On the use of satellite estimates of precipitation in initial analyses for numerical weather prediction

Open Access
Publisher Website
Google Scholar
Abstract
A 4‐dimensional data assimilation procedure to incorporate satellite estimates of precipitation into the initialization of operational numerical weather prediction models by improving the “first guess” fields for the next analysis time is described. The method is tested using a limited‐area 8‐level primitive equations model including input of sensible and latent heat from the ocean surface, both parameterized convective and large‐scale precipitation and release of latent heat, surface frictional drag and orography. The grid size is 190 km at 60°N. Precipitation is estimated from satellite infrared photographs using an adaptation of the method of Richards and Arkin (1981). This relates the area‐average precipitation rate to the fraction of the area covered by clouds whose cloud‐top temperature is less than a given value. The latent heat associated with the estimated precipitation replaces that computed by the model. It is distributed in the vertical and “fed back” via the thermodynamic equation into the system of prognostic equations. This induces vertical motion, low‐level convergence and cyclogenesis. Thus the satellite‐estimated precipitation in effect “guides” the sea‐level low and “discourages” it from straying too far from where the satellite data indicates it should be. To test the effectiveness of the technique, prognoses are produced for five 24‐h periods in the development of two intense North Pacific cyclones. Normally, prognoses would be made only for the length of time between analysis cycles (e.g. 6 h). Predictions are made both with and without satellite estimates of precipitation. The inclusion of satellite data results in a significant improvement at 1000 mb, lowering the average 24‐h mean error from 38 to 11 m, the standard error from 57 to 41 m, and S1 score from 57 to 43. Inclusion of satellite data accelerates the low and frequently deepens it.