Recent developments in hydrologic data assimilation

Data assimilation is a term that is most closely associated with the atmospheric sciences. It arises from the meteorological custom of constructing daily weather maps which show how environmental variables such as pressure and wind velocity vary over space (see Daley [1991] for a good historical review). Such maps are useful both for short‐term forecasting, where they provide initial conditions for numerical weather prediction models, and for more long‐term climatic analysis, where they can be used to reveal regional trends [ Bengtson and Shukla , 1988; National Research Council, 1991; Shubert et al, 1993]. For many years weather maps were derived from a limited number of measurements collected from surface stations and radiosondes. The mapping process required a good understanding of atmospheric physics and was typically carried out by hand, using the methods of ‘subjective analysis’. In the first half of the twentieth century efforts were made to standardize and automate meteorological mapping. The quantitative procedures which grew out of these efforts were commonly called ‘objective analysis’ methods [ Gandin , 1963]. Objective analysis algorithms provide spatially distributed descriptions of atmospheric conditions which are consistent both with observed data and with kinematic (‘diagnostic’) principles (such as the geostrophic requirement that the pressure gradient and Coriolis force must balance). These algorithms produce ‘snapshots’ portraying conditions everywhere in a given region at a particular time.