Initial values of horizontal and vertical wind components, temperature, pressure, and moisture are objectively determined for use in a boundary layer forecast model. The objective analysis is accomplished by use of Sasaki's (1969 a, b) numerical variation method. This technique incor-porates the governing dynamical and thermodynamical equations, as well as observations, into the initialization process. The solution to four coupled partial differential equations with asso-ciated boundary conditions completely determines the initial map. Richardson's relaxation method is used to solve the system of equations.
The initialization and prediction is applied to severe weather occurrence in the Midwest on 10 June 1968. A 1.5km thick layer bounded by the earth's surface and encompassing a horizon-tal area of approximately 2000×2000km2 is used. The horizontal grid spacing is 190km, half of the NMC (National Meteorological Center) grid, and the vertical interval is 200 m.
Two sources of data are used in conjunction with the governing prognostic equations to determine the initial fields: 1) wind and specific humidity observations from the radiosonde network, 2) pressure and temperature fields from the NMC analysis. The pressure and tem-perature data could also be extracted from the radiosonde records to obtain a more consistent set of data. The hybrid collection of information, however, helps reveal the versatility of the variational method. The degree of confidence in each data set is controlled by appropriate choice of weights in the variational formulation.
The areal distribution of a severe storm index, wq, the product of the vertical velocity and specific humidity, at the mid-level of this model is discussed. The initial distribution and the 3hr and 6hr forecasts of the index are compared with the surface observations of severe weather.