Abstract
Observing systems simulation experiments were carried out to estimate the accuracy of temperatures diagnosed from the divergence equation when an army of nearly continuous (in time) wind observations is available. It was found that a useful estimate of temperature can be derived from high-resolution wind observations such as those obtainable from a network of wind profiling systems. Adding the divergence and vertical motion terms to the balance equation to form the complete divergence equation reduces the errors in derived temperatures and geopotential heights. Observations on an irregularly spaced grid lead to greater errors than those on a regularly spaced grid. Moderate errors are also introduced when large-scale errors in geopotential occur in the lateral boundary conditions. This suggests the need for some independent observations of temperature (from rawinsonde or temperature profiler) to prescribe the boundary conditions for the retrieval technique.
In a simulation of a possible operational system in which wind observations with random errors of 1 m s−1 are available on a 350 km grid and boundary values of geopotential height contain errors typical of a 12 h model forecast, the derived temperatures and heights on the interior of the grid contain root-mean-square errors of 1.55°C and 18.8 m, respectively.
