<p>Nowadays, the importance of coastal areas is greater than ever, with approximately 10% of the global population living in these areas. These zones are an intermediate space between sea and land and are exposed to a variety of natural (e.g. ground deformation, coastal erosion, flooding, tornados, sea level rise, etc.) and anthropogenic (e.g. excessive urbanisation) hazards. Therefore, their conservation and proper sustainable management is deemed crucial both for economic and environmental purposes. The main goal of the Greece-China bilateral research project &#8220;EPIPELAGIC: ExPert Integrated suPport systEm for coastaL mixed urbAn &#8211; industrial &#8211; critical infrastructure monitorinG usIng Combined technologies&#8221; is the design and deployment of an integrated Decision Support System (DSS) for hazard mitigation and resilience. The system exploits near-real time data from both satellite and in-situ sources to efficiently identify and produce alerts for important risks (e.g. coastal flooding, soil erosion, degradation, subsidence), as well as to monitor other important changes (e.g. urbanization, coastline). To this end, a robust methodology has been defined by fusing satellite data (Optical/multispectral, SAR, High Resolution imagery, DEMs etc.) and in situ real-time measurements (tide gauges, GPS/GNSS etc.). For the satellite data pre-processing chain, image composite/mosaic generation techniques will be implemented via Google Earth Engine (GEE) platform in order to access Sentinel 1, Sentinel 2, Landsat 5 and Landsat 8 imagery for the studied time period (1991-2021). These optical and SAR composites will be stored into the main database of the EPIPELAGIC server, after all necessary harmonization and correction techniques, along with other products that are not yet available in GEE (e.g. ERS or Sentinel-1 SLC products) and will have to be locally processed. A Machine Learning (ML) module, using data from this main database will be trained to extract additional high-level information (e.g. coastlines, surface water, urban areas, etc.). Both conventional (e.g. Otsu thresholding, Random Forest, Simple Non-Iterative Clustering (SNIC) algorithm, etc.) and deep learning approaches (e.g. U-NET convolutional networks) will be deployed to address problems such as surface water detection and land cover/use classification. Additionally, in-situ or auxiliary/cadastral datasets will be used as ground truth data. Finally, a Decision Support System (DSS), will be developed to periodically monitor the evolution of these measurements, detect significant changes that may indicate impending risks and hazards, and issue alarms along with suggestions for appropriate actions to mitigate the detected risks. Through the project, the extensive use of Explainable Artificial Intelligence (xAI) techniques will also be investigated in order to provide &#8220;explainable recommendations&#8221; that will significantly facilitate the users to choose the optimal mitigation approach. The proposed integrated monitoring solutions is currently under development and will be applied in two Areas of Interest, namely Thermaic Gulf in Thessaloniki, Greece, and the Yellow River Delta in China. They are expected to provide valuable knowledge, methodologies and modern techniques for exploring the relevant physical mechanisms and offer an innovative decision support tool. Additionally, all project related research activities will provide ongoing support to the local culture, society, economy and environment in both involved countries, Greece and China.</p>
Multi-Temporal Surface Water Classification for Four Major Rivers from the Peruvian Amazon
