Practical experience and framework for sensitivity analysis of hydrological models: six methods, three models, three criteria

Abstract

Abstract. Sensitivity Analysis (SA) and Uncertainty Analysis (UA) are important steps for better understanding and evaluation of hydrological models. The aim of this paper is to briefly review main classes of SA methods, and to presents the results of 10 the practical comparative analysis of applying them. Six different global SA methods: Sobol, eFAST, Morris, LH-OAT, RSA and PAWN are tested on three conceptual rainfall-runoff models with varying complexity: (GR4J, Hymod and HBV) applied to the case study of Bagmati basin (Nepal), and also initially tested on the case of Dapoling-Wangjiaba catchment in China.The methods are compared with respect to effectiveness, efficiency and convergence. A practical framework of selecting and using the SA methods is presented. The result shows that, first of all, all the six SA methods are effective. Morris and LH-15 OAT methods are the most efficient methods in computing SI and ranking. eFAST performs better than Sobol, thus can be seen as its viable alternative for Sobol. PAWN and RSA methods have issues of instability which we think are due to the ways CDFs are built, and using Kolmogorov-Smirnov statistics to compute Sensitivity Indices. All the methods require sufficient number of runs to reach convergence. Difference in efficiency of different methods is an inevitable consequence of the differences in the underlying principles. For SA of hydrological models, it is recommended to apply the presented practical 20 framework assuming the use of several methods, and to explicitly take into account the constraints of effectiveness, efficiency (including convergence), ease of use, as well as availability of software.

Impact Assessment of Rainfall-Runoff Simulations on the Flow Duration Curve of the Upper Indus River—A Comparison of Data-Driven and Hydrologic Models

“…Rathod and Manekar [37] identified that the imperviousness of the catchment area is the least influencing parameter. According to Wang and Solomatine [58], the time to peak is the most influential model parameter, while Ghumman et al [1] observed that the kinematic wave parameter is the most contributing hydrologic element in rainfall runoff modeling. This discussion shows that it is wise to perform a sensitivity analysis of hydrologic model parameters for each catchment in order to achieve better understanding of rainfall-runoff behavior.…”

Exploring the role of soil storage capacity for explaining deviations from the Budyko curve using a simple water balance model

“…We varied the four parameter β , Smax, FCfrac and kres within a defined limits (Table 3) using the shuffled complex evolution SCE-UA (Duan et al, 1994) uniform sampling scheme. The parameter ranges were defined in close accordance with other 235 studies (Beck et al, 2016;Osuch, 2015;Piotrowski et al, 2017;Wang and Solomatine, 2018).…”