Small agricultural reservoirs detection with satellite data and OpenStreetMap integration for sustainable water management: a contribution to the CASTLE project

Meteorological unpredictability, exacerbated by severe events caused by climate change, poses significant problems for water resource management (IPCC, 2023). Climate change has increased the frequency and severity of droughts, especially in mid-latitude regions, where reduced precipitation coupled with rising temperatures is expected to exacerbate water scarcity (https://doi.org/10.1007/s40641-018-0093-2). In this regard, Small Agricultural Reservoirs (SmARs) offer a strategic response, as they are designed to collect and store water for use in irrigation and other agricultural applications. This is the context in which the research activity described here is developed, contributing to the research project CASTLE - Creating Agricultural reSilience Through smaLl rEservoirs. Despite their importance, the lack of comprehensive national databases for SmARs remains a major obstacle to their efficient management. Prior to this study, for example only eight of Italy's twenty regions had SmARs inventories, often based on non-standardised and incomparable approaches (https://indicatoriambientali.isprambiente.it/it/pericolosita-sismica/invasi-artificiali). This fragmentation of information makes the analysis and management of SmARs challenging. A possible option to overcome this problem is represented by satellite data, which provides accurate and continuous information over large geographical areas. Sentinel-2 satellite imagery - part of the European Space Agency's Copernicus programme - was particularly well suited to this study. The objective of this research was to develop a methodology for detecting Small Agricultural Reservoirs from satellite imagery with integration of OpenStreetMap (OSM) and the ESA World Cover 2021 dataset and creating a comprehensive inventory of the existing reservoirs in Italy. The system was validated in Tuscany with the use of the ground truth database of LaMMA - CNR IBIMET (https://geoportale.lamma.rete.toscana.it/difesa_suolo/#/viewer/372). Integration with OSM helped eliminate false positives such as ponds, glaciers, large dams, rivers, and canals, which spectral indices alone cannot distinguish from SmARs due to their similar reflectance characteristics, as they are also water surfaces. The ESA World Cover data were used to exclude urbanized areas, which were irrelevant to this study. The combined use of open-source data has enabled the development of a replicable methodology adaptable to various spatial scales, considerably enhancing the identification and mapping of SmARs. This strategy will help to manage agricultural water resources more efficiently and increase resilience to climate change.