Effect of DEM Uncertainty on the Distributed Hydrological Model TOPMODEL

TOPMODEL is a semi-distributed hydrological model in which the distributed predictions of catchments response to rainfall are made. In this process of simulation, digital elevation models (DEM) is required to provide the values of parameters, such as topographic index, cumulative area of catchment and distance from catchment outlet; thus DEM play a dominant role in TOPMODEL implementation. Generally, DEM has inherent errors, referred to as DEM uncertainty. Lacking of knowledge about DEM errors, DEM data is often used in hydrological applications by using TOPMODEL without quantifying the effects of DEM uncertainty. In fact, the uncertainty of DEM may strongly influence the simulation results produced by TOPMODEL. Unfortunately, this effect is largely ignored in many empirical researches. This study aimed to examine the impacts of DEM uncertainty on the simulation results of TOPMODEL from the study area - Jiaokou watershed, a sub- basin with an area of 259 km² of the Yongjiang River in southeast China. This paper mainly discussed the effects in both quantitative and qualitative aspects. First, DEM uncertainty was simulated using the Monte Carlo method, and for every realization of the DEM, the topographic index, cumulative area of catchment and distance from catchment outlet were calculated. Second, TOPMODEL was tested and the results saved as the four statistical indices of the simulation: EFF (the Nash and Sutcliffe efficiency criterion), SSE (Sum of squared residuals over all time steps), SLE (Sum of squared log residuals over all time steps) and SAE (Sum of absolute errors over all time steps) under the condition of seven storm events. Third, the statistic results - min, max, range, standard deviation, and Mean Value, of the four indices from the simulated DEM were compared in each case of flood and Mean value of the four indices were picked up to evaluate the effect of uncertainty of DEM on TOPMODEL. Finally, the simulation hydrographs were compared wi- - th the hydrographs using the original DEM under each flood event. The biggest errors for the indices of EFF, SSE, SLE, and SAE were 0.0169, 2.3x10^-5, 1.1375, and 0.0033 respectively, which showed that the effect of DEM uncertainty on TOPMODEL was inconsiderable and could be ignored in the model´s application.