modeling strategies for a groundwater dominated … · modeling strategies for a groundwater...
TRANSCRIPT
MODELING STRATEGIES FOR A GROUNDWATER DOMINATED HEADWATER SYSTEM
1Latif Kalin, 1Rasika Ramesh, 2Mohamed Hantush, 3Mehdi Rezaeianzadeh, 1Chris Anderson,
1 School of Forestry and Wildlife Sciences, Auburn University, Auburn, Alabama2 National Risk Management and Research Laboratory, US Environmental Protection Agency, Cincinnati, Ohio3 National Water Center, National Oceanic and Atmospheric Administration, Tuscaloosa, Alabama
Baldwin County, AL
© Coastal Alabama river basin management plan
• 30% population growth between 2000 and 2010• Increased concerns about regional and coastal
water quality and quantity
“Headwater streams are buried more extensively than are larger streams at all levels of urban development (low, residential, suburban, and urban)” – Elmore and Kaushal (2008)
Wetland alterations
Watershed land use is an important driver of wetland function
-70
-60
-50
-40
-30
-20
-10
0
10
2/27/2011
2/28/2011
3/1/2011
3/2/2011
3/3/2011
3/4/2011
3/5/2011
3/6/2011
3/7/2011
3/8/2011
3/9/2011
3/10/2011
Wat
er le
vel d
epth
(cm
) re
l. to
gro
und
leve
l
55% Forested
87% Urban
Headwater wetland
Baldwin County, AL
• Account for over two-thirds of the total stream length in a river
• Alterations can cause large scale impacts on ecological functions
Headwater slope wetlands
The general plan
Improved water quality prediction
SWATWatershed model
•Land use•Climate •Soils
WETQUALWetland model
•Flow•Nutrients•Climate
Flow and nutrient time series Calibrate with observed wetland data
New Foley watershed
• Head watershed draining into headwater slope wetland (yellow star in map)
• Ungauged• Stage data at discernible
inflow from Aug 2013 – May 2014
Temperature stationWatershed outlet
0
0.1
0.2
0.3
0.4
0.5
0.6
Flow
(cm
s)
Observed inflow to wetland SWAT generated flow to wetland
Observed data
On inspection…
Watershed size = 0.49 sq.kmPrecipitation = 1723 mmTotal flow = 6599 mm
Where was all this extra water coming from?
0
20
40
60
80
100
120
140
Flow
dep
th (m
m)
Precipitation Flow
The problem
• Observed flows exceeded precipitation inputs• However from repeated field visits, observed flow data is correct• The $$$ question -
• Where is all this extra baseflow coming from?• Does the watershed (0.47 km2) have a very large ground watershed
which SWAT fails to account for?
Objectives
• Can flows be modeled in SWAT relatively simply without having to resort to complex groundwater models?
• Provide useful approaches using SWAT model to predict flows in small watersheds with extensive ground watersheds
11
0
0.1
0.2
0.3
0.4
0.5
Flow
(cm
s)
Observed Stormflow SWAT simulated streamflow
0
0.05
0.1
0.15
0.2
0.25
Flow
(cm
s)
Observed baseflow SWAT simulated baseflow
Taking a closer look..
ENASH = -5.6 ENASH = 0.44
Approach 1
From the data,
SWAT streamflow + [a + (b * SWAT baseflow)] ≅ Observed streamflow
Linear regression
0
0.004
0.008
0.012
0.016
0.02
0
0.05
0.1
0.15
0.2
0.25
0.3
SWAT
sim
ulat
ed b
asef
low
(cm
s)
Obs
erve
d ba
seflo
w (c
ms)
Observed baseflow SWAT simulated baseflow
Calibrating baseflow trendImportant SWAT parameters:GWDELAY = 1 dayRCHRGE_DP = 0
Observed baseflow = 13.352 * Simulated baseflow – 0.0038(R2 = 0.75)
y = 13.352x - 0.0038R² = 0.7533
0
0.05
0.1
0.15
0.2
0.25
0.3
0 0.005 0.01 0.015 0.02
Obs
erve
d ba
seflo
w (c
ms)
Trend calibrated baseflow simulated by SWAT (cms)
Observed baseflow (cms)Linear (Observed baseflow (cms))
Calibrating stormflow
0
0.1
0.2
0.3
0.4
0.5
Flow
(cm
s)
Observed stormflow SWATCUP calibrated streamflow
(R2 = 0.71, ENASH = 0.62)
Parameter_Name Fitted_Value 1: r__CN2.mgt 0.015 2: v__ALPHA_BF.gw 0.690 3: v__GW_DELAY.gw 0.362 4: v__GWQMN.gw 21.741 5: v__GW_REVAP.gw 0.199 6: v__ESCO.hru 0.842 7: v__CH_N2.rte 0.224 8: v__CH_K2.rte 32.122 9: v__ALPHA_BNK.rte 0.884 10: r__SOL_AWC(..).sol -0.151 11: r__SOL_K(..).sol 0.014 12: r__SOL_BD(..).sol -0.809 13: v__RCHRG_DP.gw 0.006
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
8/4/2013 9/23/2013 11/12/2013 1/1/2014 2/20/2014 4/11/2014
Flow
(cm
s)
Observed streamflow (cms) SWAT simulated streamflow (cms)
Calibrated baseflow + calibrated stormflow
0
0.1
0.2
0.3
0.4
0.5
0.6
0 20 40 60 80 100
Flow
(cm
s)
Percent of time equaled or exceeded
Observed streamflow (cms)
SWAT simulated streamflow (cms)
Predicted flow = 13.352 * trend calibrated baseflow – 0.0038 + calibrated stormflow(R2 = 0.74, ENASH = 0.67)
Approach 2
Baseflow
Recharge to aquifer wrchrg
When β > 0 When β < 0
Deep aquifer recharge wdeep = βdeep.wrchrg
Total shallow aquifer recharge, wrchrge,sh
Shallow aquifer
Deep aquifer
• In SWAT, RCHRGE_DP (βdeep) ranges from 0 to 1
- only considers percolation loss to deep aquifer
Calibrate with negative RCHRGE_DP
Parameter_Name Fitted_Value Min_value Max_value 1: r__CN2.mgt 0.285 0.126 0.421 2: v__ALPHA_BF.gw 0.082 0.001 0.173 3: v__GW_DELAY.gw 0.563 0.001 3.104 4: v__GWQMN.gw 41.312 25.939 43.589 5: v__GW_REVAP.gw 0.094 0.026 0.101 6: v__ESCO.hru 0.952 0.916 0.985 7: v__CH_N2.rte 0.144 0.125 0.254 8: v__CH_K2.rte 131.781 85.332 142.185 9: v__ALPHA_BNK.rte 0.386 0.232 0.740 10: r__SOL_AWC(..).sol -0.435 -0.441 -0.068 11: r__SOL_K(..).sol 0.241 -0.098 0.442 12: r__SOL_BD(..).sol 0.029 -0.217 0.134 13: v__RCHRG_DP.gw -15.293 -19.559 -13.066
00.10.20.30.40.50.6
0 20 40 60 80 100
Flow
(cm
s)
Percent of time equaled or exceeded
Observed streamflow (cms)SWAT simulated streamflow (cms)
(P-factor = 0.84, R-factor = 1.04, R2 = 0.78, ENASH = 0.75
SWAT+ANN for improved calibration
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
8/4/2013 9/23/2013 11/12/2013 1/1/2014 2/20/2014 4/11/2014
Flow
(cm
s)
Observed streamflow SWAT-ANN simulated streamflow
00.10.20.30.40.50.6
0 50 100
Flow
(cm
s)
Percent of time equaled or exceeded
Observed streamflow (cms)SWAT simulated streamflow (cms)
SWAT calibrated flow + ET + Precipitation = Improved hydrology prediction1 hidden layer, 8 nodes, log-sigmoid transfer functionENASH = 0.89
20
Results and conclusions
• Evaluated approaches for SWAT application in groundwater dominant watersheds
• Useful SWAT parameter tweak to model high groundwater systems without having to resort to complex groundwater models
• SWAT+ANN is a very useful tool for superior hydrology calibration