|
What is data assimilation?Essentially, data assimilation is the sophisticated combination of a numerical model (typically based on some system of partial differential equations) and observations. An optimal solution that “best” satis- fies the model and the observations is sought, where “best” is defined by the modeler (e.g., minimizing the errors associated with the model and the observations in some least squares sense). The obvious product of such an exercise is the optimal solution itself, which may be envisioned as either a more realistic model solution or an interpolation of the observations, depending upon perspective. Other products include: improved estimates of forcing fields and model parameters, testing of models as scientific hypotheses, assessment of redundancy in observing systems, resolution of mathematically illposed modeling problems, and estimates of model and observational errors.
|
 |