I need a robust, end-to-end workflow that uses satellite‐derived variables and machine-learning models to quantify both groundwater stress and groundwater recharge across my study area. The project covers the full pipeline—from sourcing and cleaning raw remote-sensing data (GRACE, SMAP, Landsat, GLDAS or similar), through feature engineering and model selection, to generating final geospatial layers and concise technical documentation. Key goals • Produce spatially explicit maps and time-series of groundwater stress indices as well as annual/seasonal recharge rates. • Explain the drivers: show which satellite-based predictors (e.g., precipitation, evapotranspiration, land-cover change) most influence the model outputs. • Deliver reproducible Python or R notebooks, trained model files, and clear metadata so results can be updated when new satellite scenes become available. Acceptance criteria 1. Model accuracy meets agreed benchmarks (R², RMSE or classification F1 as appropriate). 2. All code runs end-to-end on a fresh machine with only the listed dependencies. 3. Final raster layers align correctly with standard geographic projections and pass spot-checks against in-situ well data supplied later in the project. When you reply, focus on your experience with satellite hydrology, geospatial machine learning, and any prior work that combined stress-recharge assessments. Briefly outline the toolchain you prefer (e.g., Google Earth Engine, xarray, TensorFlow, scikit-learn) and how quickly you can deliver the first milestone of cleaned input data and an initial baseline model.