2017/09/19 by Sammy Metref, Alexis Hannart, Metref, Sammy +9 · 1 citation
Earth and Planetary Sciences · Environmental Science · #Atmospheric and Environmental Gas Dynamics #Climate variability and models #Computation (stat.CO) #FOS: Computer and information sciences #Meteorological Phenomena and Simulations #Methodology (stat.ME)
paper · pdf · doi:10.48550/arxiv.1709.06635
openalex publication_date 2017/09/19 · openalex created_date 2022/10/01 · openalex updated_date 2026/07/28
IIn recent years, there has been a growing interest in applying data\nassimilation (DA) methods, originally designed for state estimation, to the\nmodel selection problem. In this setting, Carrassi et al. (2017) introduced the\ncontextual formulation of model evidence (CME) and showed that CME can be\nefficiently computed using a hierarchy of ensemble-based DA procedures.\nAlthough Carrassi et al. (2017) analyzed the DA methods most commonly used for\noperational atmospheric and oceanic prediction worldwide, they did not study\nthese methods in conjunction with localization to a specific domain. Yet any\napplication of ensemble DA methods to realistic geophysical models requires the\nimplementation of some form of localization. The present study extends the\ntheory for estimating CME to ensemble DA methods with domain localization. The\ndomain-localized CME (DL-CME) developed herein is tested for model selection\nwith two models: (i) the Lorenz 40-variable mid-latitude atmospheric dynamics\nmodel (L95); and (ii) the simplified global atmospheric SPEEDY model. The CME\nis compared to the root-mean-square-error (RMSE) as a metric for model\nselection. The experiments show that CME improves systematically over the RMSE,\nand that this skill improvement is further enhanced by applying localization in\nthe estimate of the CME, using the DL-CME. The potential use and range of\napplications of the CME and DL-CME as a model selection metric are also\ndiscussed.\n