You are on page 1of 18

Multi-Site Sensitivity Analysis and Calibration of

a SWAT Model on a Selected Urban Watershed in


Metropolitan Atlanta, GA

Nahal Hoghooghi and Dr. David Radcliffe

Crop & Soil Sciences Department


 Outline
I. Introduction
II. Material and methods
 Study area
 SWAT model inputs

III. Preliminary results


 Model calibration
 Sensitivity analysis

IV. Summary
I. Introduction

 Usually in hydrologic models, calibration and validation are


conducted at the outlet of a watershed.

 It doesn't represent all hydrological patterns within the watershed.

 Main objective is to conduct the calibration of stream flow at the up


stream sub-basin as well as at the watershed outlet to improve flow
pattern predication.
 Study area
 Study area

• Big Haynes Creek watershed


o 45 km2
o 29 sub-basins
o USGS gaged station

• Sub-basin 7
o 2 km2
o UGA automated gage recorder
 Study area
Big Haynes
Creek watershed Sub-basin 7

Land use (%)

Urban 58 67

Forest 28 24

Range & hay 11 7

Water 1 0.4

Wetlands 1 0.3
Industrial 1 0
 SWAT inputs
Data type Source Description

Soil survey spatial and


Soil SSURGO 2.2
tabular data
Land use land cover
Land use NLCD 2001
classification
National elevation dataset Digital Elevation Model
Elevation
10 meter (DEM)

PRISM (Parameter-elevation
Daily precipitation and
Weather regression on independent slopes
temperature
model)
 SWAT inputs
• Reservoirs
• No point source

 SWAT run
• 4 years warm-up
• Simulation: 2012-2014
• Daily time step
 SWAT-CUP inputs
• Big Haynes watershed (SUFI2)
o 3 iterations, each 1000 runs
o Start with 22 parameters
o Observed stream flow (cm s-1) from USGS, 2012-2014
o Daily time step

• Sub-basin 7 (SUFI2 & manual calibration)


o First iteration start with 22 parameters
o Observed stream flow (cm s-1) from UGA automated gage recorder,
from July 2012 to May 2014
o Daily time step
III. Results
 Sensitivity analysis - Big Haynes watershed
Fitted
Parameter Method Min Max
value
1 TRNSRCH .bsn Replace 0 1 0.02
2 SOL_K .sol Relative -0.8 0.8 0.38
3 CN2 .mgt Relative -0.2 0.2 -0.08
4 CH_N2 .rte Replace 0.025 0.15 0.05
5 RES_EVOL .res Relative -0.2 1 0.49
6 GW_REVAP .gw Replace 0.02 0.2 0.07
7 GWQMN .gw Replace 0 1000 407.5
8 CH-K2 .rte Replace 0 30 18.6
9 REVAPMN .gw Replace 0 500 389.5
10 CH_K1 .sub Replace 0 300 285.1
 Calibration - Big Haynes watershed

NS KGE R2 p-factor StdDev-sim(StdDev-obs) Mean-sim(Mean-obs)


0.74 0.85 0.74 0.66 0.71 (0.83) 0.55 (0.56)
 Calibration- Sub-basin 7

NS KGE R2 p-factor StdDev-sim(StdDev-obs) Mean-sim(Mean-obs)


-0.84 -0.24 0.00 0.37 0.03 (0.04) 0.02 (0.04)
 Calibration- Sub-basin 7
• Manual calibration
oSurface runoff
o CN2
o SOL_AWC

o Base flow
o CH_K2
o CH_K1
o CH_N2
 Calibration- Sub-basin 7
• Manual calibration

Default value in
Parameter Fitted value
SWAT
1 CN2 .mgt 55 44
2 SOL_AWC.sol 0.13 0.26
3 CH_K2 .rte 0 50
4 CH_K1 .sub 0 100
5 CH_N2 .rte 0 0.03

NS MNS KGE R2 StdDev-sim(StdDev-obs) Mean-sim(Mean-obs)


0.23 0.32 0.29 0.18 0.03 (0.04) 0.02 (0.04)
 Calibration- Sub-basin 7
• Manual calibration
IV. Summary
 Calibrated daily stream flow showed good Nash-Sutcliffe coefficient (0.74)
at the outlet of Big Haynes watershed.
 For the sub-basin 7 MNS coefficient was 0.32 for daily time step.
 The model underestimated base-flow for the sub-basin 7 in winters, may
due to the uncertainty in precipitation data.
 In sub-basin 7, unknown point sources may cause model underestimation
during low flow condition.
In progress works
 Using calibrated flow parameters of sub-basin 7 to simulate flow at the
outlet of the watershed for both scenarios (w/o OWTS).

 Including other sources of nitrogen in the mentioned watershed (lawn


fertilizer, and manure application in pasture and hay).
 Acknowledgment

This research was funded by a grant from the


USDA-NIFA National Integrated Water Quality Program

o Dr. David Radcliffe

o Dr. Rajith Mukundan

o Anthony Gotvald

You might also like