Observing with Sentinel-1 widespread flood crises of 2022 in Pakistan and Nigeria

Flood events are a major threat to human lives and are often responsible for a substantial destruction of infrastructure. Unfortunately, the obstruction of transport links often prevents the accessibility of certain regions and the impact cannot be estimated, especially during large scale flood events. In such crises, earth observation data provide the most valuable information. Due to their cloud-independent observations, microwave satellites are well-suited for observing the flood extent in these situations. In 2022, millions of people were affected when large flood events hit Pakistan and Nigeria. Both events were covered by images taken by the European Synthetic Aperture Radar (SAR) satellite Sentinel-1, whereby the event in Pakistan was captured more frequently compared to the one in Nigeria.

The Global Flood Monitoring (GFM) component of the Copernicus Emergency Management Service (CEMS) utilises Sentinel-1 to automatically map floods on a global scale. The service relies on three independent flood mapping algorithms combined to an ensemble solution, and one of them was developed by the Technische Universität Wien (TU Wien). The algorithm (Bauer-Marschallinger et al., 2022) performs a pixel-wise Bayesian decision between flood and no-flood situation. For this, a local no-flood backscatter signature is provided based on a time-series-based harmonic model. The flood backscatter signature is defined by a linear model for water surfaces. Thanks to this setup, the algorithm provides its results without the need for any manual intervention and allows fast and lightweight computation.

This contribution analyses results of the TU Wien algorithm for the two large scale events in Pakistan and Nigeria, and will include the presentation of the affected areas, as well as the temporal progression of the flood crises. The performance evaluation of events of such magnitude generally lacks comprehensive ground-truth data and is commonly performed based on other satellite-derived data. Expanding the scope of a previous study of the Pakistan flood (Roth et al., 2022), we compare the results to other datasets being retrieved from multi-temporal data and cover the larger area of the event. The required reference data were received from a local and global flood mapping service, namely Sentinel Asia and the United Nations, respectively. Finally, the varying Sentinel-1 coverage density in respect to flood progression will be discussed to obtain insights into the impact of the satellite overpass frequency on the flood mapping quality.

 

Bauer-Marschallinger, B., Cao, S., Tupas, M. E., Roth, F., Navacchi, C., Melzer, T., ... & Wagner, W.: Satellite-Based Flood Mapping through Bayesian Inference from a Sentinel-1 SAR Datacube, Remote Sensing, 14(15), 3673, 2022.

Roth, F., Bauer-Marschallinger, B., Tupas, M. E., Reimer, C., Salamon, P., and Wagner, W.: Sentinel-1 based analysis of the Pakistan Flood in 2022, EGUsphere [preprint], https://doi.org/10.5194/egusphere-2022-1061, 2022.