Assessment of Open Access Global Elevation Model Errors Impact on Flood Extents in Southern Niger

Abstract

Recent studies have estimated annual flood loss at over two billion USD dollars in damage per year globally. This figure is likely to increase as populations grow, people move to flood prone areas, and flooding dynamics change. Digital Elevation Models are a primary input into many flood models; therefore, the accuracy and resolution of these data sets have implications for model accuracy that are not completely understood. An increasing number of DEMs are becoming open and freely available, which presents analysts with the challenge of selecting the most appropriate inputs for their specific use case, especially in countries such as Niger where these global DEMs have not been formally evaluated. This study compares the absolute vertical accuracy of four global and freely available DEMs: Shuttle Radar Topography Mission, ASTER, Advanced Land Observing Satellite, and MERIT, with reference to a high-resolution LiDAR DEM, in southern Niger. Specifically, the Height Above Nearest Drainage (HAND) model is used to investigate the impact of resolution and error on simulated flood extent. There are currently no studies that assess the vertical accuracy of global DEMs in Niger nor that investigate the impact of the DEM geometric variability on flood extents. The ALOS DEM showed the lowest Root Mean Square Error of 1.19 m when compared to the LiDAR DEM, while ASTER had the highest RMSE of 4.2 m. The results of the Height Above Nearest Drainage model showed that at floods under 3 m in depth, the higher resolution DEMs simulated less flood extent; however, once the flood depth exceeded 3 m, the higher the resolution the DEM, the larger the flood extent. These findings suggest that both absolute error and relative error, as well as resolution, should be addressed when applying a DEM to a flood model.