Abstract: Weather forecasting is formulated as an initial value problem in the work of numerical forecasting, while in the routine forecasting in the conventional way it is formulated as, an 0evolution' problem of weather process, especially that on certain specific level. It is shown in this paper that for large-scale motion these two different formulations are in fact equivalent under certain trivial conditions. It is also pointed out that for large-scale motion the three-dimensional structure of the baroclinic atmosphere at certain moment is reflected in and could be reconstructed from the evolution of the contour and temperature fields at certain level before and after that moment.