Bayesian inference for wind field retrieval

In many problems in spatial statistics it; is necessary to infer a global p roblem solution by combining local models, A principled approach to this pr oblem is to develop a global probabilistic model for the relationships betw een local variables and to use this as the prior in a Bayesian inference pr ocedure. We use a Gaussian process with hyper-parameters estimated from num erical weather prediction models, which yields meteorologically convincing wind fields. We use neural networks to make local estimates of wind vector probabilities. The resulting inference problem cannot be solved analyticall y, but Markov Chain Monte Carlo methods allow us to retrieve accurate wind fields. (C) 2000 Elsevier Science B.V. All rights reserved.