Flood inundation mapping in data-scarce areas: A case of Mbire District, Zimbabwe

Abstract

Floods are one of the most devastating weather-related hazards that are affecting millions of people over the world every year. In some poor resource areas such as Mbire District in Zimbabwe, the floods are difficult to anticipate and prepare for. Hence the need for spatial modelling of the past flood events for effective response and management. This study modelled the flood extent and depth based on data from household surveys, transect walks and a digital elevation model (DEM). A sample of 304 households was used, with 70% for calibration and 30% for validation of the flood extent. Twenty-four flood depth measurements obtained from transect walks were used to validate the modelled flood depths based on a linear regression model. The flood depth of the worst most recent flood (January 2015) at each household was combined with altitude from the DEM using the sum function, and the inverse distance weighting was applied to model the worst flood depth. The flood extent was considered as those areas where flood depth was higher than the DEM. Approximately 24% of the area was covered by floods. The modelled flood extent agreed reasonably well with what was reported during the survey (probability of detection 0.93 and accuracy level about 0.8). Most of the areas in the wards experienced flood depths greater than 2 m, especially along the major rivers. Such areas are dangerous for people, animals and properties such as boreholes, houses, schools and clinics located on the floodplain. These results can be used for planning purposes in preparing and responding to stages of the flood management cycle. However, there is a need for further research to improve the performance and applicability of the methodology applied in this study in other settings.