A series of simulation studies has been conducted in an effort to obtain information relevant to the planning of the First GARP Global Experiment. Thus far, the studies have used only the Mintz-Arakawa 1969 model, and have been concerned mainly with the utilization of IR vertical sounding measurements. The initial results indicate that temperature profiles derived from these measurements can play a valuable role, provided they are used on a continuing, day-to-day basis over an extended period. Temperature data used in this way appear to have a controlling influence on all other meteorological variables in the model, including winds and pressure in particular. Assuming a mean error of 1C in the temperature data, and assuming the coverage provided by the planned GARP satellite configuration of two polar orbiting satellites containing IR sounders with full side-scan capability, the experiments indicate that winds are determined within an error of 2 m sec−1 and pressures within an error of 2 mb. The wind and pressure determinations are significantly improved if (i) IR sounders are added to the geostationary satellites, (ii) the number of polar orbiting satellites is increased, or (iii) the errors in the temperature data are decreased. If the side-scan capability is reduced, the wind and pressure determinations are substantially worsened, and may fail entirely. The validity of these results is limited by (i) the defects of the model, especially in the tropics, (ii) the use of simulated data in place of real observations, and (iii) the fact that the experiments use a comparison with the solutions to the model as a test of the accuracy of the results, in place of a comparison with actual observations of winds and pressures.
A second series of experiments has been concerned with the relationship between the error limits specified in the global observations and the accuracy of the forecasts that will be based on these observations. The results indicate that with the present GARP data specifications, i.e., ± 3 m sec−1 in winds, ± 1C in temperature, and ± 3 mb in pressure, the forecasts begin to deteriorate on the 5th day and are misleading in major respects after the 8th day. The limit of deterministic predictability is reached in three weeks in agreement with the results of other studies. Further experiments indicate that the accuracy of the forecasts is limited primarily by the errors in the wind components, and secondarily by pressure errors. If the error limits are tightened to 1.5 m sec−1 in winds and to 2 mb in pressure, the forecasts are accurate in all major respects for 7 or 8 days. In order to secure accurate forecasts for 12 to 14 days, the error limits must be tightened to 0.5 m sec−1, 0.5 mb and 0.5C. In all cases, the limit of deterministic predictability is two or three times greater than the range of accurate forecasts.
1 Institute for Space Studies, Goddard Space Flight Center, NASA, N. Y., N. Y.
2 Computer Applications, Inc., N. Y., N. Y.