The 4D-variational data assimilation procedure (4D-var DA) is a special routine used for correction of climate/weather forecasts by tuning the climate model parameters in a way that provides the best possible fit to the available observational data. Due to a number of reasons DA introduces its own inevitable methodological error which ultimately affects the accuracy of the model forecast. The existing methods designed for the reduction of this uncertainty require a lot of computational resources. This is the reason why their usage in many climate models is restricted by some simplified versions. This project is aimed on developing a conceptually novel, robust, and efficient nonlinear variational error estimation algorithm (NOVEEA) which can estimate the inaccuracy of the DA methods and can make the corresponding corrections quite efficient computationally. The advantage of the proposed method is that the computational algorithm is based on a abstract mathematical 4D-var DA problem which allows using it in a wider geophysical context.
Mathematik, Informatik und Naturwissenschaften
Dr. Andrey Vlasenko