Distributed rainfall-runoff simulation for a large-scale karst catchment by incorporating landform and topography into the DDRM model parameters

Abstract

Runoff movement in karst catchments is of great complexity. Although many models have been developed for the rainfall-runoff simulation in karst catchments, the distributed ones, capable of taking into consideration the impacts of the spatial variability of hydrogeological factors, are few and can only be applied to well-explored karst catchments of commonly less than 100 km2 in area. In this study, the previously-published DEM (Digital Elevation Model) -based distributed rainfall-runoff model (DDRM) was applied and modified for distributed rainfall-runoff simulation in a large-scale karst catchment by assuming the main parameters of runoff generation and routing for each grid cell to be a function of karst landform and topographic index (i.e., the mathematical expression of topographic features) to represent the hydrological effects of epikarst and underground river system. For model structure investigation, three versions of the modified DDRMs, i.e., the DDRM considering only karst landform (DDRM_K), the DDRM considering only topographic index (DDRM_T), and the DDRM considering both karst landform and topographic index (DDRM_KT), were set up and compared with the original DDRM (i.e., the original DDRM, termed as DDRM_O). All four DDRMs were calibrated with a multi-objective optimization framework that uses the Kling-Gupta Efficiency of streamflow (KGEQ) and the autocorrelation function of streamflow (KGEACF) as objective functions to improve parameter identifiability. The investigation was conducted in the Xijiang Basin, which is a typical karst catchment in Southwest China with an area of 309300 km2. The results show that, compared with DDRM_O, DDRM_T barely benefits the rainfall-runoff simulation, while both DDRM_K and DDRM_KT noticeably enhance the simulation accuracy in terms of KGEQ. Due to the better performance in high flow and autocorrelation function of streamflow, DDRM_KT outperforms DDRM_K in general and has the potential to be successfully applied to large-scale karst catchments.