Input dataset for ATDT (Attribution tool for Daily-to-monthly Temperatures)

Description Input data related to the manuscript "ATDT v1.0 - Attribution Tool for Daily-to-monthly Temperatures and its application to record-breaking Northern European heatwave of July 2025". Submitted to Geoscientific Model Development in June 2026. Directory structure Unpack the file "input_data.tar.gz" with the command: "tar -xzvf input_data.tar.gz". This gives you a directory called "input_data" which contains the files that are needed to run the code "estimate_distributions.py" in the "https://github.com/aetoropa/temp-attribution-1to31days" repository. The directory "input_data" contains three subdirectories named as "regression_coefficients", "GMST_time_series" and "qr_files". Regression coefficients The directory "regression_coefficients" contains the model simulated estimates for the changes in local climate's mean and variance in response to global mean temperature change. These estimates were calculated separately for each CMIP6-model and averaged to get the multi-model mean estimate. The single-model coefficients are stored in the "single_models" subdirectory while the multi-model mean coefficients are stored in the subdirectory "model_mean". The coefficients were calculated for daily mean, maximum and minimum temperatures (tas, tasmax, tasmin) and their 3, 7, 15 and 31-day moving averages using 5 different emission scenarios (ssp119, ssp126, ssp245, ssp370 and ssp585). Time-series of global mean temperature The directory "GMST_time_series" contains both the observed (HadCRUT5) and simulated (CMIP6) time-series of global mean surface temperature (GMST). The observed time-series is in the subdirectory "observed" while the simulated time-series of single CMIP6-models are in the subdirectory "single_models" and the multi-model mean time-series are in the subdirectory "model_mean". Quantile-regression files In order to convert the samples of pseudo-observations, representing pre-industrial, present-day and future climates, to continuous probability distributions, we perform quantile regression (QR) to them. QR is performed both for multi-model mean and single-model pseudo-observations. In addition, QR can also be performed for pure observations. After QR has been performed the corresponding results are saved to NetCDF-files. This is done so that QR doesn't have to be performed every time when the code is run for a given station and a time-period, as its computational cost can be high. The best-estimate quantiles obtained from performing QR to the samples of multi-model mean pseudo-observations are stored in the "model_mean" subdirectory while the quantiles of individual models corresponding to single-model pseudo-observations are stored in the "single_models" subdirectory. Quantiles obtained from performing QR to pure observations are stored in the subdirectory "observations". Correpondence More information can be asked fromAntti ToropainenDoctoral Researcher, University of [e-mail]