Development of an inexact-variance hydrological modeling system for analyzing interactive effects of multiple uncertain parameters

Uncertainty assessment of hydrological model parameters has become one of the main topics due to their significant effects on prediction in arid and semi-arid river basins. Incorporation of uncertainty assessment within hydrological models can facilitate the calibration process and improve the degree of credibility to the subsequent prediction. In this study, an inexact-variance hydrological modeling system (IVHMS) is developed for assessing parameter uncertainty on modeling outputs in the Kaidu River Basin, China. Through incorporating the techniques of type-2 fuzzy analysis (T2FA) and analysis of variance (ANOVA) within the semi-distributed land use based runoff processes (SLURP) model, IVHMS can quantitatively evaluate the individual and interactive effects of multiple uncertain parameters expressed as type-2 fuzzy sets in the hydrological modeling system. The modeling outputs indicate a good performance of SLURP model in describing the daily streamflow at the Dashankou hydrological station. Uncertainty analysis is conducted through sampling from fuzzy membership functions under different α-cut levels. The results show that, under a lower degree of plausibility (i.e. a lower α-cut level), intervals for peak and average flows are both wider; while intervals of peak and average flows become narrower under a higher degree of plausibility. Results based on ANOVA reveal that (i) precipitation factor (PF), one of main factors dominating the runoff processes, should be paid more attention in order to enhance the model performance; (ii) retention constant for fast store (RS) controls the amount and timing of the outflow from saturated zone and has a highly nonlinear effect on the average flow; (iii) the interaction between retention constant for fast store (RF) and maximum capacity for fast store (MF) has statistically significant (p<0.05) effect on modeling outputs through affecting the maximum water holding capacity and the soil infiltration rate. The findings can help generate the optimal system inputs and enhance the model’s applicability.