2015 International SWAT Conference Pula, Sardinia, Italy Danielle A. Bressiani 1, *, R. Srinivasan 2, E. M. Mendiondo 1,3 & K. C. Abbaspour 4 1 Engineering School of São Carlos, University of São Paulo 2 Spatial Science Laboratory, Texas A&M University 3 Brazilian Center of Monitoring and Early Warning of Natural Disasters 4 Eawag, Swiss Federal Institute of Aquatic Science and Technology
Background Application of hydrological models to watersheds at a sub-daily time step is very important for better understanding of flow and water quality dynamics; The urgency of such applications and developments has increased in recent years due to increased urbanization, and higher frequency of extreme hydro-meteorological events; Climate change impact and accelerated landuse change due to population growth can exasperate the situation in the coming years; A new component to allow hourly calibration has been incorporated in the SWAT-CUP software package; This study presents the application results of an hourly SWAT model using SWAT-CUP (SUFI-2 algorithm) for calibration.
Study Area Piracicaba Watershed Area = 12,600 km 2 Mean annual rainfall = 1,405 mm Mean natural flow = 143.9 m 3 s 1 Population = 3.5 million people Pop density = 272 hab/km 2 The region is prone to severe flooding and droughts.
Model Set Up The Piracicaba Watershed ArcSWAT project was built using freely available data on the web, or provided by Government agencies and Research institutions. ASTER 30 meter resolution Landsat 5 TM, supervised classification (Molin, 2015) Rotations: Local Farmers São Paulo State Map and Legends (OLIVEIRA, 1999) and pedo-transfer functions Precip: Interpolated from 189 stations for daily (ANA, DAEE) 9 Weather Stations (INMET & USP) 16 flow gauges (ANA, DAEE) 523 sub-basins were delimited in SWAT, with an average area of 20Km 2, the modeled watershed area is of 10,454 Km 2
Calibration and Validation Sensitivity Analysis (Global and One at a Time) Manual and automated calibration using SUFI-2 (SWAT - CUP) and validation were performed: Using Daily Climate Datasets: Yearly Average Values Using Hourly Rainfall Data: Monthly Time Step Hourly Time Step Daily Time Step
Hourly Calibration in SWAT-CUP A new component to allow hourly calibration has been incorporated in the SWAT-CUP software package, using SUFI-2. The hourly results for determined sub-basin is extracted in SWAT, using the fig.fig file;
Most Sensible in the Global Sensitive Analysis Monthly: GW_Delay ; CN2 Daily: CN2, CNCOEF, ESCO, Surlag Hourly: CN2, CNCOEF, LAT_TIME, ALPHA_BF, ESCO, CH_N2, SLSUBBSN, Surlag.
One at a Time Sensitivity Analysis A One at a Time, sensitivity Analysis was conducted for the hourly time step, with 4 simulations for each parameter. With several parameters: r CN2.mgt -0.40 0.40 v CNCOEF.bsn 0.5 1.0 a GW_DELAY.gw -90.0 20.0 v LAT_TTIME.hru 0 25 v ESCO.hru 0.6 0.95 r CH_N2.rte -0.20 0.20 r CH_S1.sub -0.20 0.20 r OV_N.hru -0.2 0.2 v SURLAG.bsn 0.05 24.0 r SLSUBBSN.hru -0.15 0.15 v RCHRG_DP.gw 0.01 0.095 v CANMX.hru FRSE 0 25 r CH_L1.sub -0.2 0.2 r CH_N1.sub -0.20 0.20 r ALPHA_BF.gw -0.20 0.20 r CH_k1.sub -0.30 0.30 a GWQMN.gw -1000 130 r HRU_SLP.hru -0.2 0.2 a REVAPMN.gw -500 500 v GW_REVAP.gw 0.02 0.09
Flow ( m3/s) Flow ( m3/s) 1 13 25 37 49 61 77 89 102 114 126 140 152 164 176 188 200 212 224 237 249 261 273 285 297 309 321 333 345 357 369 Monthly Results 200 160 120 80 40 0 NSE=0.84; br2=0.79; PBias=-4.38; R2=0.88; NMSE=0.07; RSR=0.16 Simulated Observed NSE=0.80; br2=0.74; PBias=-11.91; R2=0.87; NMSE=0.09; RSR=0.20 Validation (1980-1999) Time (months) Calibration (2000-2011) 120 80 NSE=0.79; br2=0.80; PBias=-16.44; R2=0.89; NMSE=0.09; RSR=0.21 Simulated Observed NSE=0.88;bR2=0.87;PBias=-3.39; R2=0.88;NMSE=0.07;RSR=0.12 40 0 1 13 25 43Validation 70 82 94(1980-106 118 1999) 130 142 154 166 Time 178 (months) 193 205 274 286Calibration 298 310 322(2000-334 3462011) 358 370 382 800 600 400 200 NSE=0.63;bR2=0.70;PBias=-18.04;R2=0.92; NMSE=0.13;RSR=0.37 Simulated Observed NSE=0.91;bR2=0.90;PBias=-1.99; R2=0.92;NMSE=0.04;RSR=0.09 7(c) 0 Validation (1980-1999) Time (months) Calibration (2000-2011)
Flow (m3/s) Flow (m 3 /s) Flow (m 3 /s) Daily Calibrated Flow Results 400 300 NSE=0.66; br2=0.79; PBias=6.94; R2=0.84; NMSE=0.10; RSR=0.34 200 100 0 1/1/2000 1/1/2002 1/1/2004 1/1/2006 Time (days) 1/1/2008 1/1/2010 400 300 200 100 NSE=0.76; br2=0.86; PBias=3.95; R2=0.87; NMSE=0.13; RSR=0.24 95PPU Calibrated Flow Observed Flow 0 3/5/2002 3/5/2004 3/5/2006 Time (days) 3/5/2008 3/5/2010 1000 800 NSE=0.83; br2=0.89; PBias=8.29; R2=0.92; NMSE=0.09; RSR=0.17 600 400 200 0 1/1/2000 1/1/2002 1/1/2004 1/1/2006 Time (days) 1/1/2008 1/1/2010
Daily Validation and Cross Validation Results Gauge Drainage NSE br2 PBias station area (km2) 4 2,308 0.80 0.82-7.57 12 1,581 0.70 0.81 5.75 16 11,040 0.67 0.73-12.80 Validation for the Red Gauge Stations Gauge Drainage NSE br2 PBias station area (km 2 ) 1 1143 0.65 0.79 15.81 2 1920 0.80 0.88 5.49 3 2152 0.75 0.86 9.58 5 1950 0.79 0.85 1.94 6 1140 0.75 0.82-1.65 7 1950 0.78 0.86 4.82 8 2180 0.69 0.83-3.52 9 2187 0.86 0.69-2.34 10 387 0.66 0.76-0.65 11 928 0.72 0.78-1.88 13 4045 0.79 0.83 1.36 14 7327 0.83 0.88 8.61 15 8500 0.83 0.86 14.61 Cross-Validation for the Blue Gauge Stations
Piracicaba Hourly Calibration Precipitation Data Limitation: Daily: ~200 gauge stations, interpolated for each sub-basin. Sufficient data for 30 years Hourly: 16 gauge stations, not with the best spatial distribution. Reasonable Data for 1 to 7 years depending on the station Also a lot of data problems were found on the hourly precipitation data.
9/1/2011 10/1/2011 11/1/2011 12/1/2011 1/1/2012 2/1/2012 3/1/2012 4/1/2012 5/1/2012 6/1/2012 7/1/2012 8/1/2012 9/1/2012 10/1/2012 11/1/2012 12/1/2012 1/1/2013 2/1/2013 3/1/2013 4/1/2013 5/1/2013 6/1/2013 7/1/2013 8/1/2013 9/1/2013 10/1/2013 11/1/2013 12/1/2013 1/1/2014 2/1/2014 3/1/2014 11/1/2010 12/1/2010 1/1/2011 2/1/2011 3/1/2011 4/1/2011 5/1/2011 6/1/2011 7/1/2011 8/1/2011 9/1/2011 10/1/2011 11/1/2011 12/1/2011 1/1/2012 2/1/2012 3/1/2012 4/1/2012 5/1/2012 6/1/2012 7/1/2012 8/1/2012 9/1/2012 10/1/2012 11/1/2012 12/1/2012 1/1/2013 2/1/2013 3/1/2013 4/1/2013 5/1/2013 6/1/2013 7/1/2013 8/1/2013 9/1/2013 10/1/2013 11/1/2013 12/1/2013 Data Issues 160 140 120 100 80 60 40 20 0 Example of event not captured by gauge network Observed Hourly Daily First Simulation A careful analysis of the precipitation and flow data had to be conducted using Matlab comparing hourly data with maximum reasonable values and daily averages, and to the closest station. 500.0 450.0 400.0 NSE=0.45, R2=0.75, PBIAS=32, RSR=0.55 Before calibration and treatment 350.0 300.0 250.0 200.0 150.0 Observed Hourly Daily First Simulation (after daily Calibration) 100.0 50.0 0.0 Example of Precipitation data issues Example of Flow measurement issue
Hourly Calibration R2= 0.61, NSE= 0.57, PBIAS=19.3 The other calibrated location, and cross-validated ones, have similar results, with NSE>0.5
Flood Modeling for Forecasting We are using Eta Model (a state-of-the-art atmospheric model used for research and operational purposes) The model is simulated by CPTEC-INPE (Center for Weather Forecasting and Climate Research/ Brazilian National Institute of Space Research) There is a ensemble of five members with initial time for runs at 00h/12h and forecast range of 72h. Also another Eta model run for all considering ocean ridge; Providing 6 possible future conditions.
Final Remarks SWAT can now be automatically calibrated using SWAT-CUP for hourly time-step; The Piracicaba SWAT Model was calibrated for hourly time steps (with limited data); Calibration of this model resulted in satisfactory and reasonable agreements between observed and modeled; The aim is to have an application of the SWAT hourly Piracicaba model for flood forecasting; Still road ahead: Improvements in the calibration and on using the ensembles is necessary; This developed method foresees future applications which can help the real time operational decision making for disaster risk reduction of hydrological extremes at strategic river basins
ACKNOWLEGMENTS
Thank you very much! For further information: danielle.bressiani@usp.br (Danielle Bressiani) daniebressiani@gmail.com