1. A Soil-water Balance and Continuous Streamflow Simulation Model that Uses Spatial Data from a Geographic Information System (GIS)
Advisor: Dr. David Maidment
Research Sponsor: Hydrologic Engineering Center

2. Overview
Hydrology review: Event-based vs. continuous simulation models
Research objective
Study area
Spatial Data
radar precipitation, USDA soils
spatial analysis, parameter estimation
information transfer from GIS to an external model
Hydrologic process representation
Summary

3. Hydrologic Simulation
EVENT SIMULATION
SHOW SLIDE WITH LOSS RATES ETC.!!
HMS Basin Schematic
Well-known computer programs: HEC-1, HMS, TR-20
Sub-basin hydrograph methods:
Loss rate
Transform
Baseflow recession
Make sure that we are on the same page in terms of terminology.

4. Hydrologic Simulation(cont.)
CONTINUOUS SIMULATION
evaporation
rainfall
Well known computer
programs:
USGS PRMS/Stanford
NWS/Sacramento
HEC Continuous Simulation Model
interception and depression storage
surface
runoff
soil root zone
soil transmission zone
Neglecting snowmelt
Evaporation/Deep percolation are sinks that are difficult to quantify
subsurface
runoff
groundwater storage zone(s)
leakage

5. Event vs. Continuous Simulation
simple
infiltration losses are a sink
difficult to initialize
Continuous
more physical processes represented
antecedent storm conditions are known
complex -- many more parameters
PROS &
CONS
Neither model provides mass closure for the entire hydrologic cycle.
hydrologic/hydraulic design
flood forecasting
real time water control
water resources planning
climate change impacts on streamflow
Event
Continuous
APPLICATIONS

6. Research Objectives
Develop and test a practical model that uses data describing spatial variability of soils and rainfall
- Develop GIS/hydrology procedures applicable anywhere in the U.S.
- Does added information improve runoff estimates? -- Particularly with regard to the validation stage.
Make results reproducible by automating and working from standard databases.
What spatial scales and modeling complexity are practical and useful?
GEWEX Continental Scale International Project
SGP 97
ARM CART experiments
spatial heterogeneity
put simply -- I am interested in modeling soil moisture dynamics and their influence on storm runoff.
Current large scale models used in climate simulation use statistical distributions to describe precipitation and soil variability. This works well when it doesn’t matter exactly where within a box the net vertical (rainfall - evaporation) and horizontal fluxes (runoff) occur. But does not make sense for estimating flow at a specific point and time.

7. Study Area Little Washita River Watershed 600 km2
Climate: moist and subhumid
Why choose Little Washita?
NWS NEXRAD Stage III rainfall data
Higher resolution soils data than is generally available.
Site of numerous hydrologic and remote sensing studies -- data available for calibration and validation.

8. Problem Description climate station streamflow gage
Illustrate the problem complexity that can occur even with a simple model.
Contrast to a traditional lumped model.
climate station
streamflow gage

9. NEXRAD Precipitation Data
Stage III Product
4 km x 4 km grid
hourly estimates
Composite of information from 17 radars and 500 rain gages

10. Density of Precipitation Gages
114 Oklahoma Climate
Stations (Density ~ 1 gage/ 1600 km2)
100 NEXRAD Cells Per Gage

11. USDA STATSGO Soils Data
Mapunit OK002
Map unit: grouping of map components.
Components: typically identify soils with similar properties.
Mapunit is by no means homogeneous.
There are about 16,000 soil series in the United States
# of polygons = 1800
# of mapunits = 219
#of polygons per mapunit = 8.2
# of components per mapunit = 11.2
# of layers per component = 3.1

12. Spatial Variability in Soils
STATSGO
County Level Data 2 1
Polygon 1: Mapunit OK151
89 % Sandy loam
6% Loam
2% Silty clay loam
2% Clay
1% Loamy sand
Polygon 2: Mapunit OK103
56 % Loam
30% Silt Loam
14% Sandy loam
Polygon 2 only comprises 10% of all polygons
in mapunit OK103.

13. Main GIS Procedures Assumed inputs:
Coverage of Modeling Units (I.e. NEXRAD Cells, Thiessen polygons)
Watershed boundaries
Flowlength grid
STATSGO/SSURGO coverage w/ component and layer tables
1. Calculate soil component properties using attribute tables and lookup table
Component properties dBase file
NEXRAD cell/watershed
shape file
2. Intersect the precipitation cells with
the watershed boundary
3. Determine the component names and component percentages in each NEXRAD cell.
4. Determine the average flowlength from each NEXRAD cell the watershed outlet.

14. Texture Name to Soil Parameters
719 STATSGO texture names
12 standard
USDA classes
Soil parameters
Cobbly
bouldery
gravelly
mucky
very, extremely

15. Tabulation of Component Names and Percentages
Selected cell actually intersects two mapunits -- one of these mapunits has 14 components and another has5 components.
Why make hydrologic calculations in an external model? Because making computations within a GIS is extraordinarily inefficient.

16. External Model for Hydrologic Calculations

17. GIS as a Pre-processor for Hydrologic Models
+ Add spatial information
+ Automate/Create a Reproducible Product
- Increase computational burden
Accounting for spatial variability in a simple way
increases the computational burden in the Little Washita
by a factor of :
55 cells * 10 components/cell = 550
Good for automation
Good for adding spatial information
computational burden can increase quickly
85 is 845 km2
86 is 1556 km2
113 is 825 km2
127 is 936 km2
total = 4163 km2
LESSON: KEEP MODEL SIMPLE

19. Soil-water Balance Model
infiltration
evaporation
direct
runoff
root zone
transmission
zone
percolation
USGS, US Army Corps of Engineers, National Weather Service all have their own continuous simulation models of this type -- conceptual storage models.
* Not distinguishing between interception and depression storage; Not considering snow melt; Not considering hillslope effects (saturation excess mechanism); Only one subsurface reservoir for baseflow
Parameters are typically estimated by comparing modeled and observed hydrograph responses under different conditions I.e examine the prediction errors when the soil moisture defecit is too small ; estimate impervious cover by estimating direct runoff when antecedent conditions are extremely dry.
Parameters in a typical soil zone model are difficult to estimate based on physical data -- maximum soil storage, maximum infiltration capacity (for the watershed in the case of the Stanford model, maximum percolation (Ks?), maximum tension storage
Using simple models to represent infiltration, evaporation, and percolation for which parameters can be estimated from soils databases. What parameters am I interested in : depth, relative permeability, water retention parameters
GW Reservoir(s)
subsurface
runoff

20. Green-Ampt Infiltration Modelq r qi qo
soil depth
Actual profile q r qi Dq f qo L
soil depth
Idealized profile

21. Infiltration Rate as a Function of Initial Moisture Content
Infiltration and Precipitation Rates vs. Time
for a Loam Soil 7
Initial Eff. Saturation = 0.7
Direct Runoff = 5.1 cm
Initial Eff. Saturation = 0.2
Direct Runoff = 3.5 cm 6 5 4
Rate (cm/hour) 3 2 1 1 2 3 4 5 6 7 8 9 10 11
time (hours)

22. Percolation/Redistributionqi qi qo qo
t = 2
t = 1
t = 3
t = 1
t = 2
t = 3
Layer 1
soil depth
Layer 2
Even though the appearance is simple, the first two models become somewhat complicated when both evaporation and redistribution occur at the same time.

23. Evaporation Evaporation depends on many factors including :
energy available at the surface
water content of the soils
soil type
vegetation characteristics
atmospheric conditions
Left graph is April 30, 1996
Right graph is March 15, 1996
Si has already been reduced by absorption, scattering, and reflection in the atomosphere.

25. Evaporation E = f(q)*PE PE changes as soil dries out.
Seasonal effects?
PE changes as soil dries out.
Penman-Monteith is widely cited alternative but how do you determine surface resistance, especially when the soil begins to dry out?
moisture extraction
function, f(q) q wp c fc 1
soil moisture fraction
In reality, evaporation from a leaf is a function of air temp, humidity, solar radiation, soil moisture defecit. Transpiration is a function of density of leaves in an area.
Ideas for computing rs as a function of soil moisture. . .

26. Questions and Data Related to Evaporation
How to account for the following factors using a simple
(daily) model?
How to quantify influence of the moisture state
on the evaporation rate.
To what depth(s) does surface drying influence soil
moisture ?
Is it possible to account for seasonal effects?
EBBR : Energy Budget Bowen Ratio
SWATS : Soil Water and Temperature Systems
SMOS : Surface Meteorological Observation Stations
ARM Data Streams

27. Bowen Ratio Method Lo Rn = Si(1-a) + Li-Lo Si aSi Li
Rn + H + lE + G = 0

28. June 14, 1997

29. SWATS Data - heat dissipation sensors calibrated against matric
potential
- water retention curve used
to estimate soil water
- measurements at eight depths

30. Summary
Utilize spatial data describing soils and rainfall in a hydrology model.
GIS programs are used to automate parameter estimation.
Evaluate soil-water balance model using both observed soil moisture and runoff data.
Data availability determines model complexity.
Good news is -- utilizing spatial information
Bad news is -- the list of factors not being considered is probably longer than the list of factors that is being considered.
Comforting factors -- dominant factors are being considered, calibration
For large areas, state of the art inputs in terms of space-time resolution.
Questions on paper shared with Maidment
* Not considering the effects of agricultural practices
* Not considering the effects of detention ponds.