Post-traitement stochastique des précipitations \njournalières issues de réanalyses: application à la \nréanalyse CFSR au Canada.
Bibliographic record
Abstract
Il est communément admis que la disponibilité de séries journalières de précipitations observées est \nindispensable pour plusieurs applications. Au Canada, comme pour beaucoup d’autres pays, la densité des \nstations de mesure est faible et les historiques sont courts. Le développement des modèles numériques \nde temps performants au cours de ces dernières décennies offre la possibilité de se tourner vers de \nnouveaux jeux de données. Les réanalyses, en particulier, présentent l’avantage d’assimiler tout au long \nd’une période donnée divers types d’observations, offrant ainsi un contrôle continu de la dynamique de \nl’atmosphère et donc une bonne représentation de la météorologie. Ces dernières, disponibles sur des \ngrilles couvrant l’ensemble du globe avec une résolution spatio-temporelle donnée, peuvent présenter \ndes erreurs de diverses natures (p.ex., biais, erreur de représentativité). Il est alors difficile d’utiliser \ndirectement ces données comme proxy pour des applications nécessitant des données locales. Dans ce \ncontexte, le présent projet s’intéresse à post-traiter les données de précipitations journalières issues d’une \nréanalyse nommée Climate Forecast System Reanalysis (CFSR), afin de proposer des séries non biaisées, \nayant des caractéristiques locales (par opposition au point de grille) et ce notamment aux endroits \ndépourvus d’observations. \nLe projet s’articule autour de trois axes principaux: i) analyser un modèle probabiliste basé sur \ndes approches de régression afin de post-traiter les précipitations journalières de la réanalyse CFSR \nen se basant sur des stations d’observation; ii) intégrer la structure spatio-temporelle du processus de \nprécipitations dans ces mêmes modèles pour améliorer l’estimation des séries post-traitées; iii) et proposer \ndes champs journaliers de précipitations en combinant les modèles du second axe à des modèles spatiaux. \nLe post-traitement développé donne des résultats très encourageants quant à la correction systématique \ndu biais des sorties de réanalyses, mais aussi concernant la représentation de plusieurs caractéristiques \nlocales des précipitations. Cette étude ouvre, par ailleurs, des perspectives très intéressantes \nà la fois méthodologique (p.ex: implémentation pour les précipitations extrêmes), mais aussi en termes \nde champs d’utilisation avec l’application de ces approches aux scénarios de modèles climatiques pour \nl’analyse de l’évolution des précipitations locales sous un climat changeant. It is widely recognized that the availability of observed daily precipitation series is essential for several \napplications. The most important challenge that many countries face, including Canada, is to characterize \nhistorical precipitation considering the low station density in many of their regions and the short sample \nsize. Reanalysis, generated by Numerical Weather Prediction methods assimilating past observations, is \nan attractive alternative as they provide coherent, spatially and temporally continuous meteorological \nfields for a specific period and domain. However, reanalysis, available on grids covering the whole globe, \npresent errors of various natures (e.g., bias, representativeness error) that prevent from their direct use \nas a proxy for applications that require local data. In this context, the present project is interested in \npost-processing the daily precipitation data from one reanalysis, Climate Forecast System Reanalysis \n(CFSR), in order to propose unbiased series, with local characteristics (as opposed to grid points), even \nat places without observations. \nThe project conducted here was organized around three major axes: i) to analyze a probabilistic \nmodel based on regression approaches in order to post-treat the daily CFSR precipitation at sites with \nobservations; ii) to consider the spatio-temporal structure of the precipitation process into these same \nmodels to improve the estimation of post-processed series at the daily scale; and (iii) to propose daily \nprecipitation fields by combining the second-axis models with spatial models so that to propose posttreated \ndaily series at each grid point of the domain. \nThe developed stochastically based post-treatment bring very encouraging results by systematically \ncorrecting CFSR biases but also by providing good representation of several local characteristics of the \nprecipitation process. This study also opens very interesting perspectives, as regards the improvement \nof the methodology (e.g., explicit implementation of the extreme precipitation) but also concerning the \nenlargement of the application fields. For example, the current approach could be applied to climate \nmodel 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.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.008 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.004 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".