Forecast-errors in the 24- and 48-hour barotropic 500-mb prognostic charts prepared by JNWPU for the winter of 1957 were investigated. Certain large-scale forecast-errors were found to persist from day-to-day in fairly localized geographical areas. In general, the numerical prognoses exhibited a tendency to forecast excessively high values off the southeastern coasts and excessively low values off the northwestern coasts of the continents. Little forecast-error was observed in the continental interiors.

A Fourier analysis revealed that the forecast-errors were largely due to incorrect phase forecasts for wave numbers 1 through 4.

The forecast-errors associated with wave number 1 usually “positioned” the large-scale systematic forecast-errors near Japan and Scandinavia and contributed materially to their intensities.

Relationships between forecast-error fields and topography, geography, synoptic situation, and “non-adiabatic” heating are discussed.

Hemispheric fields of “non-adiabatic” heating, computed using a two-level graphical model, are shown.

Charts illustrating the dependence of “non-adiabatic” heating on the general circulation are also shown.