Analysis of uncertainty associated with defining the objective functions for the calibration of a SWAT model for the Narmada River basin

The development of a hydrological model for simulating streamflow with minimum uncertaintyrequires awareness of the different processes occurring in the river basin. The distributedhydrologic models have highly uncertain input data including land use, soil characteristics,climate data, and discharge data from gauge stations. Further, the identification of the mostappropriate set of sensible parameters is important in hydrological modeling. Uncertaintiesassociated with the model parameter estimation are reduced by calibration and validation. Thisstudy focuses on the uncertainty added by different objective functions to the calibration ofmodels. Six objective functions (R2, NSE, BR2, PBIAS, KGE, and RSR) are selected forcalibrating nine subbasins of the Narmada River Basin. We performed calibration anduncertainty analysis for monthly and daily timesteps with SUFI-2 (Sequential Uncertainty Fittingversion 2) using SWAT- Calibration and Uncertainty Procedure (SWAT_CUP). Propagation ofmodel input uncertainty to outputs is evaluated by using 95 % prediction uncertainties (95PPU)in SWAT-CUP. The comparison of 95 PPU band with observed data is performed by computingthe p-factor and r- factor. The results show that all the objective functions give satisfactorycalibration and validation results for monthly and daily time steps but in different parameterranges. Amongst the selected objective functions, NSE produces the best results in our studyarea. The results of the ongoing work will be presented at the conference.