Detecting olive grove abandonment with Sentinel-2 and machine learning: The development of a web-based tool for land management

The abandonment of rural areas is an important environmental and socio-economic issue in Europe, threatening the stability profitability agricultural production. identification quantification abandoned land key for temporal spatial monitoring process applying alternative management measures. Italy one most European countries production high quality olive oil, accounting a large slice current certificated (i.e., PDO, PGI). In this study, we present machine learning model Random Forest) tree groves using field observations NDVI time series, tested typical agroecosystem central dominated by groves. An application smartphones able to record geographic position was developed used collect points, which turn were utilised train model. data from January December 2020, calculated on Sentinel-2 images, extracted each point gap-filled obtain 10-days interval series. Forest annual series as features classified sampling points test dataset with accuracy 0.85. showed higher capacity classifying cultivated than sensitivity being equal 0.88 specificity 0.82. Results demonstrated applicability combined approach discriminating groves, case that parcels destined cultivation are known. A web-based tool implemented support management.