Neural networks input-based models to calibrate the mean precipitation of an ensemble prediction system
Autor(a) principal: | |
---|---|
Data de Publicação: | 2016 |
Tipo de documento: | Tese |
Idioma: | eng |
Título da fonte: | Biblioteca Digital de Teses e Dissertações do INPE |
Texto Completo: | http://urlib.net/sid.inpe.br/mtc-m21b/2016/08.05.13.14 |
Resumo: | Although ensemble forecasting systems provide richer forecasts by adding probabilistic concepts to single deterministic forecasts, they have intrinsic shortcomings caused by the lack of full comprehension of the relationship between meteorological variables. It is especially noticed in medium and large-scale forecasts, whose effects of chaotic behavior of the atmosphere drastically increase as the target forecasting date is lengthened. Improvements on weather forecasting systems can be done either by the meteorology staff concerning physical aspects of weather behavior as well as by implementing computational statistical methods in order to tune the weather forecasting model output. The purpose of this work is to compute, along the forecast horizon, a more accurate precipitation value than the ensemble mean precipitation by post-processing INPE/CPTEC's ensemble prediction output. To achieve the goal, some prognostic fields and derived data are combined and submitted as explanatory variables to an artificial neural network system. Experiments were guided in an exploratory way such that several computational models were generated and thereafter assessed. The study was individually performed at some grid points located within the boundaries of La Plata Basin. Results indicate that the application of this methodology presented values closer to actual values when compared to the ensemble mean precipitation. It also shows that the inclusion of the ensemble mean precipitation itself, as well as data from adjacent grid points, improve the calibration process of the target grid point. In addition, the exploratory approach detects different artificial network models to fit specific location and lead-time. Although this input-driven system computes less than ideal forecasting values, it performs better than the mean output of the ensemble model, which is widely used in various weather forecasting products. |
id |
INPE_d08c0a2219b6dc0ad0a7755397acb5ca |
---|---|
oai_identifier_str |
oai:urlib.net:sid.inpe.br/mtc-m21b/2016/08.05.13.14.01-0 |
network_acronym_str |
INPE |
network_name_str |
Biblioteca Digital de Teses e Dissertações do INPE |
spelling |
info:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/doctoralThesisNeural networks input-based models to calibrate the mean precipitation of an ensemble prediction systemModelos de redes neurais baseados em entrada para calibrar a precipitação média de um sistema de previsão por conjuntos2016-08-24Rafael Duarte Coelho dos SantosChristopher Alexander Cunningham CastroMarcos Gonçalves QuilesNandamudi Lankalapalli VijaykumarCarlos Henrique Quartucci ForsterFábio Dall CortivoJosé Roberto Motta GarciaInstituto Nacional de Pesquisas Espaciais (INPE)Programa de Pós-Graduação do INPE em Computação AplicadaINPEBRmachine learningneural netsstatistical weather forecastingrainaprendizado de máquinaredes neurais artificiaisprevisão do tempochuvaAlthough ensemble forecasting systems provide richer forecasts by adding probabilistic concepts to single deterministic forecasts, they have intrinsic shortcomings caused by the lack of full comprehension of the relationship between meteorological variables. It is especially noticed in medium and large-scale forecasts, whose effects of chaotic behavior of the atmosphere drastically increase as the target forecasting date is lengthened. Improvements on weather forecasting systems can be done either by the meteorology staff concerning physical aspects of weather behavior as well as by implementing computational statistical methods in order to tune the weather forecasting model output. The purpose of this work is to compute, along the forecast horizon, a more accurate precipitation value than the ensemble mean precipitation by post-processing INPE/CPTEC's ensemble prediction output. To achieve the goal, some prognostic fields and derived data are combined and submitted as explanatory variables to an artificial neural network system. Experiments were guided in an exploratory way such that several computational models were generated and thereafter assessed. The study was individually performed at some grid points located within the boundaries of La Plata Basin. Results indicate that the application of this methodology presented values closer to actual values when compared to the ensemble mean precipitation. It also shows that the inclusion of the ensemble mean precipitation itself, as well as data from adjacent grid points, improve the calibration process of the target grid point. In addition, the exploratory approach detects different artificial network models to fit specific location and lead-time. Although this input-driven system computes less than ideal forecasting values, it performs better than the mean output of the ensemble model, which is widely used in various weather forecasting products.Embora os sistemas de previsão por conjuntos forneçam previsões mais ricas, acrescentando conceitos probabilísticos às previsões determinísticas simples, eles têm deficiências intrínsecas causadas pela falta de plena compreensão da relação entre variáveis meteorológicas. Isto é especialmente notado nas previsões de média e grande escala, cujos efeitos de comportamento caótico da atmosfera aumentam drasticamente à medida que a data alvo de previsão é estendida. Melhorias nos sistemas de previsão do tempo podem ser feitas tanto pela equipe de meteorologia em relação aos aspectos físicos do comportamento do tempo, bem como através da aplicação de métodos computacionais estatísticos, visando ajustar a saída do modelo de previsão. O objetivo deste trabalho é calcular, ao longo do horizonte de previsão, um valor de precipitação mais preciso do que a precipitação média do sistema de previsão por conjunto através do pós-processamento das saídas do modelo do INPE/CPTEC. Para atingir esta meta, alguns campos de prognóstico e dados derivados são combinados e apresentados como variáveis explanatórias a um sistema de rede neural artificial. Os experimentos foram orientados de forma exploratória onde vários modelos computacionais foram gerados e posteriormente avaliados. O estudo foi realizado individualmente em alguns pontos de grade localizados dentro dos limites da Bacia da Prata. Os resultados indicam que a aplicação desta metodologia apresentou valores mais próximos da realidade do que a média de precipitação do sistema de previsão por conjuntos. Mostra também que ao incluir a própria precipitação média, bem como dados de pontos de grade adjacentes o processo de calibração melhora no ponto de grade alvo. Além disso, a abordagem exploratória traz uma melhora ainda maior pois detecta diferentes modelos de redes neurais para locais e dias de previsão específicos. Embora este sistema baseado em entrada calcule valores de previsão inferiores aos ideais, ele tem um desempenho melhor do que a média do modelo de previsão por conjunto, que é amplamente usado em vários produtos de previsão de tempo.http://urlib.net/sid.inpe.br/mtc-m21b/2016/08.05.13.14info:eu-repo/semantics/openAccessengreponame:Biblioteca Digital de Teses e Dissertações do INPEinstname:Instituto Nacional de Pesquisas Espaciais (INPE)instacron:INPE2021-07-31T06:55:11Zoai:urlib.net:sid.inpe.br/mtc-m21b/2016/08.05.13.14.01-0Biblioteca Digital de Teses e Dissertaçõeshttp://bibdigital.sid.inpe.br/PUBhttp://bibdigital.sid.inpe.br/col/iconet.com.br/banon/2003/11.21.21.08/doc/oai.cgiopendoar:32772021-07-31 06:55:12.052Biblioteca Digital de Teses e Dissertações do INPE - Instituto Nacional de Pesquisas Espaciais (INPE)false |
dc.title.en.fl_str_mv |
Neural networks input-based models to calibrate the mean precipitation of an ensemble prediction system |
dc.title.alternative.pt.fl_str_mv |
Modelos de redes neurais baseados em entrada para calibrar a precipitação média de um sistema de previsão por conjuntos |
title |
Neural networks input-based models to calibrate the mean precipitation of an ensemble prediction system |
spellingShingle |
Neural networks input-based models to calibrate the mean precipitation of an ensemble prediction system José Roberto Motta Garcia |
title_short |
Neural networks input-based models to calibrate the mean precipitation of an ensemble prediction system |
title_full |
Neural networks input-based models to calibrate the mean precipitation of an ensemble prediction system |
title_fullStr |
Neural networks input-based models to calibrate the mean precipitation of an ensemble prediction system |
title_full_unstemmed |
Neural networks input-based models to calibrate the mean precipitation of an ensemble prediction system |
title_sort |
Neural networks input-based models to calibrate the mean precipitation of an ensemble prediction system |
author |
José Roberto Motta Garcia |
author_facet |
José Roberto Motta Garcia |
author_role |
author |
dc.contributor.advisor1.fl_str_mv |
Rafael Duarte Coelho dos Santos |
dc.contributor.advisor2.fl_str_mv |
Christopher Alexander Cunningham Castro |
dc.contributor.referee1.fl_str_mv |
Marcos Gonçalves Quiles |
dc.contributor.referee2.fl_str_mv |
Nandamudi Lankalapalli Vijaykumar |
dc.contributor.referee3.fl_str_mv |
Carlos Henrique Quartucci Forster |
dc.contributor.referee4.fl_str_mv |
Fábio Dall Cortivo |
dc.contributor.author.fl_str_mv |
José Roberto Motta Garcia |
contributor_str_mv |
Rafael Duarte Coelho dos Santos Christopher Alexander Cunningham Castro Marcos Gonçalves Quiles Nandamudi Lankalapalli Vijaykumar Carlos Henrique Quartucci Forster Fábio Dall Cortivo |
dc.description.abstract.por.fl_txt_mv |
Although ensemble forecasting systems provide richer forecasts by adding probabilistic concepts to single deterministic forecasts, they have intrinsic shortcomings caused by the lack of full comprehension of the relationship between meteorological variables. It is especially noticed in medium and large-scale forecasts, whose effects of chaotic behavior of the atmosphere drastically increase as the target forecasting date is lengthened. Improvements on weather forecasting systems can be done either by the meteorology staff concerning physical aspects of weather behavior as well as by implementing computational statistical methods in order to tune the weather forecasting model output. The purpose of this work is to compute, along the forecast horizon, a more accurate precipitation value than the ensemble mean precipitation by post-processing INPE/CPTEC's ensemble prediction output. To achieve the goal, some prognostic fields and derived data are combined and submitted as explanatory variables to an artificial neural network system. Experiments were guided in an exploratory way such that several computational models were generated and thereafter assessed. The study was individually performed at some grid points located within the boundaries of La Plata Basin. Results indicate that the application of this methodology presented values closer to actual values when compared to the ensemble mean precipitation. It also shows that the inclusion of the ensemble mean precipitation itself, as well as data from adjacent grid points, improve the calibration process of the target grid point. In addition, the exploratory approach detects different artificial network models to fit specific location and lead-time. Although this input-driven system computes less than ideal forecasting values, it performs better than the mean output of the ensemble model, which is widely used in various weather forecasting products. Embora os sistemas de previsão por conjuntos forneçam previsões mais ricas, acrescentando conceitos probabilísticos às previsões determinísticas simples, eles têm deficiências intrínsecas causadas pela falta de plena compreensão da relação entre variáveis meteorológicas. Isto é especialmente notado nas previsões de média e grande escala, cujos efeitos de comportamento caótico da atmosfera aumentam drasticamente à medida que a data alvo de previsão é estendida. Melhorias nos sistemas de previsão do tempo podem ser feitas tanto pela equipe de meteorologia em relação aos aspectos físicos do comportamento do tempo, bem como através da aplicação de métodos computacionais estatísticos, visando ajustar a saída do modelo de previsão. O objetivo deste trabalho é calcular, ao longo do horizonte de previsão, um valor de precipitação mais preciso do que a precipitação média do sistema de previsão por conjunto através do pós-processamento das saídas do modelo do INPE/CPTEC. Para atingir esta meta, alguns campos de prognóstico e dados derivados são combinados e apresentados como variáveis explanatórias a um sistema de rede neural artificial. Os experimentos foram orientados de forma exploratória onde vários modelos computacionais foram gerados e posteriormente avaliados. O estudo foi realizado individualmente em alguns pontos de grade localizados dentro dos limites da Bacia da Prata. Os resultados indicam que a aplicação desta metodologia apresentou valores mais próximos da realidade do que a média de precipitação do sistema de previsão por conjuntos. Mostra também que ao incluir a própria precipitação média, bem como dados de pontos de grade adjacentes o processo de calibração melhora no ponto de grade alvo. Além disso, a abordagem exploratória traz uma melhora ainda maior pois detecta diferentes modelos de redes neurais para locais e dias de previsão específicos. Embora este sistema baseado em entrada calcule valores de previsão inferiores aos ideais, ele tem um desempenho melhor do que a média do modelo de previsão por conjunto, que é amplamente usado em vários produtos de previsão de tempo. |
description |
Although ensemble forecasting systems provide richer forecasts by adding probabilistic concepts to single deterministic forecasts, they have intrinsic shortcomings caused by the lack of full comprehension of the relationship between meteorological variables. It is especially noticed in medium and large-scale forecasts, whose effects of chaotic behavior of the atmosphere drastically increase as the target forecasting date is lengthened. Improvements on weather forecasting systems can be done either by the meteorology staff concerning physical aspects of weather behavior as well as by implementing computational statistical methods in order to tune the weather forecasting model output. The purpose of this work is to compute, along the forecast horizon, a more accurate precipitation value than the ensemble mean precipitation by post-processing INPE/CPTEC's ensemble prediction output. To achieve the goal, some prognostic fields and derived data are combined and submitted as explanatory variables to an artificial neural network system. Experiments were guided in an exploratory way such that several computational models were generated and thereafter assessed. The study was individually performed at some grid points located within the boundaries of La Plata Basin. Results indicate that the application of this methodology presented values closer to actual values when compared to the ensemble mean precipitation. It also shows that the inclusion of the ensemble mean precipitation itself, as well as data from adjacent grid points, improve the calibration process of the target grid point. In addition, the exploratory approach detects different artificial network models to fit specific location and lead-time. Although this input-driven system computes less than ideal forecasting values, it performs better than the mean output of the ensemble model, which is widely used in various weather forecasting products. |
publishDate |
2016 |
dc.date.issued.fl_str_mv |
2016-08-24 |
dc.type.status.fl_str_mv |
info:eu-repo/semantics/publishedVersion |
dc.type.driver.fl_str_mv |
info:eu-repo/semantics/doctoralThesis |
status_str |
publishedVersion |
format |
doctoralThesis |
dc.identifier.uri.fl_str_mv |
http://urlib.net/sid.inpe.br/mtc-m21b/2016/08.05.13.14 |
url |
http://urlib.net/sid.inpe.br/mtc-m21b/2016/08.05.13.14 |
dc.language.iso.fl_str_mv |
eng |
language |
eng |
dc.rights.driver.fl_str_mv |
info:eu-repo/semantics/openAccess |
eu_rights_str_mv |
openAccess |
dc.publisher.none.fl_str_mv |
Instituto Nacional de Pesquisas Espaciais (INPE) |
dc.publisher.program.fl_str_mv |
Programa de Pós-Graduação do INPE em Computação Aplicada |
dc.publisher.initials.fl_str_mv |
INPE |
dc.publisher.country.fl_str_mv |
BR |
publisher.none.fl_str_mv |
Instituto Nacional de Pesquisas Espaciais (INPE) |
dc.source.none.fl_str_mv |
reponame:Biblioteca Digital de Teses e Dissertações do INPE instname:Instituto Nacional de Pesquisas Espaciais (INPE) instacron:INPE |
reponame_str |
Biblioteca Digital de Teses e Dissertações do INPE |
collection |
Biblioteca Digital de Teses e Dissertações do INPE |
instname_str |
Instituto Nacional de Pesquisas Espaciais (INPE) |
instacron_str |
INPE |
institution |
INPE |
repository.name.fl_str_mv |
Biblioteca Digital de Teses e Dissertações do INPE - Instituto Nacional de Pesquisas Espaciais (INPE) |
repository.mail.fl_str_mv |
|
publisher_program_txtF_mv |
Programa de Pós-Graduação do INPE em Computação Aplicada |
contributor_advisor1_txtF_mv |
Rafael Duarte Coelho dos Santos |
_version_ |
1706809359007219712 |