Flood Risk Mapping in Africa: Exploring the Potentials and Limitations of SRTM Data in the Lower Limpopo, Mozambique

Abstract: Many regions in Africa are presently faced with an increasing flood risk due to impending climate change and population growth. One useful mitigation strategy to decrease this risk would be to map it, so that urban planning, warnings systems and emergency response subsequently could be designed to reduce societal vulnerability. This is, however, not widely feasible on the African continent, as developing countries often lack access to the topography and discharge data required to produce high- quality flood risk maps. To seek a way around this problem, on-going research is investigating the possibility of obtaining alternative model inputs, by using global datasets of elevation, derived from remote sensing, and methods to estimate flood flows. This thesis presents a case study within this context where the aim was to determine the accuracy of an African catchment-scale flood map, produced with the satellite product SRTM (Shuttle Radar Topography Mission) as topography input, and to explore the potentials and limitations of such a model scheme. Two high-magnitude floods, occurring in year 2000 and 2013 in the Lower Limpopo Basin (Mozambique), were modelled for inundation extent, using a no-channel 2D model built for the LISFLOOD-FP flood modelling software. Flood water levels were also simulated to assess the models vertical performance. Model outcomes were evaluated against satellite imagery and recordings of high watermarks, adjusting the value representing the roughness of the floodplain to optimize flood extent correspondence. Due to different hydrograph dynamics, simulations of the two floods required different values of roughness (0.02 and 0.09 s m-1/3) to reach maximum accuracy (F = 0.59 and 0.64, respectively). However, the results also indicated that a model calibrated with a flood of relatively low return period potentially could be used to map rare flood events. Simulation inaccuracies were mainly attributed to (1) reservoirs and streams, temporarily connecting to the river system during high flow conditions, (2) limitations of the topography data, in terms of recognizing riverbed geometry and floodplain micro-topography, and (3) cloud cover, reducing the accuracy of flood extent reference data. The vertical simulation accuracy, with an average error of ± 2 m, was well within the uncertainty bounds of input data. Errors were in this case ascribed the SRTM’s representation of high slope terrain and possible radar speckles in urban areas. The findings of this study indicate that there is high potential in using SRTM data for mapping of high-magnitude flood risk in Africa, but also that consideration to river system complexity is crucial.