A data fusion model for meteorological data using the INLA-SPDE method DOI Creative Commons
Stephen Jun Villejo, Sara Martino, Finn Lindgren

и другие.

Journal of the Royal Statistical Society Series C (Applied Statistics), Год журнала: 2025, Номер unknown

Опубликована: Фев. 27, 2025

Abstract We present a data fusion model designed to address the problem of sparse observational by incorporating numerical forecast models as an additional source improve predictions key variables. This is applied two main meteorological sources in Philippines. The approach assumes that different are imperfect representations common underlying process. Observations from weather stations follow classical error model, while forecasts involve both constant multiplicative bias and additive bias, which spatially structured time-varying. To perform inference, we use Bayesian averaging technique combined with integrated nested Laplace approximation. model’s performance evaluated through simulation study, where it consistently results better more accurate parameter estimates than using only or regression calibration, particularly cases data. In application, proposed also outperforms these benchmark approaches, demonstrated leave-group-out cross-validation.

Язык: Английский

A data fusion model for meteorological data using the INLA-SPDE method DOI Creative Commons
Stephen Jun Villejo, Sara Martino, Finn Lindgren

и другие.

Journal of the Royal Statistical Society Series C (Applied Statistics), Год журнала: 2025, Номер unknown

Опубликована: Фев. 27, 2025

Abstract We present a data fusion model designed to address the problem of sparse observational by incorporating numerical forecast models as an additional source improve predictions key variables. This is applied two main meteorological sources in Philippines. The approach assumes that different are imperfect representations common underlying process. Observations from weather stations follow classical error model, while forecasts involve both constant multiplicative bias and additive bias, which spatially structured time-varying. To perform inference, we use Bayesian averaging technique combined with integrated nested Laplace approximation. model’s performance evaluated through simulation study, where it consistently results better more accurate parameter estimates than using only or regression calibration, particularly cases data. In application, proposed also outperforms these benchmark approaches, demonstrated leave-group-out cross-validation.

Язык: Английский

Процитировано

0