Scenario-Testing and Prediction Capabilities of the SaveCrops4EU Agricultural Digital Twin Component

This abstract presents the scenario-testing and forecastingcapabilities of the SaveCrops4EU DTC Agriculture platform. Thescenario-based representation of crop conditions relies on numericalsimulations generated with eCLM, a land-surface model with advancedagronomic features designed to produce alternative and plausibletrajectories of crop states (e.g., LAI, biomass, phenology, yield)under varying weather and management practices. The proposed approachcouples the spatiotemporal coverage of EO data with the mechanisticknowledge embedded in crop-growth models to (i) constrain model statesusing observations and (ii) quantify uncertainty through ensemblemethods. These scenario-testing capabilities are demonstrated for thewell-instrumented ICOS site of Selhausen (winter wheat) in Germany,using Copernicus seasonal atmospheric forecasts during the 2018drought year.The in-season crop-yield forecasting capabilities of the DTC solutionbuild on a suite of machine- and deep-learning techniques trained onatmospheric conditions and land-surface features derived from multipleremote-sensing products available in Copernicus catalogues andgenerated in the Monitoring Pillar of the platform. The data-drivenapproaches are evaluated using wheat and maize yield statisticsaggregated at the NUTS3 level for Germany and Hungary. In addition,explainability tools are integrated to support users in interpretingmodel outputs. Finally, we discuss how the scenario-testing andforecasting components of SaveCrops4EU can be scaled across Europeanagricultural regions to enable farmers and policymakers to assessadaptation strategies under changing climate conditions.