The Initial Structure and Resulting Error Growth in the NCAR GCM Produced by Simulated, Remotely Sensed Temperature Profiles

Abstract

The error produced by an observational system of remotely sensed temperature profiles is partially simulated with aid of an operational retrieval scheme. A given temperature distribution is converted to radiances and then back to temperatures. A comparison is made between the retrieved temperatures and the original values. The sensitivity of the retrieval scheme to various input parameters, such as guess profiles, the statistical coefficients used in the retrieval scheme, and cloudiness, is examined. Experiments with the placement of a reference level from which to integrate the geopotential field are performed. The relative growth rate of the simulated initial error is examined by forecasting with two initial states, one case with the observational error and one central case. Error growths are calculated for different reference levels, clear, and cloudy cases. The results show that the generated error fields do not seriously contaminate forecasts of the large-scale baroclinic waves for periods up to one week, providing the retrievals are free of the effects of clouds and an accurate guess is used. In general, the results support the need for reference-level information, and if forecast error in the low troposphere is to be a minimum, placing of the reference level at sea level.