The Impact of Climate Data Uncertainty on Calibration of the SWAT Model

Transcription

1 The Impact of Climate Data Uncertainty on Calibration of the SWAT Model SWAT conference, Toulouse, France, July B. Kamali, Jamshid Mousavi, Karim Abbaspour, Hong Yang Eawag: Das Wasserforschungs-Institut des ETH-Bereichs

2 s Outlet Positions Parameterizati Parameterization on Parameterizati on Search Algorithm 1 Successful application of hydrologic How can models the depends structure on of Calibration the SWAT Process govern results of simulator? Outlet Positions Nonuniqueness Nonuniqueness Calibration Uncertainty Model Structure structure Calibration Input climate Uncertainty data data Search Algorithm Different sources of climate data are available for a watershed 1. Which data sources are more reliable? Model Structure 2. Can adding more quantitative climate information be helpful? Input Data 2 What is the impact of climate data on model output? Material Results conclusion 1

3 Study area Karkheh River Basin (KRB): Located in South Western of Iran Area: km 2 Climate: Semi -Arid It is the third largest catchment in the country It is considered as the food basket of Iran Four sources of climate data were available for the catchment Material Results conclusion 2

4 Climate data Climate Sources 12 (C1) (C2) 3 and with 4 (C3&C4) 923 local stations with 30 grid points Iranian Ministry of Energy Weather C3: (Climate Service Research of Iranian Unit) Meteorological CRU Organization C4: GFDL-ESM2M (GCM) Dataset Dataset 1 Dataset 2 Dataset 3 Dataset 4 Dataset 5 Dataset 6 Climate C1 C2 C3 C4 C1& C2 Best C1&C2 Material Results conclusion 3

5 SWAT Structuring o o The results of first Set-up were not convincing Simulated discharge values were far from observed discharge values Three important issues governed the structure of simulator 1: Stream Network Construction in SWAT Initial drainage area: hec Initial drainage area: 8000 hec O1 O1 Wrong Placement O2 Manual Adjustment O2 Material Results conclusion 4

6 SWAT Structuring 2: Considering discharge values after intersection of tributaries P3 O3 P1 P2 P1 and P2: Underestimated P3: far from real position Increasing resolution through changing threshold imposed many calculation cost Add manual outlets Material Results conclusion 5

7 Model Structuring 3. Lack of sufficient data may lead to wrong information The resolution in structure of SWAT should be compatible with its P2 ability in representing watershed characteristics o Outlet P1 underestimates O4 o Outlet P2 matches O4, but is wrong Modelers should prevent from imposing unnecessary information resolution and calculation costs o Wrong information lead to wrong Modelers should prevent from parameter giving adjustment wrong information to O4 P1 the model Material Results conclusion 6

8 Final Observed outlets o After model restructuring, 12 outlets used for calibration o 8 outlets on major rivers o 4 outlets on tributaries The performance of different climate sets were compared in terms of calibrating 12 discharge outlets Material Results conclusion 7

9 Evaluation the performance of different datasets Before calibration o Dataset1 with fewer numbers of stations performed better o Adding more stations does not necessarily lead to better simulations o Qualitative information is more helpful than quantitative br2 value Before Calibration DataSet1 DataSet2 DataSet3 DataSet4 DataSet5 DataSet6 9 climate stations + 23 climate stations How do these datasets perform after calibration? Material Results conclusion 8

10 Evaluation the performance of different datasets.7 After Calibration Before Calibration 0.7 After Calibration.6.5 o Dataset4 did not much improved o Datasets 1&3&6 have similar performance br2 value o Datasets 2&5 have similar performances and in between DataSet1 DataSet2 DataSet3 DataSet4 DataSet5 DataSet6 0.1 DataSet1 DataSet2 DataSet3 DataSet4 DataSet5 DataSet6 Different Datasets have similar mean performances How do they perform in final adjusted parameters? Material Results conclusion 9

11 Comparing the final parameter sets 1.5 br2 SLSUBBSN 1.5 Small CV in br2 br2 CN CV 0.5 CV 0 CV Z1 Z2 Z3 Z4 Z5 Z6 Z7 Z8 br2 REVAPMN -1 Z1 Z2 Z3 Z4 Z5 Z6 Z7 Z8 Large CV in adjusted 1.5 parameter sets br2 CH-N CV CV CV Z1 Z2 Z3 Z4 Z5 Z6 Z7 Z8 0 Z1 Z2 Z3 Z4 Z5 Z6 Z7 Z8 Material Results conclusion 10

12 Conclusion o o o o Each dataset adjusts the model with different parameter sets There are uncertainties within adjusted parameters obtained from different sources These uncertainties will propagate in model output These uncertainties need to be considered at the first step before the results are used for next management policies Material Result Conclusion 11

13 Thanks For Your Attention! Eawag: Das Wasserforschungs-Institut des ETH-Bereichs

14 Observed variable Algorithm Function SWAT and SWAT-CUP Elevation map Land use & Soil map Climate data SWAT Calibration br2 SWAT-CUP SUFI2 Algorithm Discharge Values Material Results conclusion 3

15 Structure of presentation Introduction Conceptual Hydrologic hydrologic modeling models and calibration are useful tools to: Model Analyze structure hydrologic and characteristics climate data of watersheds Material Water resource management SWAT Climate and change SWAT-CUP impacts Study Area; Successful Data Preparation; application of hydrologic models depends on Calibration Process Results Model re-structuring; Climate data impact Conclusion Material Results conclusion 12

16 Observed variable Algorithm Function SWAT and SWAT-CUP Elevation map Land use & Soil map Climate data SWAT Calibration br2 SWAT-CUP SUFI2 Algorithm Discharge Values Material Results conclusion 3