SWAT-CUP. New Features

Transcription

1 SWAT-CUP. New Features 2011 International SWAT Conference, Toledo Karim C. Abbaspour Eawag: Swiss Federal Institute of Aquatic Science and Technology

2 SWAT-CUP SWAT Calibration and Uncertainty Analysis Procedures - Calibration, - Validation, - Sensitivity analysis (One-at-a-time and global), - Uncertainty analysis, - Visualize the watershed and outlet locations, - Parallel processing

3 TxtInOut SWAT-CUP Parameters New SWAT Inputs SWAT Outputs SWAT_Edit.exe SWAT2009.exe SWAT_Extract.exe PSO SUFI-2 MCMC ParaSol GLUE Output

4 Issues w.r.t Calibration of Distributed Watershed Models - Parameterization (most important, difficult, and neglected aspect of calibration) - Positioning observed outlets in SWAT (causes the biggest headache in calibration) - Time constraint (makes some projects impossible to build and run) - Objective function definition (most surprising aspect of calibration) - Non-uniqueness (Uncertainty) (most difficult part to quantify and communicate) - Parameter conditionality (most disappointing aspect of calibration)

5 ntain R?tibach Sulzbach Parameterization Basin Map am s ershed Boundary linlands chl3 linlands chl4 her_m olas se2 her_m olas se3 braunerde ht_m oraen_g1 ht_m oraen_g2 ht_m oraen_g3 ht_m oraen_g4 n end braunerde lung M ola_m orah 1 M ola_m orah 2 M ola_m orah 3 Niederwisenbach Gr?tbach Thur Ellikerbach Binnenkanal Thur T?gelbach Seebach Dorfbach M?lit?beli L?tzelmurg Seebach Binnenkanal M?libach Dorfbach Dorfbach Tuenbach Aubach Beerenbach Lauche Murg Aum?hlebach T?felstobel Aspibach Maltbach Kaabach Chrebsbach M?hlr?tibach Alpbach H?rachbach Chemibach Gillgraben Feldbach Giessen Hartenauer Bach Chrebsbach Dietfurterbach Krinauerbach Rickenbach Furtbach Br?bach Uze Rindalbach Ibach Lederbach Ulisbach Stelzbach Furtibach Steintaler Bach R?tibach Jomerbach Ruerbach Josenbach Gerenbach Nestenbergbach Aachbach Niderwilbach Wiss Thur M?ttlibach Widenbach Dorfbach Goldbach Necker L?tisbach Ijentaler Bach Sorenbach Wissbach Schwindelbach Luteren Glatt T?fenbach Murbach R?telbach R?telbach Sangenbach Diezisbach Leistbach Lauftenbach Andwiler Bach Chellenbach Urn?sch N?rigbach Chrinnbach Giessbach Fitzibach Tosbach T?fentobelbach Wisenbach Sonderbach W issbach Gr?ppelensee Wattbach Sitter Schwarz Wilenbach Kaubach L?chlibach Fallbach Wildhauser Thur Goldibach S?gibach Zungbach Schwendibach Wissbach Rotbach Ibach Br?elbach N M ola_m orah 4

6 Parameterization x <parname>.<ext> <hydrogrp> <soltext> <landuse> <subbsn> Slope Where x= v means the existing parameter value is replaced by a given value, a means the given value is added to the existing parameter value, and r means the existing parameter value is multiplied by (1+the given value);

7 Parameterization Soil parameters Parameter identifiers r SOL_K().sol r SOL_K(1).sol Description K of all soil layers in all HRUs K of Layer 1 of all HRUs r SOL_K(1).sol FSL r SOL_K(1).sol FSL PAST r SOL_K(1).sol PAST 1-15 K of layer 1 in HRUs with soil texture FSL K of layer 1 of HRUs with soil texture FSL and landuse PAST K of layer 1 of subbasin 1 to 15 in HRUs with landuse PAST v SOL_K(1).sol 0-10 K of layer 1 for HRUs with slope 0-10

8 Parameterization Management parameters Parameter identifiers v HEAT_UNITS{rotation no,operation no} v CNOP{[],1}.mgt v CNOP{2,1,plant_id=33}.mgt v CNOP{[],1,plant_id=33}.mgt v CNOP{[],1,plant_id=33} mgt Description Management parameters that are subject to operation/rotation must have both specified change an operation's parameters in all rotations [] means every rotation that has the specified operation. Operation filters could be applied as shown The changes are applied to crop 33 only To modify just one file, it could be given as shown

9 Parameterization Rainfall parameters Parameter identifiers v precipitation(1){ }.pcp1.pcp v precipitation(1-3){ }.pcp1.pcp v precipitation( ){ , }.pcp v precipitation( ){ , , }.pcp Description (1) means column number 1 in the pcp file { } specifies year and day (1-3) means column 1, 2, and3 { } specifies year and day ( ) means all columns (all stations) { , } means 1977 days 300 and 301 ( ) means all columns from day 1 to day 361 of 1977, and from day 1 to day 365 of 1978, and day 3 of 1979

10 Parameterization par_inf.txt r CN2.mgt v ALPHA_BF.gw v GW_DELAY.gw v GWQMN.gw v GW_REVAP.gw v ESCO.hru v OV_N.hru v SLSUBBSN.hru v HRU_SLP.hru v CH_N2.rte v CH_K2.rte v ALPHA_BNK.rte r SOL_AWC(1).sol r SOL_K(1).sol Global parameterization

11 Parameterization par_inf.txt r CN2.mgt AGRR 12,13,19, r CN2.mgt FRST 12,13,19, r SOL_BD().sol 12,13,19, r SOL_AWC().sol 12,13,19, v EPCO.hru v ESCO.hru v SLSUBBSN.hru r CN2.mgt AGRR 1-11,14-18,20,22, r CN2.mgt FRST 1-11,14-18,20,22, r SOL_BD().sol 1-11,14-18,20,22, r SOL_AWC().sol 1-11,14-18,20,22,

12

13

14 Observed Outlet Positioning The biggest calibration headache!

15

16

17 Time Constraint Calibration takes a long time! Build projects with coarser resolution - Not run enough times to properly calibrate or do uncertainty - Grid computing analysis - Cloud computing - Parallel processing Rouholahnejad, Elham et al. Parallelizing SWAT Calibration in Windows using SUFI2 Program Environmental Modelling and Software, March 2011

18

19 Speedup Ideal Alberta Danube Server Number of processors 24 CPU 16 GB RAM Danube 69,875 Alberta 2,698

20 Speedup 25 1:1 line 20 SWAT project with 2698 file SWAT project with 69,875 file Number of processors

21 Future Development - Parallel Processing on GPU! - More visualization - More optimization algorithms - More features as requested by users.