Modeling floods in the Middle Zambezi Basin using remote sensing and hydrological modeling techniques.

Abstract

Mbire District, in the Middle Zambezi Basin, experiences floods annually. The study aimed at applyingremote sensing and hydrodynamic modeling tools to map, and understand flood processes in order to improve flood management in the district. In determining the spatial and temporal variation of flood inundation in the district, NDVI MODIS images for the period 2005-2006, 2008-2009, 2013-2014 and 2014-2015 were processed in ILWIS GIS environment. The above period was classified by the Civil Protection Unit and ZINWA as flood years. Validation of the MODIS derived flood areas was done using 68 GCPs collected using participatory GIS mapping methods. A binary logistic regression modelthrough the SPSS software was used to determine the spatial variation of flood hazard as a function of environmental factors. The results were confirmed using hydrologic modeling techniques, where the HEC-HMS model helped to quantify the peak flow and runoff contributed by the three sub-basins in the Mbire District (Angwa, Musengezi and Lower Manyame). The HEC-RAS model was used to map inundated areas for the Lower Manyame Basin for the flood return periods, 10, 25, 50 and 100 years. Flood mapping using MODIS images showed that the maximum areas flooded is 1934km2 on 16 January 2006, and 1 895 km2 on 8 January 2015. A good agreement of (R2=0.86) between GCPs and MODIS derived flooded area for 8 January 2015 was recorded. Environmental factors that significantly explained flooding are distance from water bodies (p<0.05). Simulations through the HEC-HMS model indicated an average yearly observed flow rates of 15.6 * 107 m3 16.2 * 107 m3, and 25.7 * 107 m3 for Lower Manyame (Mapomha), Musengezi (C109) and Angwa (Angwa) basins respectively. These flows were against an average discharge of 19.7 * 107 m3/yr.,18.3 * 107 m3/yr. and 25.4 * 107 m3/yr. for Lower Manyame, Musengezi, and Angwa respectively. Model performance was evaluated and theefficiency for Musengezi showed a RMSE of 5.25 %, RBIAS of 0.04 % for Angwa the RMSE was 3.94 % and RBIAS of -0.003 % and Manyame gave a RMSE of 5.25 % and RBIAS of 0.07 %. The HEC-RAS simulated inundated areas are 56.3 km2, 57.3 km2, 58.4 km2, 58.7 km2, 59.1 km2 for the 2008-9 season, 10 year, 25 year, 50 year and 100 year return floods respectively and these are in and around Chikafa, Hunyani and Mushumbi Pools areas. The study concludes that Mbire District is vulnerable to floods hence the need for a flood protection measure framework that provides practical and feasible solutions, basic constructional guidelines for the protection of settlements and agricultural lands as well as a near real-time monitoring framework provided by this study.