Relevance of ESA World Cover Map in Flood Classification Using SAR Images

Flood detection has become a crucial issue in the current decade, prompting the development of numerous solutions. Satellite image-based flood detection has emerged as a popular approach. However, adverse weather conditions, particularly cloud cover, often compromise the quality of satellite imagery. This limitation underscores the significance of Synthetic Aperture Radar (SAR) images, which can penetrate through adverse weather conditions, including clouds. Moreover, the polarization of SAR images varies across different land types, further emphasizing the need for accurate classification methods. In this context, the European Space Agency (ESA) provides a valuable resource in the form of the world cover map for SAR images, enhancing the precision of flood classification efforts. In our study, only $0.6 \%$ of the dataset, consisting of 10 images, was utilized for training, with the remaining $99.4 \%$ (1,621 images) reserved for testing. The analysis reveals a significant improvement in flood classification accuracy, increasing from 0.799 without world cover map feature to 0.8343 with the inclusion of world cover map feature. This demonstrates the significance of ESA world cover maps in flood classification, underscoring its importance in satellite-based flood disaster management strategies.