Comparison of multi-objective and single objective calibration for SWAT model: a case study on Musi river basin, India

ABSTRACT Calibration of hydrological models for watersheds is critical considering the hydrological processes involved. The Soil and Water Assessment Tool (SWAT) is one such popular model and requires proper calibration, without which models have difficulty in proper simulation of runoff. The present study aims to utilize multi-objective calibration framework using Non-Dominated Genetic Algorithm- II (NSGA-II) and SWAT-Calibration Uncertainty Procedures (SWATCUP) for calibration. The study is conducted on Musi river basin located in India (10,000 Sq km) for seven years from 2013–2016. It includes an initial warm-up period of three years, the calibration period from 2015–2016, and validation period from 2014–2015. NSGA-II aims to optimize the multiple objective functions i.e. Nash Sutcliffe Efficiency (NSE) and Percentage Bias (PBias). The Monthly simulations results are expressed in terms of statistical parameters NSE, R2 and PBias for calibration and validation period. The results indicate satisfactory performance. Further, NSGA-II results are compared with SWATCUP (Sequential Uncertainty Fitting ver.2 (SUFI-2). We find NSGA-II performance is better than SWATCUP. The sensitive analysis indicates that CN2, GW_DELAY, GW_REVAP, ALPHA_BF, RCHRG_DP, and CH_K2 are very sensitive whereas SURLAG, ESCO, SLSUBBS, HRU_SLP are observed to be least sensitive.