Analog ensemble data assimilation and a method for constructing analogs with variational autoencoders
This work addresses data assimilation challenges in fields like weather forecasting by introducing incremental improvements through machine learning techniques for ensemble generation.
The paper tackled the problem of generating ensemble perturbations for data assimilation methods like EnOI and EnVar by proposing analog-based approaches, including a new method using variational autoencoders (VAEs) to construct analogs. In tests on a multiscale Lorenz-96 model, the analog methods improved EnOI performance, with the VAE-based method performing as well as a full ensemble square root filter and showing robustness across tuning parameters.
It is proposed to use analogs of the forecast mean to generate an ensemble of perturbations for use in ensemble optimal interpolation (EnOI) or ensemble variational (EnVar) methods. A new method of constructing analogs using variational autoencoders (VAEs; a machine learning method) is proposed. The resulting analog methods using analogs from a catalog (AnEnOI), and using constructed analogs (cAnEnOI), are tested in the context of a multiscale Lorenz-`96 model, with standard EnOI and an ensemble square root filter for comparison. The use of analogs from a modestly-sized catalog is shown to improve the performance of EnOI, with limited marginal improvements resulting from increases in the catalog size. The method using constructed analogs (cAnEnOI) is found to perform as well as a full ensemble square root filter, and to be robust over a wide range of tuning parameters.