Abstract
To the extent that the stratosphere wind field is close to geostrophic, the thermal wind is a good approximation to the vertical wind shear (vertical variation of the horizontal wind). And since the thermal wind is proportional to the horizontal temperature gradient, the possibility exists of determining it from satellite radiance observations. Several different methods are developed here for retrieving thermal winds directly from the horizontal gradients of satellite radiance observations, without first retrieving the horizontal temperature gradient. The methods are applied to the determination of thermal winds in the upper troposphere and lower stratosphere over the White Sands Missile Range area. A special series of about 30 concurrent sets of radiance observations from the NDAA-4 VTPR instrument and wind shears from radiosonde observations (for ground truth) distributed throughout one year, is used for these tests. The results obtained with these direct methods are compared with results obtained with 1) a traditional method, in which temperature profiles are first retrieved from the satellite radiances and the thermal winds are then obtained from the horizontal gradients of the retrieved temperatures; and 2) a linear regression between observed radiance gradients and observed wind shears. The latter method serves as an estimate of the upper limit of accuracy to be obtained by any method based on a linear combination of radiance gradients.
The results indicate that the direct methods may be divided into two groups, with much better retrievals for one of these groups. The probable reasons for these differences are identified. The best direct methods yield results comparable to the traditional method. In comparison with ground truth none of the methods is particularly skillful. The lack of skill in these particular cases is attributed mainly to the modest wind shears contained in the sample. Errors associated with trying to measure relatively small horizontal radiance gradients over relatively small horizontal distances result in residual uncertainty nearly as large as the variance of the sample. it is suggested that much better results would be obtained if some of the better methods were to be applied over greater horizontal distances or to regions with larger wind shears.
