A Case Study of Forecast Sensitivity to Data and Data Analysis Techniques

Abstract

In this study we examine the sensitivity of forecast to individual components of the First GARP (Global Atmospheric Research Programme) Global Experiment database as well as to some modifications in the data analysis techniques. Several short assimilation experiments (0000 GMT 18 January 1979 through 0000 21 January) are performed in order to test the effects of each database or analysis change. Forecasts are then generated from the initial conditions provided by these experiments. The 0000 21 January case is chosen for a detailed investigation because or the poor forecast skill obtained earlier over North America for that particular case. Specifically, we conduct experiments to test the sensitivity of forecast skill to: 1) the addition of individual satellite observing system components; 2) temperature data obtained with different satellite retrieval methods; and 3) the method of vertical interpolation between the mandatory pressure analysis levels and the model sigma levels. For the single case examined, TIROS-N infrared land retrievals produced operationally are found to degrade the forecast, while the use of TIROS-N retrievals produced with a physical inversion method as part of an analysis/forecast cycle results in an improved forecast. The use of oceanic VTPR (Vertical Temperature Profile Radiometer) satellite retrievals also results in an improved forecast over North America. The forecast is also found to be sensitive to the method of vertical interpolation between the mandatory pressure analysis levels and the model sigma levels.