Assessing and improving the simulation of runoff and design flood estimation in urban areas using the ACRU and SCS-SA models.

Abstract

Urbanisation is increasing at a rapid rate. Pervious and vegetated land is increasingly being replaced by impermeable surfaces (roads, pavements, driveways, parking lots, etc.) resulting in large portions of total imperviousness in catchments. The expansion of urban areas alters the natural underlying surface condition affecting catchment characteristics. The most common impacts of urbanisation on the hydrology of a catchment are increased runoff volumes, reduced baseflows owing to less infiltration taking place and a decrease in catchment response time. These changes can result in increased flood risk and subsequent damage to urban infrastructure and affect livelihoods. Therefore, accurate modelling of runoff and estimation of design floods of highly urbanised areas is necessary, especially in the often neglected catchments with informal settlements and infrastructure and in peri-urban catchments. Peri urban areas are defined as those areas located adjacent to a city area and have a mix of both rural and urban characteristics. Two rainfall-runoff models, namely the ACRU and the Visual SCS-SA model, were selected for application on catchments with typical South African urban conditions. The models have been developed and tested in urban catchments, however not extensively. The study areas are located in the South African urbanised cities of Tshwane and Pietermaritzburg. ACRU is a daily time step conceptual and physically-based agro-hydrological model that is relatively more data intensive compared to the simpler SCS-SA model. Therefore, information systems such as Remote Sensing (RS) and Geographic Information System (GIS) have been explored to aid as data sources and tools for acquiring model input parameters, at a more accurate level. The ACRU default values by Tarboton and Schulze (1992) and impervious area estimations derived by Loots (2020) were initially used to estimate the ACRU impervious parameters. Additionally, the pixel-based land cover classification method using satellite images was carried out in detail for this study as an attempt to map impervious surfaces and obtain impervious ACRU parameters with improved accuracy. Impervious land use classes were also extracted from the 2018 South African National Land Cover Database (SANLC), 2018 Global Man-made Impervious Surface (GMIS) and the 2010 Global Artificial Impervious Areas (GAIA). In order to use the ACRU and SCS-SA models confidently, the simulated results need to be verified against reliable observed data for each impervious scenario, if observed data is available. QGIS was used to obtain and process data into information required for the selected models. Several model input data such as slope, elevation, and catchment rainfall were estimated through GIS. The models over simulated observed design floods for the urbancatchments. Obtaining reliable observed data (rainfall and runoff), and satellite images with good resolution proved to be a consistent challenge throughout the study and could have contributed to the poor performance of the models. Urban area data dating back to the1990s was extracted from the GAIA method for most of the simulation period and a trend in impervious area expansion linked to urbanisation was detected and analysed against simulated streamflow from the urban catchments.