On temperature initialization in primitive equation forecast models

Four procedures of specifying model initial temperature were described and tested in the present study. It was found that the use of observed temperatures along with a proper vertical interpolation scheme was not only acceptable, but produced less error than the use of temperatures derived from geopotential height through the hydrostatic equation did. Use of the difference form of the hydrostatic equation would produce unacceptable errors in the initial temperatures, unrealistic horizontal and vertical distribution of temperature, and these errors would influence the calculation of the pressure gradient force, resulting in substantial, artificial disturbances within the model domain. In addition, an approach to check the initial data was described. Taking advantage of the fact that the geostrophic wind in sigma coordinates should be nondivergent, geopotential height and temperature were used to calculate the pressure gradient force terms and an initial divergence of the geostrophic wind. This approach can be used for comparing different initialization schemes for identical input data.