Constructing DAMIP-based detrended reanalyses for event attribution: design of the ATTRICI-DAMIP dataset

Event attribution studies increasingly rely on single–model counterfactual simulations (e.g. from the Detection and Attribution Model Intercomparison Project; DAMIP), or on statistical detrending to separate anthropogenic from natural variability (e.g. from the ATTRIbuting Climate Impacts framework; ATTRICI). However, there is currently no standardized observatio–constrained dataset that combines reanalysis with multi‑model forced trends from such existing intercomparison projects.

Here we propose, for the first time, to develop an ATTRICI–DAMIP dataset at a global scale, consisting of detrended reanalyses that approximate a counterfactual climate without anthropogenic forcing. Our main goals are to (i) derive anthropogenic ("hist–ant") trends from DAMIP historical (all forcings) and hist–nat (only natural forcings) experiments, (ii) apply these trends to remove the anthropogenic signal from multiple reanalysis products using the ATTRICI framework, and (iii) assess the suitability of the resulting fields for specific climate attribution applications.

We compute multi–model anthropogenic trends for key variables, including daily mean, maximum and minimum temperature, and precipitation, from DAMIP ensembles. Trends are estimated over 1950–2014, primarily using pattern scaling, and expressed on a common 0.5° spatial grid. Corresponding hist–ant trends are then subtracted from reanalysis fields to construct detrended pseudo–counterfactual time series. The resulting ATTRICI–DAMIP products are evaluated by comparing (a) the amplitude and phasing of variability against reanalyses, (b) trends and variability against hist–nat simulations, and (c) attribution metrics such as changes in distribution tails and Fraction of Attributable Risk for selected regional case studies.

We anticipate that the ATTRICI–DAMIP dataset should retain the realistic day–to–day variability and synoptic structures of reanalyses, while substantially reducing long–term anthropogenic trends in temperature and related variables. Our preliminary analyses using the ERA5 reanalysis and four process–based models (IPSL–CM6A–LR, MPI–ESM1–2–HR, GFDL–ESM4 and MRI–ESM2–0) indicate strong consistency between model–derived anthropogenic warming patterns and reanalysis trends over Europe. Here we expect to demonstrate how this new global–scale dataset can be used to quantify the anthropogenic contribution to recent high–impact events, particularly in under–studied regions such as West Africa and South-East Asia.

This work aims to provide a transparent, reproducible framework to merge DAMIP–based forced responses with reanalysis using the ATTRICI protocol, producing a new class of counterfactual datasets for climate attribution studies and supporting operational attribution and impact modelling.