Title:Data-driven stochastic Lie transport modelling of the 2D Euler equations

Abstract:Stochastic modelling of coarse-grid SPDEs of the two-dimensional Euler equations, in the framework of Stochastic Advection by Lie Transport (SALT) [Cotter et al, 2019], is considered. We propose and assess a number of models as stochastic forcing. The latter is decomposed in terms of a deterministic basis (empirical orthogonal functions) multiplied by temporal traces, here regarded as stochastic processes. In particular, we construct the stochastic forcing from the probability density functions (pdfs) and the correlation times obtained from a fine-grid data set. We perform uncertainty quantification tests to compare the different stochastic models. In particular, comparison to Gaussian noise, in terms of ensemble mean and ensemble spread, is conducted. Reduced uncertainty is observed for the developed models. On short timescales, such as those used for data assimilation [Cotter et al, 2020a], the former models show a reduced ensemble mean error and a reduced spread. Estimating the pdfs yielded stochastic ensembles which rarely failed to capture the reference solution on short timescales, as is demonstrated by rank histograms. Overall, introducing correlation into the stochastic models improves the quality of the coarse-grid predictions with respect to white noise.