Protected agriculture mapping at continental scale for highlighting hotspots of altered hydrological processes

The presence of agricultural catchments used for protected agriculture is growing worldwide. Monitoring studies document regional-scale and long-term impacts of widespread greenhouse horticulture production on the hydrological cycle, agricultural chemicals fate, and plastic pollution. Still, methodological tools that allow for continuous and continental mapping of greenhouses are lacking. In this study, we present one such tool developed upon integrating an improved version of the random forest-based Open field and Protected Agriculture land cover Classifier (OPAC), of Sentinel-2 L2A data, with the computational power and storage capability of the Google Earth Engine (GEE) platform. The GEE platform allowed for (1) exploiting a recently deployed state-of-the-art cloud mask, (2) performing a time series analysis-based image aggregation step to generate an optimal scene to classify, while preserving the coherent information of each pixel in the collection and (3) integrating user-defined masks. The tool was applied at the pan-European level constrained on agricultural areas mapped in the CORINE Land Cover 2018 product. Against manually selected ground-truth pixels, OPAC achieved an overall accuracy of 92 % for the class of protected agriculture. The study reveals that in the summer period of 2019, protected agriculture covered 5300 km2 of the pan-European agricultural area. The largest extensions of protected agriculture were in Turkey, Spain and Italy. The output land cover map can reveal locations where protected agriculture is altering the hydrological cycle and agricultural contaminants' fate in the environment as well as it can reveal potential sources of plastic pollution. The exploitation of the output map can thus expand the capability of modeling tools used for evaluating scenario analyses aiming to achieve a sustainable management of agricultural catchments.