A bias-corrected & downscaled massive ensemble to diagnose uncertainty in climate impact projections

Abstract:

Projections of climate change and climate impacts requires bias-corrected, downscaled output from ensembles of earth system models (ESMs). Potential impacts are uncertain due to modeling differences between ESMs, internal variability stemming from the chaos of the earth system, and differences in the historical reference datasets used to bias-corrected and downscale ESM output. Here, we introduce the Bias-Corrected and Downscaled Massive Ensemble (BCD-ME), a set of over 1,400 projections of daily mean and maximum temperature. The BCD-ME samples model and internal uncertainty with up to 86 runs from 12 Large Ensembles and uncertainty in the reference dataset by using 4 different reanalysis products to bias-correct and downscale output. Output is organized by Global Warming Levels (GWL), accounting for differences between forcing scenarios and ESM climate sensitivities. The ensemble contains 20-year daily time series for each GWL on a uniform 1-degree grid, bias-corrected using Quantile Delta Mapping, and statistics of 20-year time series for each GWL on a uniform 0.25-degree grid, downscaled using Quantile-Preserving Localized Analog Downscaling. The BCD-ME is stored in Analysis-Ready, Cloud-Optimized (ARCO) Zarr stores, enabling efficient computation of the 97 TB (uncompressed) dataset.

Draft: "A bias-corrected & downscaled massive ensemble to diagnose uncertainty in climate impact projections" (25 February 2026);  preprint available via EarthArXiv, submitted for publication in Nature Scientific Data.