Assimilation of satellite data in models for energy and water cycle research

Abstract

Numerical models provide a detailed, self-consistent, fairly accurate picture of atmospheric processes. If there are sufficient data, assimilation techniques can keep the model's representation in line with the actual evolving atmospheric state, enabling detailed diagnosis and understanding of atmospheric phenomena. Satellite data from temperature sounders, and planned doppler wind lidars, are most useful through model assimilation. Moisture in the models is controlled by the parametrisations of sub-grid-scale processes. The model reaches an equilibrium between evaporation and condensation, whose latent heating is an important driver of the dynamical circulations. Assimilation can force the model away from this equilibrium state, damaging the model's representation of atmospheric processes. Examples are given where incorrectly processed data have done this. Even with accurate data, assimilation is not the way to correct systematic deficiencies in the model. Rather, we learn about the model, and should be able to improve it and our understanding.