Abstract
Data assimilation is a relevant framework to merge a dynamical model with noisy observations. When various models are in competition, the question is to find the model that best matches the observations. This matching can be measured by using the model evidence, defined by the likelihood of the observations given the model. This study explores the performance of model selection based on model evidence computed using data-driven data assimilation, where dynamical models are emulated using machine learning methods. In this work, the methodology is tested with the three-variable Lorenz' model and with an intermediate complexity atmospheric general circulation model (a.k.a. the SPEEDY model). Numerical experiments show that the data-driven implementation of the model selection algorithm performs as well as the one that uses the dynamical model. The technique is able of selecting the best model among a set of possible models and also to characterize the spatio-temporal variability of the model sensitivity. Moreover, the technique is sensitive to differences in the model dynamics which are not reflected in the moments of the climatological probability distribution of the state variables. This suggests the implementation of this technique using available long-term observations and model simulations.
Highlights
Data assimilation (DA) methods aim to provide the best estimation of the state of a dynamical system based on a set of noisy and partial observations
Model selection using data assimilation has been introduced in Carrassi et al (2017), applied in the context of climate change 375 detection and attribution in Hannart et al (2016), and to complex or large state model evaluation in Metref et al (2019)
It consists in putting in competition two or more dynamical models and use observations to compute a likelihood, called Contextual Model Evidence (CME), for each model, attributing the highest probability to the model which provides forecasts that better match the observations over a period of time
Summary
Data assimilation (DA) methods aim to provide the best estimation of the state of a dynamical system based on a set of noisy and partial observations (see Carrassi et al, 2018; Reich, 2019; Van Leeuwen et al, 2019, and references therein). One application that has recently received increasing attention is the use of data assimilation methods for model optimization and model selection. The former is concerned with obtaining better estimates for model parameters and configuration with the ultimate goal of quantifying and reducing model error and dispersion in various applications (e.g., Schirber et al, 2013; Ruiz et al, 2013; Ruiz and Pulido, 2015; Lauvaux et al, 2019; Kotsuki et al, 2020).
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call