Assess land degradation status based on Earth Observation driven proxy indicator

Land degradation is a global topic in climate change debates resulted from different types of human activities as well as from physical processes. Resilient, healthy soils are important to help reduce the ecological and economic impact of environmental change and extreme conditions. The development of adequate and broadly applicable indicators and thresholds is challenged by the great diversity of European soils and climate, as well as different political, economic, and social conditions which lead to different priority settings for targets and indicators. This work built upon current environmental awareness (e.g CAP, SDGs, etc.) to design a methodological framework for environmental performance metrics related to land degradation. The framework was oriented towards a data-driven environmental metric approach leveraging Copernicus Sentinel-2, existing open-access databases such as LUCAS (Land Use/Cover Area frame statistical Survey) and GEOSS (e.g., Soil Grids) vast dataset archives to provide metrics for environmental actors. Based both on the international literature and European commission documentations this work is focused on the combination of vital importance indicators of soil degradation and soil health. A novel deep learning architecture was implemented to support the final knowledge extraction with a pixel-based spatial resolution of 10m for the determination of Soil Organic Carbon (SOC) and Clay content. The above indicators are used as enhanced geospatial inputs to a soil erosion modelling approach providing improved predictions. A proper approach was followed for the SOC:clay ratio generation and with the soil erosion product combination to provide an ambitious land degradation index. An agricultural area in Northern Greece was used as a demonstration test site area for the proposed methodology.