• PC-based GIS tool for watershed modeling

– KINEROS & SWAT (modular)

• Investigate the impacts of land cover change on runoff, erosion, water quality

• Targeted for use by research scientists, management specialists

– technology transfer

– widely applicable

Introduction

• Develop landscape assessment tool for land managers using indicators

• Advance scientific understanding of principles governing watershed response to change

land cover changes in the US and associated impacts on runoff volume, water quality

• Investigate historical changes using repeat imagery (San Pedro, Catskill/Delaware)

• Investigate spatially distributed hydrologic processes using single scenes (Las Vegas)

• Forward looks with simulations

Objectives of the Research

Introduction

• Hydrologic Modeling in GIS

– The Shape and characteristics of the earth’s surface is useful for many fields of study.

– Understanding how changes in the composition of an area will affect water flow is important!

What happens when residential development occurs?How does this affect the watershed?How can these affects be mitigated?

– Best Management Practices (BMPs)

Introduction

Introduction

Topographic Maps

A watershed boundary can be sketched by starting at the outlet point and following the height of land defining the drainage divides using the contours on a map.

Outlet Point

Traditional watershed delineation had been done manually usingContours on a topographic map.

Introduction

Introduction: Terminology

• Drainage system - The area upon which water falls and the network through which it travels to an outlet.

• Drainage Basin - an area that drains water and other substances to a common outlet as concentrated flow (watersheds, basins, catchments, contributing area).

• Subbasin - That upstream area flowing to an outlet as overland flow

• Drainage Divide - The boundary between two basins. This is an area of divergent flow.

GIS Background

GIS Background

• Raster Data Structure– Much of the data we will use in this class

will be “Raster” data.– Raster formatted data is much more

suitable for many types of landscape modeling, including hydrologic analysis.

– Inputs such as elevation can only be processed as a raster data set

– Raster is Faster, Vector is Corrector

GIS Background

Raster

Vector

Real World

GIS Background: DEMs

• Digital Elevation Models (DEM)– A DEM is a digital representation of the

elevation of a land surface.– X,Y and Z value– The USGS is the major producer of DEM’s

in the Nation

GIS Background: DEMs

•DEM’s consist of an array of data representing DEM’s consist of an array of data representing elevation sampled at regularly spaced intervalselevation sampled at regularly spaced intervals

XX

YY

ELEVATION VALUES

GIS Background: DEMs

• Two Scales of DEMs Available– 1:24,000 Scale

Level 1 - 30 meter spacing– Errors up to 15 meters inherent in data– Developed using automated methods from air photos– Systematic errors evident as banding– Not appropriate for hydrologic modeling

Level 2- 30 meter spacing– Matches map accuracy of 1:24,000 scale quads– Developed by scanning published quads– Appropriate for hydrologic modeling

– 1 Degree (~250,000) scale - 93 meter spacingAppropriate for regional analysis (not for AGWA)

GIS Background: DEMs

• Preprocessing DEMs– DEMs typically require some type of

preprocessing prior to hydrologic modeling to remove errors inherent in the data. This type of processing can greatly increase the accuracy of a DEM..

– Primary error found in DEMs are “Sinks”A sink is an erroneous depression created by

the DEM interpolation routineSinks are usually small and cause drainage

basins to be incorrectly delineated

GIS Background: DEMs

100 100 100 100

100 97 96 95100

1001009998100

100 100 100 100 101

Stream

100-meter Elevationcontour

Sink Area

100 Cell Elevation

Cells containing the contour are assigned the value of the contour,all other cells are interpolated. Sinks are always possible in areaswhere contours converge near a stream.

An example Sink

100

• Generating Surface Parameters– Flow across a surface will always be in the

steepest down-slope direction– Known as “Flow Direction” this is the basis of all

further watershed modeling processes.– Once the direction of flow is known it is possible to

determine which and how many cells flow into any given cell!

– This information is used to determine watershed boundaries and stream networks.

GIS Background: Surface Parameters

• Generating Surface Parameters - Flow Direction– In ArcView Spatial Analyst, the output of a

Flow Direction is a grid whose values can range from 1 to 255 based on the direction water would flow from a particular cell. The cells are assigned valued as shown below.

GIS Background: Flow Direction

32 64128

16 1

8 4 2Target Cell

• Generating Surface Parameters - Flow Direction– If a cell is lower than its eight neighbors, that cell is

given the value of its lowest neighbor and flow is defined towards this cell.

– If a cell has the same slope in all directions the flow direction is undefined (lakes)

– If a cell has the same slope in multiple directions and is not part of a sink the flow direction is calculated by summing the multiple directions

GIS Background: Flow Direction

GIS Background: Flow Direction

2 2 2 1 128

1 1 1 12864

3264128128128

128 64 64 32 80

Flow Direction Surface

100 100 100 100 94

100 97 96 95100

1001009998100

100 100 100 100 101

Original Surface

• Generating Surface Parameters - Flow Accumulation– If we know where the flow is going then we can

figure out what areas (cells) have more water flowing through them than others.

– By tracing backwards up the flow direction grid we can figure the number of cells flowing into all cells in a study area

– Accumulated flow is calculated as the accumulated number of all cells flowing into each downslope cell.

GIS Background: Flow Accumulation

• Generating Surface Parameters - Flow Accumulation– For an accumulation surface the value of

each cell represents the total number of cells that flow into an individual cell

– Cells that have high accumulation are areas of concentrated flow and may be used to identify stream channels.

GIS Background: Flow Accumulation

Flow Direction Surface

GIS Background: Flow Accumulation

0 0 0 0 18

0 3 8 150

00220

0 0 0 0 0

Flow Accumulation Surface

2 2 2 1 128

1 1 1 12864

3264128128128

128 64 100 32 80

GIS Background: Flow Accumulation

Flow Accumulation Surface

DEM Flow Direction Flow Accumulation

AGWA

• Integrated with US-EPA Analytical Tool Interface for Landscape Assessment (ATtILA)

• Simple, direct method for model parameterization

• Provide accurate, repeatable results

• Require basic, attainable GIS data– 30m USGS DEM (free, US coverage)– STATSGO soil data (free, US coverage)– US-EPA NALC & MRLC landscape data

(regional)

• Useful for scenario development, alternative futures simulation work.

Objectives of AGWA

Natural Condition

Decreased Vegetation

Increased VelocityIncreased Runoff

Increased Erosion

Decreased Water Quality

Land cover changeDegradation

UrbanizationWoody plant invasion

infiltration interception evapotranspiration surface roughness

soil moistureflood hazard groundwater recharge

Land Cover & Hydrologic Response

Navigating Through AGWA

Subdivide Watershed Into Model Elements

SWAT KINEROS

generate rainfall input files

Thiessen map from…Gauge locationsPre-defined continuous record

Storm Event from…Pre-defined return-period / magnitude“Create-your-own”

Intersect Soils & Land Cover

Generate Watershed Outline

Navigating Through AGWA, Cont’d…

Subwatersheds & ChannelsContinuous Rainfall Records

prepare input data

Run The Hydrologic Model & Import Results

Display Results

For subwatershed elements:•Precipitation (mm)•Evapotranspiration (mm)•Percolation (mm)•Surface Runoff (mm)•Transmission Losses (mm)•Water Yield (mm)•Sediment Yields (t/ha)

Channel & Plane ElementsEvent (Return Period) Rainfall

For Plane & Channel Elements:•Runoff (mm, m3)•Sediment Yield (kg)•Infiltration (mm)•Peak runoff (mm/hr, m3/sec)•Peak sediment discharge (ks/sec)

ArcView working directory –secondary or temporary

coverages, grids, and tables

Spatial data –primary coverages and grids

Simulation input/output –Separate directories for

each simulation

AGWA directory –primary tables, AV project

file, and model executables

Suggested File Structure for AGWA

Hydrologic Modeling & AGWA

AGWAGIS DataRainfall

RunoffErosion

Assumptions

PROCESS

runoff, sediment hydrographtime

runo

ff

STATSGONALC, MRLCUSGS 7.5' DEM

Conceptual Design of AGWA

Build Model Input Files

Derive Secondary Parameterslook-up tables

Characterize Model Elementsf (landcover, topography, soils)

Discretize Watershedf (topography)

View Model Resultslink model to GIS

Build GIS Database

PRODUCTS

ContributingSource Area

Gravelly loam Soil Ks = 9.8 mm/hr G = 127 mm Por. = 0.453

inte

nsity

time

10-year, 30-minute event

1992 NALC Hillshade DEM STATSGO

ForestOak WoodlandsMesquite WoodlandsGrasslandsDesertscrubRiparianAgricultureUrbanWaterBarren / Clouds

Land Cover

0 5 10 km

N

GIS Data Layers for AGWA Upper San Pedro Basin, SE Arizona

CSA: 2.5% (6.9 km2)44 watershed elements29 channel elements

CSA: 20% (55 km2)8 watershed elements5 channel elements

CSA: 5% (13.8 km2)23 watershed elements15 channel elements

CSA: 10% (27.5 km2)11 watershed elements7 channel elements

0 5 10 km

N

the influence of CSA on watershed complexityAutomated Watershed Characterization

Note channel initiationPoint changing with CSA

64

74

5431 41

1121

51

24

14

4434

9484

0 10 20 km

N

11

14

pseudo-channel 11

channel 14

Abstract Routing Representation

to channel 64

Watershed Configuration for SWAT channel and subwatershed hydrology

Watershed Configuration for KINEROS

71

7372

74

71

73

72

74

0 5 kmN

Abstract Routing Representation

upland, lateral and channel elements in cascade

Characterizing the Watershed

complex topography land cover soils high spatial variability complex watershed response

Characterizing the Watershed

• Homogeneous planes

• Hydrologic parameters represent intersections of topo., cover, soil

• Information loss as f (geometric complexity)

• Scaling issues

Watershed modeling relies on condensing spatialdata into appropriate units for representing processes

leaves plenty of room for error!

BEGIN PLANE ID = 71, LEN = 1303.0, AREA = 10783378.3 SL = 0.029, MAN = 0.052, X = 593519.0, Y = 3505173.5 CV = 0.92, PRINT = 1 KS = 7.94, G = 118.14, DIST = 0.3, POR = 0.459, ROCK = 0.43 FR = 0.49, 0.33, 0.17, SPLASH = 24.42, COH = 0.006, SMAX = 0.93 INTER = 2.56, CANOPY = 0.133, PAVE = 0.00END PLANE BEGIN PLANE ID = 72, LEN = 765.0, AREA = 4357163.9 SL = 0.043, MAN = 0.054, X = 591637.8, Y = 3507025.3 CV = 0.93, PRINT = 1 KS = 7.77, G = 116.95, DIST = 0.3, POR = 0.459, ROCK = 0.43 FR = 0.49, 0.33, 0.16, SPLASH = 24.61, COH = 0.006, SMAX = 0.93 INTER = 2.85, CANOPY = 0.112, PAVE = 0.00END PLANE BEGIN PLANE ID = 73, LEN = 945.0, AREA = 7405044.9 SL = 0.038, MAN = 0.052, X = 593864.3, Y = 3507560.5 CV = 0.95, PRINT = 1 KS = 8.19, G = 114.97, DIST = 0.3, POR = 0.459, ROCK = 0.43 FR = 0.5, 0.33, 0.16, SPLASH = 24.91, COH = 0.006, SMAX = 0.93 INTER = 2.6, CANOPY = 0.137, PAVE = 0.00END PLANE

71

73

72

74

KINEROS ParameterLook-Up Table

NLCD

Land cover A B C D Cover

High intensity residential (22) 81 88 91 93 15

Bare rock/sand/clay (31) 96 96 96 96 2

Forest (41) 55 75 80 50

Shrubland (51) 63 77 85 88 25

Grasslands/herbaceous (71) 80 87 93 70

Small grains (83) 65 76 84 88 80

CURVE NUMBERHydrologic Soil Group

Curve Number From MRLC

Higher numbers result in higher runoff

N

Contributing Source Area: 2000 acres - ~5% of total watershed area20 subwatershed elements19 channels

STATSGO ID: AZ061Grassland & desertscrubModerate relief

Sample Watershed Configuration - SWAT

Watershed ID: 7Area: 11.8 km2Slope: 3.7 %Cover: 12.8 %Ks: 18.1 mm/hrCN: 71.8Soil Hyd. Group: BMultiple Soil Horizons

N

Contributing Source Area: 2000 acres - ~5% of total watershed area33 planes - 7 upland elements - 25 lateral element19 channels

STATSGO ID: AZ061Grassland & desertscrubModerate relief

Sample Watershed Configuration - KINEROS

Watershed ID: 73Area: 7.45 km2Slope: 3.53 %Width: 945 mLength: 7876 mInterception: 2.60 mmCover: 13.70 %Manning's n: 0.052Pavement: 0.00 %Splash: 24.91Rock: 0.43Ks: 6.67 mm/hrSuction: 115 mmPorosity: 0.459Max saturation: 0.93Cv of Ks: 0.95Sand: 50 %Silt: 33 %Clay: 17 %Distribution: 0.30Cohesion: 0.006

0 1 2 3 5 rainfall depth (mm)

Summer convective stormAugust 11, 2000 high spatial variability high temporal variability difficult to characterize flashy runoff response short duration (45 min)

Winter frontal stormJanuary 13, 2001 low spatial variability low temporal variability lead to little or no runoff long duration (3 hours)

0 5 kmN

Rainfall Characteristics in SE Arizona

What Could Possibly Go Wrong??

SYSTEMIC ERRORS

These are “hidden” & include:

• Poor conceptual model

• Programming errors AGWA, SWAT, KINEROS

• Poor process representation

• Errors in GIS data Land cover, soils

• Assumptions in the look-up tables

PROCESSING ERRORS

These are “visible” & include:

•Errors in GIS data DEM

•Lack of input data GIS, rainfall

•AGWA fails to characterize watershed

Rainfall-Runoff Process in SE Arizona

Spatial Distribution of Rain Gauges

###

##

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

##

#

#

#

#

#

#

#

#

#

#

##

#

#

#

#

#

#

#

#

#

#

#

#

##

#

#

#

#

##

#

#

#

#

#

##

#

#

#

#

#

#

#

#

#

#

##

#

#

#

#

#

#

#

#

##

##

# #

#

##

#

#

#

####

#

#

#

#

#

#

#

#

#

#

#

#

#

Typical Distribution Upper San Pedro Basin 6100 km2

15 rain gauges

High Density Watershed Walnut Gulch Exp. WS 148 km2

89 rain gauges

after Osborne et al., 1985**

Design Storms stored in AGWA

5 yr 30 min10 yr 30 min100 yr 30 min5 yr 60 min10 yr 60 min100 yr 60 min

“create your own”

Sample Design Rainfall Events for KINEROS

Time (min)

Inte

nsity

(m

m/h

r)

100-year, 60-minute

5-year, 30-minute

10-year, 60-minute

** Data reduced “on the fly” for watershed area

0 20 40 60

0

40

80

120

160

Limitations of GIS - Model Linkage

• Model Parameters are based on look-up tables- need for local calibration for accuracy

• Subdivision of the watershed is based on topography- prefer it based on intersection of soil, lc, topography

• No sub-pixel variability in source (GIS) data- condition, temporal (seasonal, annual) variability- MRLC created over multi-year data capture

• No model element variability in model input-averaging due to upscaling

• Most useful for relative assessment unless calibrated

AGWA Gone Awry

Problems Running AGWA

- DEM pre-processing sinks indeterminate boundary converging flow lack of defined flow path (big flat area)

- User error clicking on hill slope

- Data coverage no overlap in GIS data suitability of GIS data f (scale, model)

when good software goes bad

30m DEM – Level I USGS - unfiltered, unfilled - contains many sinks - has poor drainage - watershed delineation fails

10m DEM from Air Photos - still contains some sinks - exhibits better drainage - boundary is correct - there are still internal failures

10m DEM improved - filtered using high filter - filled to remove 222 sinks - no sinks at the end - good drainage pattern - watershed succeeds; good boundary

smoothing

Watersheds Generated From Different DEMs

Same points, tolerances, settingsnote differences in results

30m DEM

10m DEM

Parallel channels affect drainage

Sinks interrupt flow

Streams Generated From Different DEMs

Example of DEM Processing To Avoid Problems

Problem: 30m DEM contains sinks, poor flow direction, and cannot create correct watershedSolution: Filter and fill the DEM before analysis

Negative:

• streams running across the watershed divide

• boundary extends beyond the correct position

smoothing can be good & bad

Positives: • better definition of channels• no sinks• hydrologic connectivity

Get the Best Available Data!

USGS level 1 USGS level 2

inaccurate flow

minimum CSA lower for Level II

DEM error… banding better boundary

bottom linejunk in = junk out

User Error

User selected a watershed outlet missed the channel and grabbed a separate basin could also fail to generate a watershed entirely

user selected here

should have selected over here

AGWA Helps With This Problem - AGWA uses a search radius to find maximum flow accumulation - can move the point downstream - use a point coverage to specify outlet

user selects hereAGWA uses here

Lack of GIS Data Coverage

Commensurate Land Cover, DEM, STATSGO - simple to determine - AGWA can handle small errors through averaging

Land cover does not extend fully

Watershed boundary

Relative vs. Absolute Change

• Availability of repeat classified imagery for change detection- NALC

• Calibration data set for absolute change analysis- USGS runoff gauging station- Internal validation preferable to calibrating solely on outlet

• Plenty of rainfall data for the time periods - NWS gauges- Potentially NEXRAD radar data

• Confounding effects of land cover change & rainfall data- Uniform vs. distributed rainfall

Extra Slides Follow

AGWA Processing Time

Discretization levelWatershed Area (km2)

Boundary Delineation

CSA 20% CSA 10% CSA 2.5%

150 * 0:03 0:22 0:25 0:37

150 0:56 0:28 0:35 0:43

750 1:18 0:48 1:13 1:30

1940 2:03 2:50 2:45 3:20

3370 3:03 5:37 5:43 6:13

7550 6:50 9:05 9:30 10:36

* Data was clipped to a small buffer around watershed

benchmarks on a PIII, 866 MHz, 256 Mb RAM

Curve Number Modeling

+

Rai

nfal

l

Determining Curve Numbers