New Publication | Environmental Research Letters
How much of a climate extreme is down to chance?
Hot days, warm spells and heavy rainfall are routinely quantified with a standard set of 26 climate extreme indices defined by the Expert Team on Climate Change Detection and Indices (ETCCDI), which underpin impact assessments from global maps down to single river catchments. Yet how often such extremes occur depends not only on the climate itself, but also on internal climate variability: the chaotic sequence of weather that could just as well have unfolded differently. Observations capture only one such sequence, and multi-model ensembles mix it with differences between models, so this uncertainty is rarely accounted for. Writing in Environmental Research Letters, Lukas Brunner and colleagues at Universität Hamburg, together with Benjamin Poschlod (TUHH, IGWS), who conceived the study, now map it systematically.
The team analysed 50 simulations of the MPI-ESM climate model that share the same external forcing and differ only in their initial conditions, so that every difference between them reflects internal variability. Even in 20-year averages over 1995–2014, the effect is substantial. Over Europe, about 15 % of days exceed the local 90th percentile of daily maximum temperature on average, but at a given location the lowest and highest of the 50 simulations can differ by more than 10 percentage points. Globally, internal variability amounts to roughly 10–20 % of the mean for this index and 20–40 % for the duration of warm spells. For the hottest day of each year, the spread is particularly large over the US Great Plains and from the Mediterranean to central Asia. For heavy rainfall, it scales with the local climate: over Germany, the annual rainfall on very wet days typically varies by about 20 mm, or 10 %, between simulations.
Which of these numbers matters depends on the question: documenting the observed past, estimating the expected climate, or preparing for a plausible worst case. Since no static map can serve all purposes, the authors provide the Variability Atlas in three layers. A browser-based app lets users select an index, a metric and a region without any coding. A Python repository with pre-calculated indices and notebooks supports tailored analyses, such as picking the simulations with the highest or lowest regional mean to build spatially consistent storylines. The full index time series from 1850 to 2014 are archived at the World Data Center for Climate. Given the coarse model resolution of about 150 km, the results are best read as a first estimate and a lower bound.
The central message: even when only the mean matters, the role of internal variability should be acknowledged.
Read the full open-access article: Brunner, L., Mithal, V., Borchert, L. F., Sillmann, J., and Poschlod, B. (2026): The variability atlas: how internal climate variability affects the estimation of climate extreme indices. Environmental Research Letters 21, 184022. https://doi.org/10.1088/1748-9326/ae9c57
Variability Atlas (interactive web app): https://lukasbrunner.github.io/variability_atlas
Code and data: https://github.com/lukasbrunner/etccdi_internal_variability (scripts and pre-calculated indices) and https://doi.org/10.26050/WDCC/ETCCDI-MPIGE (full index time series)
