Being able to make a forecast for the weather in the coming week or more saves us billions of euros every year and helps protect lives and property. Over the last years, a rapid increase in computing power and new observations have led to a continuous improvement of predictability, although individual forecasts can be surprisingly inaccurate at times. The biggest challenge lies in identifying the limits of predictability and to find the physically best possible forecast. At the KU, Prof. Dr. Tijana Janjic is contributing to this endeavor with her research on data assimilation.
In a sense, we are all doing data assimilation every day: If you want to cross a street, you need information on the speed of approaching cars, so you will watch them for a moment. Coupled with the experience of what the average speed is, we then estimate if it is possible to cross the street or whether we should wait for the cars to pass. This can lead to errors in judgment if our observations were not correct or incomplete or the “prediction model” for the average driver did not fit. In meteorology, errors in the initial condition or in the forecast model are also the most common causes of incorrect forecasts.
In order to more accurately predict weather extremes or the melting of arctic ice, information in the form of heterogenous data must be combined with numeric models of dynamic systems. This is done by data assimilation enabling a closer analysis of processes and forecast of their trends. For this, a predictive model is continuously linked with observational data to achieve the most accurate analysis of the atmosphere.
In the field of data assimilation, Prof. Janjic is concerned with the further development of data science algorithms by incorporating physical conservation laws and solving from optimization problems in the environmental sciences. The quantification of uncertainties of predictions, numeric models and observations also play a key role. Professor Janjic is also an associate editor of the Journal of Advances in Modeling Earth Systems (JAMES) and of the Quarterly Journal of the Royal Meteorological Society (QJRMS).