Development of large-scale data assimilation techniques

Data assimilation (DA) is a computational technique for improving observation and numerical models, estimating unobservables, and enhancing future forecasts by combining limited observational data and simulation models based on Bayesian statistics. DA has been developed mainly in the meteorological and oceanic fields and is indispensable to today's weather forecasting. However, DA itself is a general mathematical theory and can be applied to a wide range of problems, not only in weather forecasting.

In DA, we need to evaluate the "posterior distribution" of model parameters and variables, but accurately evaluating the posterior distribution is a hard task because it requires exponentially increasing computations due to the "curse of dimensionality." We are interested in the mathematical theory of large-scale DA techniques and explore the applicability of DA through various problems.