Bayesian inference for wind field retrieval

Ian T. Nabney, Dan Cornford, Christopher K. I. Williams

Research output: Contribution to journalArticle

Abstract

In many problems in spatial statistics it is necessary to infer a global problem solution by combining local models. A principled approach to this problem is to develop a global probabilistic model for the relationships between local variables and to use this as the prior in a Bayesian inference procedure. We show how a Gaussian process with hyper-parameters estimated from Numerical Weather Prediction Models yields meteorologically convincing wind fields. We use neural networks to make local estimates of wind vector probabilities. The resulting inference problem cannot be solved analytically, but Markov Chain Monte Carlo methods allow us to retrieve accurate wind fields.
Original languageEnglish
Pages (from-to)3-11
Number of pages9
JournalNeurocomputing
Volume30
Issue number1-4
DOIs
Publication statusPublished - Jan 2000

Fingerprint

Monte Carlo Method
Markov Chains
Weather
Statistical Models
Markov processes
Monte Carlo methods
Statistics
Neural networks

Keywords

  • Bayesian inference
  • surface winds
  • spatial priors
  • Gaussian processes

Cite this

Nabney, I. T., Cornford, D., & Williams, C. K. I. (2000). Bayesian inference for wind field retrieval. Neurocomputing, 30(1-4), 3-11. https://doi.org/10.1016/S0925-2312(99)00136-8
Nabney, Ian T. ; Cornford, Dan ; Williams, Christopher K. I. / Bayesian inference for wind field retrieval. In: Neurocomputing. 2000 ; Vol. 30, No. 1-4. pp. 3-11.
@article{e8d49439d557469fa9b77df018945eff,
title = "Bayesian inference for wind field retrieval",
abstract = "In many problems in spatial statistics it is necessary to infer a global problem solution by combining local models. A principled approach to this problem is to develop a global probabilistic model for the relationships between local variables and to use this as the prior in a Bayesian inference procedure. We show how a Gaussian process with hyper-parameters estimated from Numerical Weather Prediction Models yields meteorologically convincing wind fields. We use neural networks to make local estimates of wind vector probabilities. The resulting inference problem cannot be solved analytically, but Markov Chain Monte Carlo methods allow us to retrieve accurate wind fields.",
keywords = "Bayesian inference, surface winds, spatial priors, Gaussian processes",
author = "Nabney, {Ian T.} and Dan Cornford and Williams, {Christopher K. I.}",
year = "2000",
month = "1",
doi = "10.1016/S0925-2312(99)00136-8",
language = "English",
volume = "30",
pages = "3--11",
journal = "Neurocomputing",
issn = "0925-2312",
publisher = "Elsevier",
number = "1-4",

}

Nabney, IT, Cornford, D & Williams, CKI 2000, 'Bayesian inference for wind field retrieval', Neurocomputing, vol. 30, no. 1-4, pp. 3-11. https://doi.org/10.1016/S0925-2312(99)00136-8

Bayesian inference for wind field retrieval. / Nabney, Ian T.; Cornford, Dan; Williams, Christopher K. I.

In: Neurocomputing, Vol. 30, No. 1-4, 01.2000, p. 3-11.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Bayesian inference for wind field retrieval

AU - Nabney, Ian T.

AU - Cornford, Dan

AU - Williams, Christopher K. I.

PY - 2000/1

Y1 - 2000/1

N2 - In many problems in spatial statistics it is necessary to infer a global problem solution by combining local models. A principled approach to this problem is to develop a global probabilistic model for the relationships between local variables and to use this as the prior in a Bayesian inference procedure. We show how a Gaussian process with hyper-parameters estimated from Numerical Weather Prediction Models yields meteorologically convincing wind fields. We use neural networks to make local estimates of wind vector probabilities. The resulting inference problem cannot be solved analytically, but Markov Chain Monte Carlo methods allow us to retrieve accurate wind fields.

AB - In many problems in spatial statistics it is necessary to infer a global problem solution by combining local models. A principled approach to this problem is to develop a global probabilistic model for the relationships between local variables and to use this as the prior in a Bayesian inference procedure. We show how a Gaussian process with hyper-parameters estimated from Numerical Weather Prediction Models yields meteorologically convincing wind fields. We use neural networks to make local estimates of wind vector probabilities. The resulting inference problem cannot be solved analytically, but Markov Chain Monte Carlo methods allow us to retrieve accurate wind fields.

KW - Bayesian inference

KW - surface winds

KW - spatial priors

KW - Gaussian processes

UR - http://www.scopus.com/inward/record.url?scp=0033991598&partnerID=8YFLogxK

U2 - 10.1016/S0925-2312(99)00136-8

DO - 10.1016/S0925-2312(99)00136-8

M3 - Article

VL - 30

SP - 3

EP - 11

JO - Neurocomputing

JF - Neurocomputing

SN - 0925-2312

IS - 1-4

ER -