This presentation reports on a Le Studium Data Analyst fellowship (March–October 2026) hosted by the JUNON project at La Technopole d’Orléans. JUNON is a joint initiative of BRGM, INRAE, CNRS and the Universities of Orléans and Tours. The project develops algorithms that automate the extraction, interpretation and analysis of geospatial and environmental datasets describing the natural resources of the Loire Valley (Centre-Val de Loire, France). The system is a virtual model of the environment that stays in sync with real-world observations through continuous data streams. Its architecture is modular and layered, and it organises data as a hierarchy of entities, sub-assets (soil, water, climate), static properties and live streams. Python is the main tool. It connects heterogeneous sources through REST APIs and ETL/ELT pipelines: Hub’Eau/ADES piezometric levels, ERA5 and Météo-France meteorological data, SAFRAN effective rainfall and evapotranspiration, Lig’Air air-quality records, and satellite imagery. The implementation uses the open-source FIWARE platform (Orion Context Broker, NGSI-LD), with MongoDB storage, Docker containers and Wirecloud dashboards. A proof of concept on water quantity covers four scenarios: monitoring, 2D/3D visualisation, prediction of groundwater levels and availability for irrigation, and detection of drought-vulnerable areas. A web application provides interactive visualisation of historical data and multi-step forecasting of groundwater levels. The forecasts use NARX, Random Forest, autoregressive and XGBoost models with exogenous climate inputs (temperature, rainfall, potential evapotranspiration), and model performance is evaluated with MAE, MSE and RMSE.