Sensitivity of flood dynamics to different soil information sources in urbanized areas

Abstract

This study focused on the sensitivity of flood dynamics to soil hydraulic properties derived from three different soil databases in urbanized area: (1) upscaled locally observed soil texture data based on the soil-landscape relationships (SMLS); (2) the SoilGrids250m open data source (SMSG); and (3) the FAO soil map (SMFAO). However, soil in the urban areas has two major conditions related to land cover and soil physical structure for which the information is not available in soil databases: the effect of fragmented vegetation cover and the effect of compaction of bare soil. Non-built-up areas can be covered by fragmented vegetation (grass and shrubs) which generally has high infiltration rates. In contrast, bare areas such as dirt roads and footpath can be heavily compacted and have typically low infiltration rates. We used Pedotransfer functions (PTFs) with satellite-derived vegetation cover and bare soil to predict soil hydraulic properties related to the uncompacted and compacted scenarios. Infiltration dynamics was derived from the predicted soil hydraulic properties for these soil information sources, which then determined the flood dynamics in the catchment. The flash flood modeling was done using the integrated flood modeling system using openLISEM (Bout and Jetten, 2018) for the whole of Kampala (Uganda) using the 25th of June 2012 flood event. In the distributed openLISEM hydrological model, these two urban soil conditions have been treated separately. We have evaluated the sensitivity of flood dynamics to three different soil databases under both uncompacted and compacted urban soil conditions by using different flood indicators such as catchment water balance, infiltration rate, flood depth and duration, flooded area and flood volume, and the average number of structures affected. The result of the study indicates that soil hydraulic properties needed for the distributed hydrological model are better predicted when using the SMSG and SMLS, which resulted in better infiltration simulation. Consequently, compared to an earlier simulation that was verified with stakeholders and accepted for drainage system design, the accuracy of the simulated flood extent map was better when using SMSG and SMLS. Moreover, soil compaction significantly reduces infiltration and consequently increases the flood depth and duration, and therefore, must be included in the urban flash flood modeling study.