In recent years, the integration of deep learning techniques with remote sensing technology has revolutionized the way natural hazards, such as floods, are monitored and managed. However, existing methods for flood segmentation using remote sensing data often overlook the utility of correlative features among multispectral satellite information. In this study, we introduce a progressive cross-attention network (ProCANet), a deep learning model that progressively applies both self- and cross-attention mechanisms to multispectral features, generating optimal feature combinations for flood segmentation. The proposed model was compared with state-of-the-art approaches using the Sen1Floods11 dataset and our bespoke flood data generated for the Citarum River basin, Indonesia. Our model demonstrated superior performance with the highest intersection over union (IoU) score of 0.815. Our results in this study, coupled with the ablation assessment comparing scenarios with and without attention across various modalities, open a promising path for enhancing the accuracy of flood analysis using remote sensing technology.

Monitoring vegetation health and its response to climate conditions is critical for assessing the impact of climate change on urban environments. While many studies simulate and map the health of vegetation, there seems to be a lack of high-resolution, low-scale data and easy-to-use tools for managers in the municipal administration that they can make use of for decision-making. Data related to climate and vegetation indicators, such as those provided by the C3S Copernicus Data Store (CDS), are mostly available with a coarse resolution but readily available as freely available and open data. This study aims to develop a systematic approach and workflow to provide a simple tool for monitoring vegetation changes and health. We built a toolbox to streamline the geoprocessing workflow. The data derived from CDS included bioclimate indicators such as the annual moisture index and the minimum temperature of the coldest month (BIO06). The biophysical parameters used are leaf area index (LAI) and fraction of absorbed photosynthetically active radiation (FAPAR). We used a linear regression model to derive equations for downscaled biophysical parameters, applying vegetation indices derived from Sentinel-2, to identify the vegetation health status. We also downscaled the bioclimatic indicators using the digital elevation model (DEM) and Landsat surface temperature derived from Landsat 8 through Bayesian kriging regression. The downscaled indicators serve as a critical input for forest-based classification regression to model climate envelopes to address suitable climate conditions for vegetation growth. The results derived contribute to the overall development of a workflow and tool for and within the CoKLIMAx project to gain and deliver new insights that capture vegetation health by explicitly using data from the CDS with a focus on the City of Constance at Lake Constance in southern Germany. The results shall help gain new insights and improve urban resilient, climate-adaptive planning by providing an intuitive tool for monitoring vegetation health and its response to climate conditions.