Repositorio Institucional
Repositorio Institucional
CONICET Digital
  • Inicio
  • EXPLORAR
    • AUTORES
    • DISCIPLINAS
    • COMUNIDADES
  • Estadísticas
  • Novedades
    • Noticias
    • Boletines
  • Ayuda
    • General
    • Datos de investigación
  • Acerca de
    • CONICET Digital
    • Equipo
    • Red Federal
  • Contacto
JavaScript is disabled for your browser. Some features of this site may not work without it.
  • INFORMACIÓN GENERAL
  • RESUMEN
  • ESTADISTICAS
 
Artículo

Evaluation of machine learning techniques for forecast uncertainty quantification

Sacco, Maximiliano Antonio; Ruiz, Juan JoseIcon ; Pulido, Manuel ArturoIcon ; Tandeo, Pierre
Fecha de publicación: 10/2022
Editorial: John Wiley & Sons Ltd
Revista: Quarterly Journal of the Royal Meteorological Society
ISSN: 0035-9009
Idioma: Inglés
Tipo de recurso: Artículo publicado
Clasificación temática:
Meteorología y Ciencias Atmosféricas

Resumen

Ensemble forecasting is, so far, the most successful approach to produce relevant forecasts with an estimation of their uncertainty. The main limitations of ensemble forecasting are the high computational cost and the difficulty to capture and quantify different sources of uncertainty, particularly those associated with model errors. In this article we perform toy-model and state-of-the-art model experiments to analyze to what extent artificial neural networks (ANNs) are able to model the different sources of uncertainty present in a forecast. In particular, those associated with the accuracy of the initial conditions and those introduced by the model error. We also compare different training strategies: one based on a direct training using the mean and spread of an ensemble forecast as target, and the other ones rely on an indirect training strategy using an analyzed state as target in which the uncertainty is implicitly learned from the data. Experiments using the Lorenz'96 model show that the ANNs are able to emulate some of the properties of ensemble forecasts like the filtering of the most unpredictable modes and a state-dependent quantification of the forecast uncertainty. Moreover, ANNs provide a reliable estimation of the forecast uncertainty in the presence of model error. Preliminary experiments conducted with a state-of-the-art forecasting system also confirm the ability of ANNs to produce a reliable quantification of the forecast uncertainty.
Palabras clave: CHAOTIC DYNAMIC MODELS , FORECAST , NEURAL NETWORKS , OBSERVATION LIKELIHOOD LOSS FUNCTION , UNCERTAINTY QUANTIFICATION
Ver el registro completo
 
Archivos asociados
Tamaño: 4.925Mb
Formato: PDF
.
Solicitar
Licencia
info:eu-repo/semantics/restrictedAccess Excepto donde se diga explícitamente, este item se publica bajo la siguiente descripción: Creative Commons Attribution-NonCommercial-ShareAlike 2.5 Unported (CC BY-NC-SA 2.5)
Identificadores
URI: http://hdl.handle.net/11336/204394
DOI: http://dx.doi.org/10.1002/qj.4362
URL: https://rmets.onlinelibrary.wiley.com/doi/10.1002/qj.4362
Colecciones
Articulos(IMIT)
Articulos de INST.DE MODELADO E INNOVACION TECNOLOGICA
Citación
Sacco, Maximiliano Antonio; Ruiz, Juan Jose; Pulido, Manuel Arturo; Tandeo, Pierre; Evaluation of machine learning techniques for forecast uncertainty quantification; John Wiley & Sons Ltd; Quarterly Journal of the Royal Meteorological Society; 148; 749; 10-2022; 3470-3490
Compartir
Altmétricas
 

Enviar por e-mail
Separar cada destinatario (hasta 5) con punto y coma.
  • Facebook
  • X Conicet Digital
  • Instagram
  • YouTube
  • Sound Cloud
  • LinkedIn

Los contenidos del CONICET están licenciados bajo Creative Commons Reconocimiento 2.5 Argentina License

https://www.conicet.gov.ar/ - CONICET

Inicio

Explorar

  • Autores
  • Disciplinas
  • Comunidades

Estadísticas

Novedades

  • Noticias
  • Boletines

Ayuda

Acerca de

  • CONICET Digital
  • Equipo
  • Red Federal

Contacto

Godoy Cruz 2290 (C1425FQB) CABA – República Argentina – Tel: +5411 4899-5400 repositorio@conicet.gov.ar
TÉRMINOS Y CONDICIONES