Dépôt numérique
RECHERCHER

Multisite and multivariable statistical downscaling using a Gaussian copula quantile regression model.

Ben Alaya, Mohamed Ali; Chebana, Fateh; Ouarda, Taha B. M. J. (2016). Multisite and multivariable statistical downscaling using a Gaussian copula quantile regression model. Climate Dynamics , vol. 47 , nº 5-6. p. 1383-1397. DOI: 10.1007/s00382-015-2908-3.

Ce document n'est pas hébergé sur EspaceINRS.

Résumé

Statistical downscaling techniques are required to refine atmosphere–ocean global climate data and provide reliable meteorological information such as a realistic temporal variability and relationships between sites and variables in a changing climate. To this end, the present paper introduces a modular structure combining two statistical tools of increasing interest during the last years: (1) Gaussian copula and (2) quantile regression. The quantile regression tool is employed to specify the entire conditional distribution of downscaled variables and to address the limitations of traditional regression-based approaches whereas the Gaussian copula is performed to describe and preserve the dependence between both variables and sites. A case study based on precipitation and maximum and minimum temperatures from the province of Quebec, Canada, is used to evaluate the performance of the proposed model. Obtained results suggest that this approach is capable of generating series with realistic correlation structures and temporal variability. Furthermore, the proposed model performed better than a classical multisite multivariate statistical downscaling model for most evaluation criteria.

Type de document: Article
Mots-clés libres: climate downscaling; Gaussian copula; quantile regression; temperature; precipitation; multisite; multivariable
Centre: Centre Eau Terre Environnement
Date de dépôt: 07 mai 2018 18:32
Dernière modification: 07 mai 2018 18:32
URI: http://espace.inrs.ca/id/eprint/6314

Actions (Identification requise)

Modifier la notice Modifier la notice