Data used in the extended abstract "Towards a distributional autoencoder for climate counterfactuals" submitted to the Climate Informatics 2026 conference. CESM2 LE Data The data is derived from the Community Earth System Model 2 Large Ensemble (CESM2 LE), which can be accessed through https://gdex.ucar.edu/datasets/d651056/ (Danabasoglu, G., Deser, C., Rodgers, K., and Timmermann, A.: CESM2 Large Ensemble, NSF National Center for Atmospheric Research [data set], https://doi.org/10.26024/kgmp-c556, 2020b.) europe_10percent_masked_stacked_TREFHT_JJA.nc pcsts_and_fGMT_not_standardized_LE.nc Individual CESM2 Simulations Freely running simulations This data is derived from freely running CESM2 simulations that are accessible through: https://doi.org/10.5281/zenodo.18172330 ETH_europe_10percent_masked_stacked_TREFHT_JJA.nc pcsts_and_fGMT_not_standardized_ETH.nc Nudged circulation simulations This data is derived from circulation nudged CESM2 simulations that are accessible through: https://doi.org/10.5281/zenodo.18172330 ETH_europe_10percent_masked_stacked_TREFHT_nudged_JJA.nc