Data Assimilation with a One-Level, Primitive Equation Spectral Model

Abstract

Techniques for inserting data into spectral models have been developed and are shown to be capable of assimilating data. Real 500 mb winds and heights are inserted into a free-surface spectral model at locations which simulate the global observing network. A series of 6-day assimilation experiments are undertaken which are designed to test the effect of a semi-implicit time-stepping algorithm, different frequencies of data insertion, and various adjustment mechanisms on the quality of the resulting analyses. Scrutiny of these experiments, in both real and spectral space, gives insight into the model's response to inserted data and suggests methods by which the assimilation may he improved. Some methods, which are particularly suited to spectral models, are subjected to testing and found to be effective.