How do you downscale CMIP6 GCM data?
Answer: Pair quality-controlled local observations with an overlapping historical CMIP6 simulation, then load a future SSP run from the same GCM and ensemble member. Harmonize the variable, units, calendar, and time step; calibrate and validate a downscaling method on historical data; then apply the fitted method to the future run. Save the corrected series together with its model, scenario, member, calibration period, and method metadata.
Recommended workflow
- Choose a CMIP6 variable, model, ensemble member, historical run, and SSP run.
- Prepare observed station data and confirm that the historical GCM period overlaps it.
- Extract the relevant GCM grid cell or cells and convert model values to the observation units.
- Calibrate Delta, QM, or EQM and evaluate performance on historical data not used for fitting when the record permits.
- Apply the selected correction consistently to the future SSP data and document the complete provenance.
Quality checks
Check dates, missing values, calendars, wet-day treatment, seasonal behavior, extremes, and whether the corrected historical series improves relevant evaluation metrics. Results should normally be compared across multiple GCMs or scenarios rather than relying on one projection.
Follow this CMIP6 downscaling workflow in SD-GCM with Delta, Quantile Mapping, or EQM.