Modeling the Ungauged Basin Using TOPMODEL in...
Transcript of Modeling the Ungauged Basin Using TOPMODEL in...
Modeling the Ungauged BasinUsing TOPMODEL in GRASS GISBrent Fogleman 30 November, 2009
Purpose
Develop a predictive hydrological model to be
used by land managers at Fort Bragg in order for
them to better understand how the hydrological
characteristics of the watershed are affecting soil
erosion.
Area of InterestRainfall: the 30‐year average (1971‐2000) annual rainfall at Fort Bragg is
47.34 inches
Cumulative rainfall for study period (01Jan08‐31Dec08) is
51.81 inches
Topography: flat to rolling
Cover: intermittent grasses, small shrubs and pine
Soil: 72.4 percent sand, 15.3 percent silt, and 12.3 percent clay
Use: heavy use by helicopters and tactical vehicles has caused depletion
of cover and has disturbed the soil
Falcon LZ
Water out
Water in
Falcon LZ
Wetland
6’3”
Water out
TOPMODELWhat is it?
• A semi‐distributed rainfall runoff model
• Basic assumption: all points in a catchment with the same topographic index
value will respond in a hydrologically similar way
• Theory assumptions:
1. saturated zone can be approximated by successive steady state
2. hydraulic gradient of the saturated zone can be approximated by local
surface topographic slope
3. distribution of downslope transmissivity with depth is an exponential
function of storage deficit or depth to the water table
Upslope contributing area
qi = Totanβ exp(-Di/m)
Where:To is transmissivityDi is local storage deficitm is a model parameter controlling the rate of decline of transmissivity
ai = upslope contributing area
r is recharge rate
ai
qi = airtanβ
r
Topographic index = a / tanβ
TOPMODELWhy?
– Renders good estimation of run‐off
– Suited for small catchments
– Best suited to catchments with shallow soils and moderate
topography which do not suffer from excessively long dry
periods
Software: GRASS(Geographical Resources Analysis Support System)
Why?
– Employs TOPMODEL function
– Developed by the U.S. Army Construction Engineering Research
Laboratory (CERL)
– Taught at NCSU
– Open source
Additional Software
Matlab
– Script for calculating potential evapotranspiration (PET)
Python
– Scripts for formatting rainfall data
MS Excel
– Calculations and hydrographs
Data• USGS Gauge 02102908 Flat Creek Near Inverness, NC
– One year of daily precipitation and discharge data (01 Jan 2008 to 31 Dec 2008)
• NC State Climate Office
– One year of mean daily temp used in the calculation of PET
• DEM USGS Seamless Server
– 1/3 arc second (10m) resolution
– Datum:NAD83
– Projection: Geographic Coordinate System (lat/lon)
Data Pre-Processing• DEM
– download from USGS Seamless Server
– import into GRASS
– reproject to WGS84, UTM Zone 17N
– create drainage direction map using command r.watershed with threshold=50000
– convert basin to vector format
– extract more detailed streams from flow accumulation
– thin streams and convert to vector format
– remove small spurs or "dangling lines" from the thinning process
– create a text file with the coordinates to the gauge/pourpoint
Data Pre-ProcessingIdentify the PourPoint
668195.00E 3893180.00N
Data Pre-Processing• DEM
– delineate the watershed and convert it to vector format using command r.water.outlet
– create a text file to reclassify the raster to indicate that values (1) are retained and values (0) are set to NULL
– convert reclassified basin raster to vector
– compute the watershed area
– create a MASK of the watershed using the reclassified basin as the MASK
– create a basin elevation map with MASK on
– create topographic wetness index with watershed MASK on
Flat Creek (gauged)• 19.6 sq km (4844.6 acres)
Falcon LZ (ungauged)• 13 sq km (3213.6 acres)
Elevation
Topographic Wetness IndexFlat Creek Falcon LZ
Perceptual Model• During heavy rain events, water quickly finds its way into small
channels of near‐impermeable red clay
• During storm events, water builds volume and speed in a short time
Model Parameters• GRASS version allows for 15 model parameters…….TOO MANY!
• I used 8 model parametersA Total catchment area [m^2]qs0 Initial subsurface flow per unit area [m/h]
(close to baseflow)lnTe Areal average of ln(T0) = ln(Te) [ln(m^2/h)]
(transmissivity)m Model parameter [m]Sr0 Initial root zone storage deficit [m]Srmax Maximum root zone storage deficit [m]td Unsaturated zone time delay per unit storage deficit [h]
( > 0.0 )vch Main channel routing velocity [m/h]vr Internal subcatchment routing velocity [m/h]
Model Input Preparationparameters.txt
Flat Creek 10# A# [m^2]1.961E+007
# qs0 lnTe m Sr0 Srmax td/alpha vch vr0.000075 8 0.025 0.0 0.1 80.0 3000 3000
# infex K psi dtheta0 0 0.0 0.0
# nch0
# d Ad_r0 0
Model Input Preparationinput.txt
# ntimesteps: Number of timesteps# dt: Time increment (timestep) [h] (in this case, day)# R: Rainfall [m/dt]# Ep: Potential evapotranspiration [m/dt]##
# ntimesteps dt366 1.0
# R Ep0.0000 0.001349700.0000 0.000805380.0000 0.000707110.0000 0.000819400.0000 0.00113164
GRASS GUI
Model Output# r.topmodel output file for "Flat Creek 10"Qt_peak: 3.64E+05 tt_peak: 254 Qt_mean: 1.82E+04 ncell: 196054 nidxclass: 27 ndelay: 0 nreach: 1 lnTe: 8.00E+00 vch: 3.00E+03 vr: 3.00E+03 lambda: 7.95E+00 qss: 1.05E+00 qs0: 7.50E‐05 tchAd 1.96E+07 Total flow Total flow/unit area Overland Flow/unit area Subsurface flow/unit area Vertical flux Mean saturation deficit timestep Qt qt qo qs qv S_mean
1 1.47E+03 7.50E‐05 0.00E+00 7.50E‐05 0.00E+00 2.39E‐01 2 2 1.47E+03 7.48E‐05 0.00E+00 7.48E‐05 0.00E+00 2.39E‐01 3 3 1.46E+03 7.46E‐05 0.00E+00 7.46E‐05 0.00E+00 2.39E‐01 ……..
Observed Max and Min Discharge
Nash-Sutcliffe 0.09
TM_FC10 0.090241.96E+07
# qs0 lnTe m Sr0 Srmax td/alpha vch vr0.000075 8 0.025 0 0.1 80 3000 3000
# infex K psi dtheta0 0 0 0
# nch0
# d Ad_r0 0
Subsurface flow is 90.8% of total flowOverland flow is 9.8% of total flow
Nash-Sutcliffe 0.09
TM_FC10 0.090241.96E+07
# qs0 lnTe m Sr0 Srmax td/alpha vch vr0.000075 8 0.025 0 0.1 80 3000 3000
# infex K psi dtheta0 0 0 0
# nch0
# d Ad_r0 0
Subsurface flow is 90.8% of total flowOverland flow is 9.8% of total flow
Nash-Sutcliffe -4.92
TM_FC11 ‐4.91818961.96E+07
# qs0 lnTe m Sr0 Srmax td/alpha vch vr0.000075 8 0.025 0.04 0.001 80 3000 3000
# infex K psi dtheta0 0 0 0
# nch0
# d Ad_r0 0
Subsurface flow is 87.73% of total flowOverland flow is 12.27% of total flow
Nash-Sutcliffe -4.92
TM_FC11 ‐4.91818961.96E+07
# qs0 lnTe m Sr0 Srmax td/alpha vch vr0.000075 8 0.025 0.04 0.001 80 3000 3000
# infex K psi dtheta0 0 0 0
# nch0
# d Ad_r0 0
Subsurface flow is 87.73% of total flowOverland flow is 12.27% of total flow
Nash-Sutcliffe -1.87
TM_FC09 ‐1.868209081.96E+07
# qs0 lnTe m Sr0 Srmax td/alpha vch vr0.000075 4 0.025 0 0.1 80 3000 3000
# infex K psi dtheta0 0 0 0
# nch0
# d Ad_r0 0
Subsurface flow is 64.60% of total flowOverland flow is 35.40% of total flow
Nash-Sutcliffe -1.87
TM_FC09 ‐1.868209081.96E+07
# qs0 lnTe m Sr0 Srmax td/alpha vch vr0.000075 4 0.025 0 0.1 80 3000 3000
# infex K psi dtheta0 0 0 0
# nch0
# d Ad_r0 0
Subsurface flow is 64.60% of total flowOverland flow is 35.40% of total flow
Nash-Sutcliffe -5.93
TM_FC02 ‐5.93241671.96E+07
# qs0 lnTe m Sr0 Srmax td/alpha vch vr0.000075 4 0.0125 0.0025 0.041 60 20000 10000
# infex K psi dtheta0 2 0.1 0.1
# nch1
# d Ad_r8000 1
Subsurface flow is 61.24% of total flowOverland flow is 38.77% of total flow
Nash-Sutcliffe -5.93
TM_FC02 ‐5.93241671.96E+07
# qs0 lnTe m Sr0 Srmax td/alpha vch vr0.000075 4 0.0125 0.0025 0.041 60 20000 10000
# infex K psi dtheta0 2 0.1 0.1
# nch1
# d Ad_r8000 1
Subsurface flow is 61.24% of total flowOverland flow is 38.77% of total flow
Nash-Sutcliffe -17.15
TM_FC04 ‐17.15271.96E+07
# qs0 lnTe m Sr0 Srmax td/alpha vch vr0.000075 1.76 0.02 0.0425 0.0016 41.37 4477 4477
# infex K psi dtheta0 2 0.1 0.1
# nch0
# d Ad_r0 0
Subsurface flow is 44.47% of total flowOverland flow is 55.53% of total flow
Nash-Sutcliffe -17.15
TM_FC04 ‐17.15271.96E+07
# qs0 lnTe m Sr0 Srmax td/alpha vch vr0.000075 1.76 0.02 0.0425 0.0016 41.37 4477 4477
# infex K psi dtheta0 2 0.1 0.1
# nch0
# d Ad_r0 0
Subsurface flow is 44.47% of total flowOverland flow is 55.53% of total flow
Nash-Sutcliffe -7.98
TM_FC08 ‐7.97703911.96E+07
# qs0 lnTe m Sr0 Srmax td/alpha vch vr0.000075 0.01 0.025 0 0.1 80 3000 3000
# infex K psi dtheta0 0 0 0
# nch0
# d Ad_r0 0
Subsurface flow is 29.06% of total flowOverland flow is 70.29% of total flow
Nash-Sutcliffe -7.98
TM_FC08 ‐7.97703911.96E+07
# qs0 lnTe m Sr0 Srmax td/alpha vch vr0.000075 0.01 0.025 0 0.1 80 3000 3000
# infex K psi dtheta0 0 0 0
# nch0
# d Ad_r0 0
Subsurface flow is 29.06% of total flowOverland flow is 70.29% of total flow
Nash-Sutcliffe -13.67
TM_FC07 ‐13.6721.96E+07
# qs0 lnTe m Sr0 Srmax td/alpha vch vr0.000075 0.01 0.025 0 0.025 80 3000 3000
# infex K psi dtheta0 4 0 0
# nch0
# d Ad_r0 0
Subsurface flow is 26.66% of total flowOverland flow is 73.33% of total flow
Nash-Sutcliffe -13.67
TM_FC07 ‐13.6721.96E+07
# qs0 lnTe m Sr0 Srmax td/alpha vch vr0.000075 0.01 0.025 0 0.025 80 3000 3000
# infex K psi dtheta0 4 0 0
# nch0
# d Ad_r0 0
Subsurface flow is 26.66% of total flowOverland flow is 73.33% of total flow
Challenges• Very little documentation on running TOPMODEL in GRASS
• Parameter Estimation and Predictive Uncertainty• GRASS apparently does not include analysis tools, making it not suitable for numerous runs (time consuming and high potential for human error)
What’s next?• Determine parameters that more accurately model the hydrological response of the gauged watershed
• Apply model parameters to the ungaugedwatershed to model hydrological response
• Explore the RRMT (Rainfall Runoff Modeling Toolbox) in Matlab