Pages 873-881, ISSN1470-160X,https://doi.org/10.1016/j.ecolind.2016.05.048.

• Cabrera, Y. P. (2017). Planes de Ordenamiento Territorial Modernos paraSantander. Retrieved fromhttp://www.santander.gov.co/index.php/actualidad/item/934-pot-modernos-para-santander

• Chundi Chen, ShengjunWu, Colin DouglasMeurk, MaohuaMa, Juanjuan Zhao,mingquan Lv, Xiaoxiao Tong, Effects of local and landscape factors on exoticvegetation in the riparian zone of a regulated river: Implications for reservoirconservation, Landscape andUrban Planning, Volume 157, 2017, Pages 45-55,ISSN0169-2046,https://doi.org/10.1016/j.landurbplan.2016.06.003.

• ClavijoOtálvaro, Karen Julieth,& LópezBarrera, Ellie Anne. (2017). Propuestametodológicaderestauraciónparalavegetaciónripariaapartirdelavariaciónde lacomposiciónflorísticaendiferentesépocasclimáticasdelhumedalTorca-Guaymaral.Producción + Limpia,12(1), 49-62.https://dx.doi.org/10.22507/pml.v12n1a5

• Dixon, I., Douglas, M., Dowe, J. & Burrows, D. (2006) Tropical Rapid Appraisal of Riparian Condition: Version 1 (for use in tropical savannas). River Management Technical Guideline No. 7. Canberra, Land & Water Australia.

• Eduardo González, María R. Felipe-Lucia, Bérenger Bourgeois, Bruno Boz,ChristerNilsson,GrantPalmer,AnnaA.Sher,Integrativeconservationofriparianzones, Biological Conservation, Volume 211, Part B, 2017, Pages 20-29, ISSN0006-3207,https://doi.org/10.1016/j.biocon.2016.10.035.

• Tabacchi, E., Correll, D., Hauer, R., Pinay, G., Planty-Tabacchi, A., Wissmar, R.,. (1998) Development, maintenance and role of riparian vegetation in the river landscape. Freshwater Biology, 40, 497-516.

• Facer,C.(2018,October23).HowtoCreateaCorrelationMatrixinR.Retrievedfromhttps://www.displayr.com/how-to-create-a-correlation-matrix-in-r/

Page 20: Assessment of the riparian condition of two tributary ...

• Friendly,M.(2002,August19).Corrgrams:Exploratorydisplays forcorrelationmatrices.Retrievedfromhttp://www.datavis.ca/papers/corrgram.pdf

• Fu, Bolin& Li, Ying&Wang, Yeqiao& Campbell, Anthony& Zhang, Bai& Yin,Shubai & Zhu, Honglei & Xing, Zefeng & Jin, Xiaomin. (2017). Evaluation ofriparian conditionof SonghuaRiverby integrationof remote sensingand fieldmeasurements.ScientificReports.7.10.1038/s41598-017-02772-3.

• G.L.McCloskey,R.J.Wasson,G.S.Boggs,M.Douglas,Timingandcausesof gullyerosion in the riparian zone of the semi-arid tropical Victoria River, Australia:Management implications, Geomorphology, Volume 266, 2016, Pages 96-104,ISSN0169-555X,https://doi.org/10.1016/j.geomorph.2016.05.009.

• Hoyle,D.(2008).AutomaticPCADimensionSelectionforHighDimensionalDataand Small Sample Sizes. Retrieved fromhttp://www.jmlr.org/papers/volume9/hoyle08a/hoyle08a.pdf

• InstitutoSinchi.2009.Bosquedegaleríayripario.FichasdelospatronesdelascoberturasdelatierradelaAmazoniaColombiana.BogotáD.C.

• Jaadi, Z. (2019, Feb 28). A step by step explanation of Principal ComponentAnalysis. Towards Data Science. Retrieved fromhttps://towardsdatascience.com/a-step-by-step-explanation-of-principal-component-analysis-b836fb9c97e2

• Jansen,Amy&Alistar,Robertson&Leigh,Thompson&Andrea,Wison.(2005).Rapid appraisal of riparian condition. River and Riparian Land ManagementTechnicalGuideline.4A.16p.

• Kassambara, A. (2017, September 23). PCA - Principal Component AnalysisEssentials. Retrieved from http://www.sthda.com/english/articles/31-principal-component-methods-in-r-practical-guide/112-pca-principal-component-analysis-essentials/#biplot

• Kassambara, A. (2017, September 25). HCPC - Hierarchical Clustering onPrincipal Components: Essentials. Retrieved fromhttp://www.sthda.com/english/articles/31-principal-component-methods-in-r-practical-guide/117-hcpc-hierarchical-clustering-on-principal-components-essentials/

• KateřinaKujanová,MiladaMatoušková,ZdeněkHošek,Therelationshipbetweenriver types and land cover in riparian zones, Limnologica, Volume 71, 2018,Pages29-43,ISSN0075-9511,https://doi.org/10.1016/j.limno.2018.05.002.

• Puggini, L., & McLoone, S. (2017). Forward Selection Component Analysis:AlgorithmsandApplications.IEEETransactionsonPatternAnalysisandMachineIntelligence,39(12),2395-2408.https://doi.org/10.1109/TPAMI.2017.2648792

• R-project. (n.d.). Plotting PCA (Principal Component Analysis). Retrieved fromhttps://cran.r-project.org/web/packages/ggfortify/vignettes/plot_pca.html

• Richardson,David&Holmes,Patricia&Esler,Karen&M.Galatowitsch,Susan&Stromberg,Juliet&Kirkman,Stephen&Pyšek,Petr&J.Hobbs,Richard.(2007).Riparian vegetation: Degradation, alien plant invasions, and restoration

Page 21: Assessment of the riparian condition of two tributary ...

prospects. Diversity and Distributions. 13. 126 - 139. 10.1111/j.1366-9516.2006.00314.x.

• Rincón Carrera, E. (2009). SIATAC :. Sistema de Informacion AmbientalTerritorial de la Amazonia Colombiana. Retrieved fromhttp://siatac.co/web/guest/productos/coberturasdelatierra/fichasdepatrones?p_p_id=54_INSTANCE_K1kl&p_p_lifecycle=0&p_p_state=normal&p_p_mode=view&p_p_col_id=column-2&p_p_col_count=1&_54_INSTANCE_K1kl_struts_action=/wiki_display/view&_54_INSTANCE_K1kl_nodeName=Fichas dePatrones&_54_INSTANCE_K1kl_title=Bosquedegaleríayripario

• Steiger,J.H.(2015,February16).PrincipalComponentsAnalysis.Retrievedfromhttp://www.statpower.net/Content/312/RStuff/PCA.html

• Twidale, C. (2004). River patterns and theirmeaning. Earth-science Reviews -EARTH-SCIREV.67.10.1016/j.earscirev.2004.03.001.