Geospatial analysis of flooding events using Sentinel-1 and Sentinel-2 data: a tale of two South African cities

The frequency and severity of floods have increased in the last five decades due to climate change and human activities, significantly impacting human lives, economies, and infrastructure. South Africa is among the most affected regions, primarily due to informal settlements, limited resources, and a weak capacity to respond to the growing flood risk, with annual impacts increasing. Earth Observation data offers crucial insights for flood monitoring and risk management, yet studies and proactive measures remain limited in the country. Therefore, this paper conducted a geospatial analysis of recent flooding incidents (i.e., 2017-2022) in two South African cities using Sentinel-1 and Sentinel-2 datasets within a cloud computing environment. Specifically, we evaluated the potential of Sentinel-1 Radar and Sentinel-2 multispectral data for flood mapping using thresholding techniques and estimated the number of people affected by incorporating statistical and building-count data. The results showed that Sentinel-2 misclassified many areas due to confusion with clouds shadows. In contrast, Sentinel-1 showed greater potential for rapidly mapping floods near the incident date and estimating the number of people exposed, making it suitable for rapid flood assessments. Consequently, flooded areas derived from Sentinel-1 imagery were more realistic, indicating that about 60,000 people were cumulatively affected by flooding in eThekwini in April 2019 and October 2017, respectively. Comparatively, relatively few people (i.e., ~ 42,068 in March 2018 and 39,903 in February 2020) were affected by the various flood incidents in Johannesburg. Overall, the study has the potential to provide pertinent information on flooded areas and to aid follow-up analysis, such as infrastructure damage assessment, thereby offering prospects for informing not only disaster management and policy formulation but also critical decisions and resource allocation.