![Page 1: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/1.jpg)
Evaluating the Impact of Different Input Datasets and Model Configuration Uncertainty
on Streamflow Simulations by Using SWAT ModelGokhan Cuceloglu, Izzet Ozturk
Istanbul Technical UniversityDepartment of Environmental Engineering
34469 Maslak Istanbul, Turkey
![Page 2: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/2.jpg)
Outline
• Introduction• Aim and Scope• Study Area• Model Setup• Results• Conclusion
![Page 3: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/3.jpg)
Introduction
• Performance of the watershed modeling studies is closely associated with the model input datasets
• Numerous databases are existing and freely available to the modelers in global or regional scale
• These datasets provide required input data to run models, their resolution and accuracies are needed to be investigated.
![Page 4: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/4.jpg)
Aim and Scope• Investigate the uncertainty caused from
• Model input data• Model configuration (in terms of subbasin number)• Model parameterization
• For this purpose we setup a model for Melen Watershed
• 2 different climate dataset• 3 different LULC dataset• 3 different subbasin number
• We consider only hydrology• Streamflow at outlet of the watershed
![Page 5: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/5.jpg)
Study Area (Melen Watershed)
Melen Watershed
Black Sea
Marmara Sea
188 km
![Page 6: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/6.jpg)
Study Area (Melen Watershed)
BLACK SEA
MELEN WATERSHED
![Page 7: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/7.jpg)
Data Sources• Digital Elevation Map SRTM 90m• Soil FAO• Land Use Data Set
• Corine Land Cover• Landsat Land Use Map• STATIP (Turkish Ministry of Agriculture)
• Climate Data Set• CFSR (1979-2014)• MGM - Local Measurement (1960-2013)
• Discharge Data• Turkish State of Hydraulics
(1 station - monthly)
Compulsory Data
Calibration
![Page 8: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/8.jpg)
Model Setup
• Simulation period (1995) 2000 - 2013• 5 years warmup• SCS-CN Method• Hargreaves Method
• SWAT-CUP using SUFI-2 algorithm• 60% of streamflow used for calibration • NSE as an objective function• Performance criteria NSE, R2, PBIAS as well
as P-factor and R-factor
![Page 9: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/9.jpg)
Landuse Data
22 Classes, 100m 7 Classes, 30m
LANDSATCORINE
STATIP
32 Classes, 10m
![Page 10: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/10.jpg)
Landuse DataLanduse SWAT CORINE LANDSAT STATIPAgricultural Land-Close-grown AGRC 0.02 - -Agricultural Land-Generic AGRL 23.25 15.37 11.47Agricultural Land-Row Crops AGRR 4.87 - -Forest-Deciduous FRSD 1.38 - 9.18Forest-Evergreen FRSE 38.25 - 1.81Forest-Mixed FRST 24.24 52.97 42.75Industrial UIDU 0.34 - 0.78Orchard ORCD 3.85 - 2.82Pasture PAST 0.32 24.85 1.64Range-Grasses RNGE 1.04 - 0.94Residential-High Density URHD - 2.12 0.79Residential-Low Density URLD 0.02 2.04 0.03Residential-Medium Density URMD 1.58 - 0.85Rice RICE 0.23 - -Southwestern US (Arid) Range SWRN 0.2 2.47 0.2Transportation UTRN 0.21 - 0.36Water WATR 0.2 0.19 0.49Wetlands-Forested WETF 0.02 - -Peanut PNUT - - 23.77Poplar POPL - - 0.05Range-Brush RNGB - - 0.22Residential-Med/Low Density URML - - 1.75Wetlands-Mixed WETL - - 0.08
32.2
2
63.8
7
2.15
1.56
0.02
0.2
15.3
7
52.9
7
4.16
27.3
2
0.19
35.2
9
53.7
4
4.56 5.
82
0.08 0.49
AGRICULTURE FOREST URBAN PASTURE WETLANDS WATER
CORINE
LANDSAT
STATIP
![Page 11: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/11.jpg)
Subbasin Configuration
7 Subbasin 50 Subbasin
252 Subbasin
![Page 12: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/12.jpg)
Climate Data
Annual PrecipitationMGM : 840 mmCFSR : 1011 mm
![Page 13: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/13.jpg)
Subbasin No vs. Climate Data
![Page 14: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/14.jpg)
Subbasin No vs. Climate Data
![Page 15: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/15.jpg)
Model Configuration Setup
Subbasin Number
Landuse Dataset
Climate Dataset
7 50 252
Corine Landsat Statip
CFSR MGM
• 18 Different Model Setup7_COR_CFSR252_LAND_MGM50_STA_MGM
SubbasinNo_Landuse_Climate
![Page 16: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/16.jpg)
Methodology
SWAT Model Setup
Preliminary Analysis
Elevation Band Adj.
Parameter Adjustment
Land Use
Calibration Complete
Soil
Topography
Observed Streamflow
Climate DataEvaluation
Re-Setup Model
Evaluation
Evaluation
Change Climate Data
Evaluation of all model
outputs
(MGM)
(CFSR)
Change LULC Data
ChangeSubbasin No
![Page 17: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/17.jpg)
Comparison of Subbasin No.
0102030405060708090
100110120
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
113
120
127
134
141
148
155
162
Dis
char
ge (m
3s-
1 )
7_COR_CFSR 50_COR_CFSR 252_COR_CFSR
0102030405060708090
1 7 13 19 25 31 37 43 49 55 61 67 73 79 85 91 97 103
109
115
121
127
133
139
145
151
157
163
Dis
char
ge(m
3s-
1 )
7_STA_MGM 50_STA_MGM 252_STA_MGM
Subbasin Config.
• First Simulationsbefore calibration
• In order tocompare effects of subbasin number
• No obs. data
![Page 18: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/18.jpg)
Comparison of Landuse Data
0102030405060708090
100110120
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
113
120
127
134
141
148
155
162
Dis
char
ge (m
3s-
1 )
50_STA_CFSR 50_LAND_CFSR 50_COR_CFSR
01020304050607080
1 7 13 19 25 31 37 43 49 55 61 67 73 79 85 91 97 103
109
115
121
127
133
139
145
151
157
163
Dis
char
ge (m
3s-
1 )
7_STA_MGM 7_LAND_MGM 7_COR_MGM
• First Simulationsbefore calibration
• In order tocompare effects of landuse datasets
• No obs. data
![Page 19: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/19.jpg)
Comparison of Climate Data
020406080
100120140
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
113
120
127
134
141
148
155
162
Dis
char
ge (m
3s-
1 )
50_STA_MGM 50_STA_CFSR
020406080
100120140
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
113
120
127
134
141
148
155
162
Dis
char
ge (m
3s-
1 )
252_LAND_MGM 252_LAND_CFSR
• First Simulationsbefore calibration
• In order tocompare effects of climate data
• No obs. data
![Page 20: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/20.jpg)
Perfomance Criteria for First Sim.
Model Name R2 bR2 NS BPIAS7_COR_CFSR 0.79 0.5594 0.63 31.67_COR_MGM 0.52 0.1602 -0.22 67.17_STA_CFSR 0.78 0.5574 0.64 317_STA_MGM 0.48 0.1404 -0.25 67.47_LAND_CFSR 0.78 0.5588 0.64 317_LAND_MGM 0.51 0.1576 -0.21 66.650_COR_CFSR 0.78 0.5843 0.68 26.950_COR_MGM 0.48 0.1546 -0.13 62.350_STA_CFSR 0.78 0.5842 0.68 26.150_STA_MGM 0.45 0.1375 -0.16 62.550_LAND_CFSR 0.78 0.5845 0.68 26.350_LAND_MGM 0.48 0.1533 -0.12 61.7252_COR_CFSR 0.78 0.5853 0.67 26.8252_COR_MGM 0.46 0.1455 -0.09 59.1252_STA_CFSR 0.78 0.5852 0.68 26252_STA_MGM 0.43 0.1294 -0.12 59.2252_LAND_CFSR 0.78 0.5857 0.68 26.2252_LAND_MGM 0.45 0.1439 -0.08 58.6
![Page 21: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/21.jpg)
Flow Duration CurveFlow Duration curve for the first simulations of all 18 models
Disc
harg
e(m
3s-1
)
Observe Discharge Data
Results for CFSR dataset
Results for MGM dataset
![Page 22: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/22.jpg)
0
20
40
60
80
100
120
140
160
1801 7 13 19 25 31 37 43 49 55 61 67 73 79 85 91 97 103
109
115
121
127
133
139
145
151
157
163
Dis
char
ge (m
3s-
1 )
50_STA_CFSR50_STA_MGMObserved Data
Model Name R2 bR2 NS BPIAS50_STA_CFSR 0.78 0.5842 0.68 26.150_STA_MGM 0.45 0.1375 -0.16 62.5
Ex. time series for first simulation
![Page 23: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/23.jpg)
SWAT Model Parameters
SWAT Parameters Initial Range
r__CN2.mgt -0.5 to 0.5
r__GWQMN.gw -0.5 to 0.5
r__GW_REVAP.gw -0.5 to 0.5
r__SOL_AWC().sol -0.5 to 0.5
r__REVAPMN.gw -0.5 to 0.5
r__ESCO.hru -0.2 to 0.2
r__ALPHA_BF.gw -0.5 to 0.5
r__SOL_K().sol -0.5 to 0.5
r__SOL_BD().sol -0.5 to 0.5
480 simulations has been done for each model
• Plapse Rate is chosenas 600mm/km (Cuceloglu and Ozturk, 2017)
• SWAT-CUP used for model calibration/validation
![Page 24: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/24.jpg)
020406080
100120140160180
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
113
120
127
134
141
148
155
162
Dis
char
ge (m
3s-
1 )
50_STA_CFSR 50_STA_MGM Observed Data
020406080
100120140160180
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
113
120
127
134
141
148
155
162
Dis
char
ge (m
3s-
1 )
50_STA_CFSR 50_STA_MGM Observed Data
First Simulation(Before Calibration)
Best Simulations(After Calibration)
![Page 25: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/25.jpg)
Model Performance/Flow Duration Curve
Calibration Validation
Model Namep-factor
r-factor R2 bR2 NS BPIAS
p-factor
r-factor R2 bR2 NS BPIAS
7_COR_CFSR 0.73 1.07 0.84 0.6459 0.78 18.6 0.82 1.06 0.8 0.6225 0.78 11.97_COR_MGM 0.82 1.11 0.89 0.6903 0.86 7.5 0.8 1.03 0.63 0.3932 0.61 10.850_STA_CFSR 0.77 1.09 0.84 0.6666 0.8 14.6 0.88 1.09 0.81 0.6096 0.8 2.450_STA_MGM 0.88 1.11 0.88 0.7304 0.88 0.4 0.61 0.81 0.42 0.2218 0.34 20.6252_STA_CFSR 0.8 1.09 0.84 0.6675 0.8 14.4 0.88 1.09 0.81 0.6099 0.8 2.3252_STA_MGM 0.77 1.08 0.87 0.7049 0.86 1.6 0.58 0.8 0.42 0.2168 0.35 17.6
First Simulation(Before Calibration)
Best Simulations(After Calibration)
![Page 26: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/26.jpg)
CV Value for Streamflow
![Page 27: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/27.jpg)
• The study shows the climate data set is the mostinfluencial dataset and has a larger uncertainity in ourstudy area.
• Global climate dataset (reanalysis data) representsbetter than the local data.
• The Corine LULC dataset (freely available) yield quitesimilar results as the local landuse datasets. Detailedlocal dataset (STATIP) has the smallest uncertainty.
• Subbasin number has no significant effect on themonthly calibrated discharge data but it is important toevaluate relation between the subbasin and climatedata.
• Using elevation band is quite efficient (if the availabledataset catches the dynamic of the system)
Results Discussion
![Page 28: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/28.jpg)
Conclusion
• Effects of different input data and subbasin configuration areinvestigated in Melen Watershed, Turkey
• SWAT model has been succesfully applied to the region.• Hydrological simulations could be significantly different for
different model setup configurations and could cause largeuncertainity.
• Multisite and daily calibration are also important toinvestigate
• Nowadays it is easy (relatively) to find multiple datasetstherefore modeler should take different these datasets intoconsideration. It is strongly recommend to use different inputdatasets for modelling studies.
![Page 29: Evaluating the Impact of Different Input Datasets and ... · • Performance of the watershed modeling studies is closely associated with the model input datasets • Numerous databases](https://reader031.vdocuments.us/reader031/viewer/2022011921/603adddd88a096578e4b6fc7/html5/thumbnails/29.jpg)
Thanks for your attention...