Diagnosing Credibility of a Large-Scale Conceptual Hydrological Model in Simulating Streamflow

Abstract

A robust hierarchical operational testing scheme has been adopted to test the credibility of a newly developed large-scale conceptual hydrological model in the Subarnarekha basin, India. We have evaluated the results using M [combining Nash-Sutcliffe efficiency (NSE), and R2], pPBIAS-, and pRMSE-values [combining the jackknife method and analysis of variance (ANOVA)] along with root-mean square error (RMSE), NSE, R2, and percent bias (PBIAS). Results show that the model has passed various steps of the testing scheme successfully, and is both geographically and climatically transposable. The results of a proxy-basin test [transferring parameters into Ghatshila subbasin after calibration in Muri and crossvalidation in Jamshedpur (M=0.62)] show satisfactory RMSE (271.14 m3/s), NSE (0.42), R2 (0.45), PBIAS (30.22%), and statistically significant pRMSE-value (0.043) and pPBIAS-value (0.002) to confirm the model as geographically transposable. Satisfactory RMSE (9.71 m3/s), NSE (0.66), R2 (0.84), and PBIAS (1.66%) for dry year validation under differential split-sample test in Muri subbasin confirms the model as climatically transposable. The uncertainty analysis, using quantile regression technique, represents reasonable predictive ranges, e.g., for Muri calibration and validation 95 percent prediction uncertainty (PPU) band encloses 63% and 76% of the observations.