Modelling Time-Variant Parameters of a Two-Parameter Monthly Water Balance Model

Chao Deng,Pan Liu,Weiguang Wang,Quanxi Shao,Dingbao Wang

DOI: https://doi.org/10.1016/j.jhydrol.2019.04.027

IF: 6.4

2019-01-01

Journal of Hydrology

Abstract:Parameters are usually assumed to be stationary, and regarded as constants in hydrological modelling. However, studies have shown that model parameters may vary temporally due to human activities, and climate variability and changes. This study proposes a framework for quantifying hydrological model parameters as functions of time-variant catchment properties, aiming to improve the capability of capturing hydrograph and the extra-polative ability of hydrological models under a changing environment. The framework includes four steps: (1) estimating model parameters by using an ensemble Kalman filter based on the observations; (2) analyzing correlations between estimated parameter values and catchment properties; (3) constructing a set of parameter functions based on the identified catchment properties and the proposed functional forms; and (4) selecting and testing the best-performed model. As a demonstration of the framework, the monthly water balance for Tongtianhe and Ganjiang basins in China with different climate are simulated using a two-parameter monthly water balance model. Results from the Tongtianhe Basin show that considering the time-variant evapotranspiration parameter (C) brings improvement in low flows when water storage capacity (SC) is treated as constant since there is no significant improvement in runoff simulation when SC is treated as a time-variant parameter. The application to Ganjiang Basin shows that reasonable improvements are achieved when both C and SC are treated as time-variant, especially in low and high flows. The presented framework provides a useful tool for estimating time-variant parameters of hydrological models, thereby leading to more reliable model predictions under a changing environment.