During the 1950s, Lorenz became skeptical of the appropriateness of the linear statistical models in meteorology, as most atmospheric phenomena involved in weather forecasting are non-linear. Modern numerical weather prediction has answered  Lorenz’ challenge through combining sophisticated physics based numerical models with increasingly accurate high-dimensional data.  This process is called data assimilation and it is performed every day at all major operational centres across the world. Data assimilation  (DA) requires massive computing capabilities as realistic atmosphere-ocean models typically have billions of degrees of freedom. The objective of the ongoing research that I am involved (see details at https://www.imperial.ac.uk/ocean-dynamics-synergy/ ) is to drastically decrease the required DA computational effort by reducing the dimension of the models involved and use stochastic perturbations to account for the unresolved scales. The incorporation of observation data is done using particle filters suitably adapted to solve high-dimensional problems.

This work is part of the  Mathematics for Planet Earth  programme which the speaker will introduce briefly.