The importance of internal variability for climate extreme indices

The occurrence of climate extremes is influenced by climate forcing as well as internal climate variability: internal variability may temporarily obscure or enhance the forced signal in climate extremes. The role of signal versus noise plays an important role, for instance in the analysis of emergence. The climate extreme indices from the Expert Team on Climate Change Detection and Indices (ETCCDI) are routinely used to assess the impacts of forced change on climate extremes, but in such analyses internal variability is often ignored. We present a comprehensive catalogue of the importance of internal variability for the 27 ETCCDI indices to inform climate extreme analysis and guide impact science.

In our assessment, we use a 50-member ensemble of the CMIP6 generation MPI-ESM 1.2 LR Earth System Model for 1961-2014 to highlight combinations of regions and indices that are strongly affected by internal variability. Unlike previous work, we consider all ETCCDI indices in the same model ensemble to provide a clean identification of internal variability. Using the coefficient of variation as initial metric, we find that the total signal is strongly affected by internal variability  

• over ocean regions for temperature indices based on percentile thresholds (e.g. tx90p), 

• along quasi-zonal mid-latitude bands for absolute maximum/minimum temperature indices (e.g. txx), and 

• in characteristic (sub-)tropical “hot-spot” regions such as northern Africa, the eastern central Pacific, and the south-east of all ocean basins for precipitation-based indices (e.g. r95p). 

This grouping illustrates the differing relative importance of internal variability for the extreme signal depending on the index and the region, and sheds light on processes that contribute to the occurrence of climate extremes. Further, the catalogue provides a tangible resource that enables users of ETCCDI indices to better understand the robustness of index information they might derive from single model runs or observations. Based on our catalogue, users, e.g. impact scientists, may select suitable indices specific to their region of interest and application.