aashishpanta0/prithvi-wxc-downscaling
Reproduce NASA PrithviWxC / granite-wxc weather downscaling with portable skills and an MCP tool server.
Analyze PrithviWxC / granite-wxc NetCDF outputs (MERRA-2, NARR, PRISM ppt/tmax/tmin) for a date, region, difference map, trend, or climatology. Use when the user asks to plot, summarize, compare dates, or compute monthly/seasonal climate stats from downscaled or training files.
Use when the user asks to downscale gridded weather/climate data (MERRA-2, NARR, ERA5, CORDEX, ACCESS-CM2, or any gridded NetCDF) with this repo's PrithviWxC / granite-wxc pipeline. Covers YAML config authoring, preprocessing, scalar computation, training, and inference through the MCP tool server. Trigger on "downscale", "super-resolve", "PrithviWxC", "granite-wxc", "NARR downscaling", "ERA5 downscaling", "MERRA-2 downscaling", "run inference", "train downscaler".
Add stochastic residual refinement to NARR Prithvi-UNet downscaling — a diffusion or flow-matching model of the residual y_true - y_hat trained on top of the frozen deterministic Prithvi model, giving a bias-corrected ensemble (mean and spread) instead of one field. Use when the user asks for diffusion, flow matching, generative or stochastic refinement, residual correction, ensembles, or uncertainty for NARR downscaling. NARR only; MERRA-2 has no refinement.