Post-traitement stochastique des précipitationsjournalières issues de réanalyses: application à laréanalyse CFSR au Canada.

Il est communement admis que la disponibilite de series journalieres de precipitations observees est indispensable pour plusieurs applications. Au Canada, comme pour beaucoup d’autres pays, la densite des stations de mesure est faible et les historiques sont courts. Le developpement des modeles numeriques de temps performants au cours de ces dernieres decennies offre la possibilite de se tourner vers de nouveaux jeux de donnees. Les reanalyses, en particulier, presentent l’avantage d’assimiler tout au long d’une periode donnee divers types d’observations, offrant ainsi un controle continu de la dynamique de l’atmosphere et donc une bonne representation de la meteorologie. Ces dernieres, disponibles sur des grilles couvrant l’ensemble du globe avec une resolution spatio-temporelle donnee, peuvent presenter des erreurs de diverses natures (p.ex., biais, erreur de representativite). Il est alors difficile d’utiliser directement ces donnees comme proxy pour des applications necessitant des donnees locales. Dans ce contexte, le present projet s’interesse a post-traiter les donnees de precipitations journalieres issues d’une reanalyse nommee Climate Forecast System Reanalysis (CFSR), afin de proposer des series non biaisees, ayant des caracteristiques locales (par opposition au point de grille) et ce notamment aux endroits depourvus d’observations. Le projet s’articule autour de trois axes principaux: i) analyser un modele probabiliste base sur des approches de regression afin de post-traiter les precipitations journalieres de la reanalyse CFSR en se basant sur des stations d’observation; ii) integrer la structure spatio-temporelle du processus de precipitations dans ces memes modeles pour ameliorer l’estimation des series post-traitees; iii) et proposer des champs journaliers de precipitations en combinant les modeles du second axe a des modeles spatiaux. Le post-traitement developpe donne des resultats tres encourageants quant a la correction systematique du biais des sorties de reanalyses, mais aussi concernant la representation de plusieurs caracteristiques locales des precipitations. Cette etude ouvre, par ailleurs, des perspectives tres interessantes a la fois methodologique (p.ex: implementation pour les precipitations extremes), mais aussi en termes de champs d’utilisation avec l’application de ces approches aux scenarios de modeles climatiques pour l’analyse de l’evolution des precipitations locales sous un climat changeant. Abstract It is widely recognized that the availability of observed daily precipitation series is essential for several applications. The most important challenge that many countries face, including Canada, is to characterize historical precipitation considering the low station density in many of their regions and the short sample size. Reanalysis, generated by Numerical Weather Prediction methods assimilating past observations, is an attractive alternative as they provide coherent, spatially and temporally continuous meteorological fields for a specific period and domain. However, reanalysis, available on grids covering the whole globe, present errors of various natures (e.g., bias, representativeness error) that prevent from their direct use as a proxy for applications that require local data. In this context, the present project is interested in post-processing the daily precipitation data from one reanalysis, Climate Forecast System Reanalysis (CFSR), in order to propose unbiased series, with local characteristics (as opposed to grid points), even at places without observations. The project conducted here was organized around three major axes: i) to analyze a probabilistic model based on regression approaches in order to post-treat the daily CFSR precipitation at sites with observations; ii) to consider the spatio-temporal structure of the precipitation process into these same models to improve the estimation of post-processed series at the daily scale; and (iii) to propose daily precipitation fields by combining the second-axis models with spatial models so that to propose posttreated daily series at each grid point of the domain. The developed stochastically based post-treatment bring very encouraging results by systematically correcting CFSR biases but also by providing good representation of several local characteristics of the precipitation process. This study also opens very interesting perspectives, as regards the improvement of the methodology (e.g., explicit implementation of the extreme precipitation) but also concerning the enlargement of the application fields. For example, the current approach could be applied to climate model scenarios to provide analysis of the evolution of local precipitation in a changing climate.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.