Data assimilation is the use of observations in computer models in order to improve the reliability of the outcome of the models. An important application is found in meteorology. There, a prediction model starts by simulating the past few days and during that period, it is constantly adjusted to stay close to the available observations. When the model then goes on to simulate the next few days, the initial situation of the model will closely match the current reality and the predictions are bound to be more reliable than without data-assimilation. But there are also other ways to meaningfully combine observations and models: for calibration, for interpolating or extrapolating observations or to compute back from an observation to the likely cause of that observation.
Each simulation model is actually nothing more than an approximate description of reality. This means that there will always be differences between a model and reality. By combining models and observations, this gap can be reduced. This opens the way to all kinds of interesting applications, like using models to complement observations or to better understand observed values. Read more>>
There is a wealth of methods for data-assimilation. Giving a complete overview is hardly possible and certainly beyond the scope of this web site. Therefore, we will only give an introduction to the two main classes of methods. Read more>>
Data assimilation and calibration are used in a rich variety of applications. In all these cases, the combination of models and observations allows results that would be impossible with either models or observations alone. Below, a number of remarkable or important applications are presented. VORtech has contributed to several such applications. Read more>>
On the Internet, a lot of information can be found about data-assimilation and its applications. A number of sites is particularly interesting to check out. Read more>>