Satellite Data Assimilation Using NASA Data Systems Test 6 Observations

Abstract

Two assimilation schemes are described in which continuous indirect insertion of satellite-derived temperatures is performed, using a global primitive equation forecast model. Both schemes employ a relatively simple indirect insertion technique but utilize different methods (Noise Freezing Methods I and II) to control the noise induced within the forecast model by data insertion. Using data collected in February 1976 during NASA Data Systems Test 6, the effectiveness of these schemes is compared with that of a third scheme in which satellite-derived temperatures are assimilated using the same techniques that most operational forecast centers employ. After a 36 h start-up period, 48 h forecasts were produced using each assimilation scheme and root-mean-square errors computed for the differences between the forecast fields and upper-air observations of geopotential height. The forecasts of geopotential height made using the noise freezing methods are found to show substantial improvements over those made using conventional techniques. The forecasts produced using Noise Freezing Methods I and II are comparable with each other, and show average percent improvements over conventional forecasts ranging from 5 to 10% at 850 mb and from 10 to 15% at both 500 and 300 mb. Improvements of nearly 25% are observed for individual forecasts.