Introduction to Data Assimilation

Atmospheric data assimilation is a class of techniques used for producing descriptions of fields of air temperature, pressure, humidity, wind, etc. on a spatial grid or in terms of a finite functional representation. These are then used to initialize numerical weather forecasts or to analyze the atmosphere for other purposes. The techniques combine past, present, and even future observations in an approximate statistically optimal way. Various types of statistical or physically-based models and their corresponding adjoints are employed to relate diverse fields in both time and space and to relate what is observed to what is being analyzed. Computationally, the problem is very demanding and onstraining on the techniques that can be employed on a routine basis.