Is Institutional Democracy a Good Proxy for Model Independence?

Leduc, Martin; Laprise, René; de Elía, Ramón et Šeparović, Leo (2016). « Is Institutional Democracy a Good Proxy for Model Independence? ». Journal of Climate, 29(23), pp. 8301-8316.

Fichier(s) associé(s) à ce document :
Télécharger (4MB)


Climate models developed within a given research group or institution are prone to share structural similarities, which may induce resembling features in their simulations of the earth’s climate. This assertion, known as the “same-center hypothesis,” is investigated here using a subsample of CMIP3 climate projections constructed by retaining only the models originating from institutions that provided more than one model (or model version). The contributions of individual modeling centers to this ensemble are first presented in terms of climate change projections. A metric for climate change disagreement is then defined to analyze the impact of typical structural differences (such as resolution, parameterizations, or even entire atmosphere and ocean components) on regional climate projections. This metric is compared to a present climate performance metric (correlation of error patterns) within a cross-model comparison framework in terms of their abilities to identify the same-center models. Overall, structural differences between the pairs of same-center models have a stronger impact on climate change projections than on how models reproduce the observed climate. The same-center criterion is used to detect agreements that might be attributable to model similarities and thus that should not be interpreted as implying greater confidence in a given result. It is proposed that such noninformative agreements should be discarded from the ensemble, unless evidence shows that these models can be assumed to be independent. Since this burden of proof is not generally met by the centers participating in a multimodel ensemble, the authors propose an ensemble-weighting scheme based on the assumption of institutional democracy to prevent overconfidence in climate change projections.

Type: Article de revue scientifique
Mots-clés ou Sujets: Filtering techniques, Ensembles, Model comparison, Model evaluation/performance, Climate variability
Unité d'appartenance: Centres institutionnels > Centre pour l'étude et la simulation du climat à l'échelle régionale (ESCER)
Faculté des sciences > Département des sciences de la Terre et de l'atmosphère
Déposé par: René Laprise
Date de dépôt: 14 déc. 2016 13:34
Dernière modification: 28 févr. 2019 13:58
Adresse URL :


Voir les statistiques sur cinq ans...