Comparative performance of SWAT and HEC-HMS in modeling runoff in tropical catchments.

Abstract

Hydrologists are increasingly finding it difficult to choose from a growing pool of hydrologic models for the different model applications in data scarce regions like Uganda. In this study, performance comparison of SWAT and HEC-HMS in event and continuous modeling of rainfall runoff in two tropical catchments (Manafwa and Sezibwa) at different altitudes was done. Two methods were used to compare the models; (1) multi-criteria comparison based on three objective functions, the Nash Sutcliff Efficiency (NSE), Coefficient of determination (R2) and Percent bias (PBIAS) and (2) testing for statistical significance of the differences in the simulated results using the Wilcoxon Rank-sum test. On a continuous time scale, both models gave satisfactory results with the differences not being statistically significant. The results for calibration for HEC-HMS were; Manafwa (NSE>0.5, R2>0.69 & PBIAS<5.3), Sezibwa (NSE>0.5, R2>0.72 & PBIAS<26.32) while for SWAT; Manafwa (NSE>0.44, R2>0.68 & PBIAS<5.3) and for Sezibwa (NSE>0.5, R2>0.66 & PBIAS<16.9). On short events, the performance of HEC-HMS was better than SWAT. The results for HEC-HMS were; Manafwa (NSE>0.4, R2>0.7 & PBIAS<16.3); Sezibwa (NSE>0.51, R2>0.87 & PBIAS<6.25) while for SWAT; Manafwa (NSE>0.1, R2>0.7, PBIAS<41.9) and for Sezibwa (NSE>0.1, R2>0.74, PBIAS<16.7). Based on the Wilcoxon Rank-Sum test, for continuous simulation the difference in performance for the two models was not statistically significant (validation p value=0.09406 and 0.2008 for Manafwa and Sezibwa respectively) while during event simulation HEC-HMS performed better than SWAT (validation p value =0.0424 and 0.002924 respectively). The results show that in data limited scenarios, both SWAT and HEC-HMS give satisfactory results and there is no significant difference in their performance while using climate data processed by Monthly to daily weather converter (MODAWEC) software. The choice for the model in continuous modeling applications should therefore be based on modelers ability to properly set up and calibrate the model and availability of input data. The results further show that HEC-HMS performs significantly better than SWAT in event simulation and for such applications the former should be the first choice model.