Meteorological data assimilation techniques are motivated forward by the advance of numerical weather prediction models and the increasing rapidly observations, including the great part of uncon-ventional data obtained by remote measurement methods. There are mainly two general concepts that have been discussed repeatedly for data assimilation in meteorology. The variational (especially adjoint variational) method has been the popular and most used scheme, which, however, has a drawback that model errors (system noise) are not taken into account. Another class of methods are those described as sequential data assimilation, which are represented by Kalman filters. The introduction of Kalman filters and their developmental stages in the meteorological data assimilation field are presented in this paper, as well as that the importance and applications of Kalman filters, particularly simplified Kalman filters and ensemble Kalman filters. Due to that they have the ability to consider model errors and let assimilation results not drift away from observations, Kalman filters are paid more and more attentions, though they need
much of computational load. Compared with the current advance abroad, the developments and applications of Kalman filters in China are lagged. However, there will be a bright prospect for them with the improvements of computational conditions.