Ensembles of regional climate models are widely used to obtain climate change signals and to evaluate associated uncertainties over a specific region. These models are forced by general circulation models (GCMs) at their lateral boundaries. The most recent multi-model ensembles projects – ENSEMBLES and CORDEX – rely on a limited set of about 10 GCMs per greenhouse gas emission scenario. It was shown previously that in particular, RCM temperature responses tend to cluster according to their driving GCM. Therefore, it is important to better understand the relation among the driving models. In multi-model ensembles as large as CMIP5, in which models tend to correlate due to their similar origin, model selection or weighting becomes an important issue. This study evaluates the distribution of climate change signals in the CMIP5 ensemble for temperature and precipitation over the Greater Alpine region and shows that different reasonable methods of model selection considerably influence the resulting temperature spread in the climate change signals at the end of the century relative to 1980–2009: excluding those GCMs with a poor representation of Alpine climate leads to a difference in spread of more than 1 °C compared to a selection strategy where all models are included and given the same weight. It is highlighted that the largest amount of spread can be retained with a weighting scheme based on a cluster analysis. Furthermore, we show in bivariate analyses that our understanding of the interplay between temperature and precipitation significantly depends on the model selection strategy. Hence, this work may have important implications for current and future design and analysis of multi-model projects such as CMIP5 and CORDEX.
CITATION STYLE
Zubler, E. M., Fischer, A. M., Fröb, F., & Liniger, M. A. (2016). Climate change signals of CMIP5 general circulation models over the Alps – impact of model selection. International Journal of Climatology, 36(8), 3088–3104. https://doi.org/10.1002/joc.4538
Mendeley helps you to discover research relevant for your work.