Meteorology

In the field of data assimilation, the three dimensional variational analysis and its extension to four dimensions is a task of high priority. An alternative co-ordinate system based on the conservation of spin, is being introduced to variational analysis. Assimilation of surface data is based on traditional methods, i.e. optimum interpolation. Remotely sensed observations of e.g. snow cover, sea ice and sea surface temperature are combined with in situ observations.

The research on numerical modelling is performed in regional climate simulations, operational numerical weather prediction models and high resolution, non-hydrostatic models. The research focuses on well posed boundary conditions, parameterisations of precipitation processes and exchange of latent and sensible heat between the surface and the atmosphere. The possibility of simulating local winds, turbulence and air pollution by using a non-hydrostatic model is investigated.

Statistical methods include the computation of error statistics for numerical prognoses, correction of prognoses and statistical modelling of different forecasted or observed quantities such as lightning and precipitation amounts. The methods assume that the observations are realisations of random variables, while the selected output from the numerical models and possibly other information is included as explanatory variables. Classical and modern techniques within regression, classification and Kalman filters are used. One great advantage of using a statistical model is that the forecasts can be formulated in probabilistic terms.

Massive parallel computation is used for data assimilation and numerical atmospheric modelling. The aim of this research is to utilise the capacity of the computers in the most effective way.