Quantifying uncertainty in European climate projections using combined performance-independence weighting

Abstract

Uncertainty inmodel projections of future climate change arises due to internal variability,multiple possible emission scenarios, and differentmodel responses to anthropogenic forcing. To robustly quantify uncertainty inmulti-model ensembles, inter-dependencies betweenmodels as well as a models ability to reproduce observations should be considered. Here, amodel weighting approach, which accounts for both independence and performance, is applied to European temperature and precipitation projections from theCMIP5 archive. Two future periods representingmidand endof-century conditions driven by the high-emission scenario RCP8.5 are investigated. To inform theweighting, six diagnostics based on three observational estimates are used to also account for uncertainty in the observational record.Ourfindings show thatweighting the ensemble can reduce the interquartile spread bymore than 20% in some regions, increasing the reliability of projected changes. Themean temperature change ismost notably impactedby theweighting in theMediterranean,where it is found tobe 0.35 °Chigher than the unweightedmean in the end-of-century period. For precipitation the largest differences are found forNorthernEurope,with a relative decrease in precipitationof 2.4% and3.4% for the two future periods compared to the unweighted case. Based on a perfectmodel test, it is found thatweighting the ensemble leads to an increase in the investigated skill score for temperature and precipitationwhileminimizing the probability of overfitting.