Supporting Disaster Management: Using Satellite Remote Sensing to Assess Changes in Access to Emergency Healthcare Facilities due to Flooding

Kurzfassung

Flooding is one of the most frequent natural disasters worldwide, destroying infrastructure and causing negative health effects. Especially for immediate health consequences such as physical injuries, fast medical intervention is vital. To allocate resources, information on restrained access due to flooding is crucial for first responders. While rapid mapping via satellite remote sensing is a common tool for disaster management, many services like the International Charter “Space and Major Disasters” and the Copernicus Emergency Management Service rely on activation through authorized users, which causes restrictions and introduces delays. The German Aerospace Center (DLR) developed the Surface Water Inventory and Monitoring (SWIM) product that continuously monitors water extents on a global scale through an automated approach. Since SWIM’s application is still under development this study aims to assess its usage in emergency management for accessibility estimates. In addition, to account for common detection challenges of satellite remote sensing, e.g. flood detection in urban areas, the flood extents are propagated based on the local terrain using the FLEXTH algorithm. This method also allows to consider water depth (WD), further advancing the automated mapping process. Finally, conclusions are drawn how the flood products impact the outcomes of accessibility estimates. The findings indicate that SWIM’s performance in flood detection still has the potential to be improved in both study areas (Emilia-Romagna, Italy and Borno, Nigeria) as the detection of impassable roads (WD greater than 30 cm) showed (IoU: 16.70 % Italy, 24.43 % Nigeria). Enhancing the flood extent with FLEXTH improved the precision of impassable road detection in Italy as false positive (FP) values were deleted, while improvements using FLEXTH were marginal in Nigeria. In Emilia-Romagna region, Italy the number of citizens with no timely access to emergency healthcare was overestimated (reference data 18.80 %, SWIM 35.79 %, mSWIM 32.95 %), whereas in Borno, Nigeria the usage of rapid mapping flood extents led to an underestimation of disruption (reference data 77.66 %, SWIM 40.58 %, mSWIM 42.05 %). The overestimation in Italy can be due to misclassifications stemming from mixed backscatter because of the spatial resolution of the Sentinel-1 sensor. In Borno, urban areas present a detection challenge causing the identification of a lower number of flooded roads with Sentinel-2. The findings can be used as recommendations to guide further research for automated rapid mapping approaches. Further studies could focus on assessing the impacts of higher resolution data, i.e. DEMs on FLEXTH’ performance. Besides, comparison of results with complimentary data sources such as crowd sourced data from social media or citizen science could be beneficial to balance misclassifications and limitations arising from satellite imagery.