Statistical post-processing of dynamical surface air temperature seasonal predictions using the leading ocean-forced spatial patterns

A statistical approach is used to correct dynamical forecasts of seasonal-mean temperatures at the earth's surface (SAT) and at 850 hPa (T850) from a global dynamical model (GCM3). In essence, the approach aims to correct systematic errors in the model's response to the two main patterns o...

Full description

Bibliographic Details
Main Author: Gheti, Rares
Other Authors: Hai Lin (Internal/Cosupervisor2), Jacques F Derome (Internal/Supervisor)
Format: Thesis
Language:English
Published: McGill University 2008
Subjects:
Soi
Online Access:http://digitool.Library.McGill.CA:80/R/?func=dbin-jump-full&object_id=18670
id ftcanadathes:oai:collectionscanada.gc.ca:QMM.18670
record_format openpolar
institution Open Polar
collection Theses Canada/Thèses Canada (Library and Archives Canada)
op_collection_id ftcanadathes
language English
topic Earth Sciences - Atmospheric Sciences
spellingShingle Earth Sciences - Atmospheric Sciences
Gheti, Rares
Statistical post-processing of dynamical surface air temperature seasonal predictions using the leading ocean-forced spatial patterns
topic_facet Earth Sciences - Atmospheric Sciences
description A statistical approach is used to correct dynamical forecasts of seasonal-mean temperatures at the earth's surface (SAT) and at 850 hPa (T850) from a global dynamical model (GCM3). In essence, the approach aims to correct systematic errors in the model's response to the two main patterns of sea surface (SST) anomalies in the tropical Pacific. The SAT (or T850) anomalies that are considered are therefore those related to the tropical Pacific sea surface temperature SST anomalies, as revealed by a Singular Value Decomposition (SVD) analysis. The post-processing method is applied to mean winter forecasts made available by the Historical Forecasting Project for 30 winters (1969 to 1998). We have verified whether the statistical method improves the predictions of the SAT (or T850) patterns associated with the Southern Oscillation Index (SOI), i.e., the El-Niño Southern Oscillation, and the North Atlantic Oscillation (NAO) index. The results show that the method substantially improves the predicted SAT (T850) patterns associated with the NAO. As for the SAT (or T850) pattern associated with the SOI, there is no improvement in the forecasts, likely due to the fact that the original, uncorrected predictions are already quite good. We also examine how the above post-processing translates into actual forecast skill scores, i.e., when all the variability in the observations is considered, as opposed to only the variability linked to the SOI and NAO. We examine the regions where the post-processing has a significant impact on the ensemble forecasts and the possible reasons it is either improving or deteriorating the skill scores. Une approche statistique est utilisée dans le but de corriger des ensembles de prévisions saisonnières de la température de l'air à la surface (SAT) et à 850 hPa (T850) à partir d'un modèle dynamique global (GCM3). Le but de l'approche est de corriger des erreurs systématiques associées à la réponse du modèle aux deux principaux patrons reliés à la réponse de la température à la surface de l'océan (TSO) dans le Pacifique tropical. Les anomalies de la SAT (ou T850) reliées aux anomalies de la TSO sur le Pacifique tropical peuvent être obtenues à partir d'une analyse de décomposition des valeurs singulières (SVD). La méthode de post-traitement est appliquée aux moyennes d'ensemble de prévisions pour l'hiver. Ces ensembles de prévisions ont été fournis par le "Historical Forecasting Project" pour 30 hivers (1969 à 1998). Nous avons vérifié si la méthode statistique améliore les prévisions des patrons de la SAT (ou T850) associés à l'indice de l'oscillation du sud (SOI) ou à l'indice de l'Oscillation Atlantique Nord (NAO). Les résultats montrent que la méthode améliore les prévisions des patrons du SAT (ou T850) associés à l'indexe NAO. Par contre, les patrons du SAT (ou T850 ) associés au SOI, ne sont pas améliorés par le post-traitement statistique. Nous examinons aussi comment la méthode de post-traitement peut améliorer la qualité des prévisions “totale” i.e., quand toute la variabilité atmospherique est prise en compte, et non seulement celle associée aux indices SOI et NAO.
author2 Hai Lin (Internal/Cosupervisor2)
Jacques F Derome (Internal/Supervisor)
format Thesis
author Gheti, Rares
author_facet Gheti, Rares
author_sort Gheti, Rares
title Statistical post-processing of dynamical surface air temperature seasonal predictions using the leading ocean-forced spatial patterns
title_short Statistical post-processing of dynamical surface air temperature seasonal predictions using the leading ocean-forced spatial patterns
title_full Statistical post-processing of dynamical surface air temperature seasonal predictions using the leading ocean-forced spatial patterns
title_fullStr Statistical post-processing of dynamical surface air temperature seasonal predictions using the leading ocean-forced spatial patterns
title_full_unstemmed Statistical post-processing of dynamical surface air temperature seasonal predictions using the leading ocean-forced spatial patterns
title_sort statistical post-processing of dynamical surface air temperature seasonal predictions using the leading ocean-forced spatial patterns
publisher McGill University
publishDate 2008
url http://digitool.Library.McGill.CA:80/R/?func=dbin-jump-full&object_id=18670
op_coverage Master of Science (Department of Atmospheric and Oceanic Sciences)
long_lat ENVELOPE(30.704,30.704,66.481,66.481)
geographic Pacific
Soi
geographic_facet Pacific
Soi
genre North Atlantic
North Atlantic oscillation
genre_facet North Atlantic
North Atlantic oscillation
op_relation Electronically-submitted theses.
http://digitool.Library.McGill.CA:80/R/?func=dbin-jump-full&object_id=18670
op_rights All items in eScholarship@McGill are protected by copyright with all rights reserved unless otherwise indicated.
_version_ 1766137256011104256
spelling ftcanadathes:oai:collectionscanada.gc.ca:QMM.18670 2023-05-15T17:37:21+02:00 Statistical post-processing of dynamical surface air temperature seasonal predictions using the leading ocean-forced spatial patterns Gheti, Rares Hai Lin (Internal/Cosupervisor2) Jacques F Derome (Internal/Supervisor) Master of Science (Department of Atmospheric and Oceanic Sciences) 2008 application/pdf http://digitool.Library.McGill.CA:80/R/?func=dbin-jump-full&object_id=18670 en eng McGill University Electronically-submitted theses. http://digitool.Library.McGill.CA:80/R/?func=dbin-jump-full&object_id=18670 All items in eScholarship@McGill are protected by copyright with all rights reserved unless otherwise indicated. Earth Sciences - Atmospheric Sciences Electronic Thesis or Dissertation 2008 ftcanadathes 2014-02-16T01:00:20Z A statistical approach is used to correct dynamical forecasts of seasonal-mean temperatures at the earth's surface (SAT) and at 850 hPa (T850) from a global dynamical model (GCM3). In essence, the approach aims to correct systematic errors in the model's response to the two main patterns of sea surface (SST) anomalies in the tropical Pacific. The SAT (or T850) anomalies that are considered are therefore those related to the tropical Pacific sea surface temperature SST anomalies, as revealed by a Singular Value Decomposition (SVD) analysis. The post-processing method is applied to mean winter forecasts made available by the Historical Forecasting Project for 30 winters (1969 to 1998). We have verified whether the statistical method improves the predictions of the SAT (or T850) patterns associated with the Southern Oscillation Index (SOI), i.e., the El-Niño Southern Oscillation, and the North Atlantic Oscillation (NAO) index. The results show that the method substantially improves the predicted SAT (T850) patterns associated with the NAO. As for the SAT (or T850) pattern associated with the SOI, there is no improvement in the forecasts, likely due to the fact that the original, uncorrected predictions are already quite good. We also examine how the above post-processing translates into actual forecast skill scores, i.e., when all the variability in the observations is considered, as opposed to only the variability linked to the SOI and NAO. We examine the regions where the post-processing has a significant impact on the ensemble forecasts and the possible reasons it is either improving or deteriorating the skill scores. Une approche statistique est utilisée dans le but de corriger des ensembles de prévisions saisonnières de la température de l'air à la surface (SAT) et à 850 hPa (T850) à partir d'un modèle dynamique global (GCM3). Le but de l'approche est de corriger des erreurs systématiques associées à la réponse du modèle aux deux principaux patrons reliés à la réponse de la température à la surface de l'océan (TSO) dans le Pacifique tropical. Les anomalies de la SAT (ou T850) reliées aux anomalies de la TSO sur le Pacifique tropical peuvent être obtenues à partir d'une analyse de décomposition des valeurs singulières (SVD). La méthode de post-traitement est appliquée aux moyennes d'ensemble de prévisions pour l'hiver. Ces ensembles de prévisions ont été fournis par le "Historical Forecasting Project" pour 30 hivers (1969 à 1998). Nous avons vérifié si la méthode statistique améliore les prévisions des patrons de la SAT (ou T850) associés à l'indice de l'oscillation du sud (SOI) ou à l'indice de l'Oscillation Atlantique Nord (NAO). Les résultats montrent que la méthode améliore les prévisions des patrons du SAT (ou T850) associés à l'indexe NAO. Par contre, les patrons du SAT (ou T850 ) associés au SOI, ne sont pas améliorés par le post-traitement statistique. Nous examinons aussi comment la méthode de post-traitement peut améliorer la qualité des prévisions “totale” i.e., quand toute la variabilité atmospherique est prise en compte, et non seulement celle associée aux indices SOI et NAO. Thesis North Atlantic North Atlantic oscillation Theses Canada/Thèses Canada (Library and Archives Canada) Pacific Soi ENVELOPE(30.704,30.704,66.481,66.481)