caa a22 4500 605471436 CHVBK 20210128100333.0 cr unu---uuuuu 210128e20151001xx s 000 0 eng 10.1007/s00382-014-2445-5 doi (NATIONALLICENCE)springer-10.1007/s00382-014-2445-5 Can metric-based approaches really improve multi-model climate projections? The case of summer temperature change in France [Elektronische Daten] [Julien Boé, Laurent Terray] The multi-model ensemble mean is generally used as a default approach to estimate climate change signals, based on the implicit hypothesis that all models provide equally credible projections. As this hypothesis is unlikely to be true, it is in theory possible to obtain more realistic projections by giving more weight to more realistic models according to a relevant metric, if such a metric exists. This alternative approach however raises many methodological issues. In this study, a methodological framework based on a perfect model approach is described. It is intended to provide some useful elements of answer to these methodological issues. The basic idea is to take a random climate model and treat it as if it were the truth (or "synthetic observations”). Then, all the other members from the multi-model ensemble are used to derive thanks to a metric-based approach a posterior estimate of the future change, based on the synthetic observation of the metric. This posterior estimate can be compared to the synthetic observation of future change to evaluate the skill of the approach. This general framework is applied to future summer temperature change in France. A process-based metric, related to cloud-temperature interactions is tested, with different simple statistical methods to combine multiple model results (e.g. weighted average, model selection, regression.) Except in presence of large observational errors in the metric, metric-based methods using the metric related to cloud temperature interactions generally lead to large reductions of errors compared to the ensemble mean, but the sensitivity to methodological choices is important . Springer-Verlag Berlin Heidelberg, 2015 Climate change nationallicence France nationallicence Metrics nationallicence Multi-model nationallicence Uncertainties nationallicence Boé Julien URA1875 CNRS/CERFACS, Toulouse, France aut Terray Laurent URA1875 CNRS/CERFACS, Toulouse, France aut Climate Dynamics Springer Berlin Heidelberg 45/7-8(2015-10-01), 1913-1928 0930-7575 45:7-8<1913 2015 45 382 https://doi.org/10.1007/s00382-014-2445-5 text/html Onlinezugriff via DOI BK010053 XK010053 XK010000 Metadata rights reserved Springer special CC-BY-NC licence nationallicence 1 research-article jats NATIONALLICENCE NATIONALLICENCE NL-springer NATIONALLICENCE 856 40 https://doi.org/10.1007/s00382-014-2445-5 text/html Onlinezugriff via DOI NATIONALLICENCE 700 1- Boé Julien URA1875 CNRS/CERFACS, Toulouse, France aut NATIONALLICENCE 700 1- Terray Laurent URA1875 CNRS/CERFACS, Toulouse, France aut NATIONALLICENCE 773 0- Climate Dynamics Springer Berlin Heidelberg 45/7-8(2015-10-01), 1913-1928 0930-7575 45:7-8<1913 2015 45 382