Supplementary Data: Physics-constrained generative machine learning-based high-resolution downscaling of Greenland's surface mass balance and surface temperature

Supplementary data to Physics-constrained generative machine learning-based high-resolution downscaling of Greenland’s surface mass balance and surface temperature. We provide the training, validation and test data for the model (in one file). For convenience, we also provide the test data in a seperate file and the artificially coarsened and linearly interpolated MAR test data. The model checkpoints are available as well as the auxiliary data needed to run the code (see Github).  The reference data for the quantile data mapping (QDM, historical MAR data driven by ERA5) is also provided. We provide one downscaled realisation (constrained and not constrained) for the MAR test data (e.g. Fig.1). We also provide example data to be able to try the downscaling via the script in the GitHub repository. Namely, we provide the QDM-corrected historical SMB and TS from a CESM2 run from CMIP6 from the years 1950-2014 (example_data). This data is not shown in the corresponding paper though.