Methodologies for analyzing intrinsic and required DEM accuracy for hydrological applications of flash floods

Abstract

For extreme events like flash floods, infiltration is considered to be negligible, and the morphology of the watershed is considered to be the most significant factor. Thus, digital elevation models (DEMs) are the most efficient data source to define the catchment surface. To determine if the DEM is adapted, it is essential to analyze both the intrinsic accuracy and the required accuracy. In this paper we propose two complementary methodologies to analyze and evaluate these two kinds of DEM accuracy that relate to the hydrological applications to flash floods. The first methodology relates to the intrinsic accuracy: a diagnostic method analyzes the accuracy and stability of the extracted hydrographic network at catchment scale. To obtain correct positioning of the channels, results show the strong influence of the topographic context and the need for associating extra data on rivers, especially in flat areas. The impact of accuracy is evaluated through a scheme based on DEM grid rotation. A step by step and iterative process gives a fuzzy evaluation of the hydrographic network. These techniques also highlight the strong scale dependence of the extracted network. The second methodology relates to the required accuracy. To analyze the sensitivity to the DEM accuracy, we propose an approach by comparing the results of hydraulic models obtained from a "real" description of the river topography and from this description distorted by a numerical noise characterizing the lack of accuracy of the DEM. The impact of this noise on the overflows on the major bed is analyzed. The results show that the accuracy required by the thematicians is often greater than that strictly required by modeling, which opens interesting prospects to reduce the phases of data acquisition.