Multi-objective Evaluation And Uncertainty Analysis For The SWAT Application In West Seti River Basin,Nepal

Saroj Karki¹, Narendra Man Shakya¹, Pabitra Gurung², Luna Bharati²

1 Institute of Engineering, Pulchowk Campus, Tribhuvan University, Kathmandu, Nepal (sarojioe@gmail.com), 2 International Water Management Institute (IWMI), Kathmandu, Nepal (p.gurung@cgiar.org)

---------------------------------------------------------------------------------------------------------------------------------------------------

The application of hydrological models is increasing in response to rising demands for economic and efficient use of available water and thus for the formulation of appropriate water management strategies to cope future water challenges. These hydrological models are simplification of reality and therefore there is always some degree of associated risk or uncertainty. Proper calibration and evaluation of model is necessary to recognize which aspects of the prediction works well and which one needs improvement. It is therefore important that the modelers quantify the uncertainties and thus represent the reliability of model predictions before it is used in decision making process.

A popular semi-distributed hydrological model, Soil and Water Assessment Tool (SWAT), is used to perform hydrological analysis of West Seti River Basin located in the far western part of Nepal which is one of the least explored regions of the country. Different statistical indices & hydrological measures are implemented to assess the performance of the model. Study shows that no single measure of model performance can fully evaluate the model. Thus the analysis advocates the shift from single to multi-objective calibration. Two methods of uncertainty analysis, meta-Gaussian and Sequential Uncertainty Fitting (SUFI-2), are adopted for the estimation of uncertainty. The uncertainty analysis of the SWAT simulated daily discharge based on meta-Gaussian approach covered most of the observed data inside the 95% confidence interval but on the expense of wider prediction band while that by SUFI-2 covered less data within relatively narrower 95PPU band. In both methods, most of the peak flows as well as the recession flow didn’t fall within 95% confidence interval band indicating high uncertainty in SWAT prediction for such flows. Given the uncertainty in the model prediction, its implications in the decision making process should be properly accounted for.

Key Words: Hydrological modeling, Multi-objective evaluation, Uncertainty analysis, SUFI-2,    SWAT  

------------------------------------------------------------------------------------------------------------------------------

An Abstract selected for the 3rd SWAT Conference & Workshop in South East and East Asia (SWAT SEEA III), June, 2013. 

Bogor, Indonesia