Locating potential flood areas in an urban environment using remote sensing and GIS, case study Lund, Sweden

Abstract: Impervious surfaces, such as large parts of the urban area, reduce infiltration and increase above ground runoff (Chen et al. 2009) with risk of flooding as a result. Urban flooding has the potential to affect a large amount of people, and can cause great harm to economic interests. Therefore it is important to locate potential areas prone to flooding, and have storm water regimes in mind when planning expansions of built up areas (Barbosa et al. 2012). The focus of this paper was to develop and adapt methods to locate potential flood areas in built up areas using remote sensing and GIS analysis. The study area for this study was the city of Lund, located in southern Sweden. The methods included the use of high resolution (0.25 m) near-infrared ortho image and a high resolution (2 m) Digital Elevation Model (DEM). The ortho image was used in a land cover classification to locate and map impervious surfaces, while the DEM was used to map potential flow directions and catchment areas on the impervious surfaces. Precipitation data was used to simulate runoff volumes and was used together with the catchment areas to highlight those areas thought to be most prone to flooding. The resulting catchment areas Focus was initially put on the three catchment areas with the highest accumulated runoff volumes. These were areas with large homogenous surface cover (impervious surfaces) and continuous slope profile. In regard to accumulated runoff volume, these three areas were thought to be most prone to flooding. But in consideration to the method used to locate the areas, they could be said to simply represent the largest continuous surfaces and not necessarily the most flood prone areas. Therefore a runoff index (water level height) was introduced to filter out catchments based on runoff volume and catchment area. This procedure produced a different result with smaller catchments. The results based on the runoff index are thought to be a more accurate representation of the most flood prone areas, due the smaller size of the catchments that decrease the response time from precipitation to accumulation. Limitations of the model For the model to be more accurate, more information regarding drainage would be required as a parameter for infiltration. The land cover classification also introduced a certain amount of misclassification, e.g. in regards to bare soil and impervious surfaces (Lu and Weng 2004). To increase the accuracy, a more specific classification scheme could have been used for impervious surfaces (Han and Burian 2009). Another limiting factor is the flow direction algorithm; the use of a more advanced representation of flow direction could have produced a more realistic result (Zhou et al. 2011).