Abstract
A study of systematic errors in the 24 and 48 hour forecasts of cyclones by the currently operational NMC LFM-II model has been completed for the 1978–79 winter season (1 October 1978–30 April 1979). All available LFM-II 0000 and 1200 GMT forecast cycles are verified over the North American continent and adjacent oceans.
For each cyclone, defined to be a low pressure region with one or more closed isobars analyzed at 4 mb intervals on the LFM-II forecast and verification charts, the following information is tabulated: 1) latitude and longitude of the cyclone center, 2) central pressure and 3) 1000–500 mb thickness over the cyclone center. Errors in forecast central pressure, thickness and position are then calculated.
The LFM-II generally underforecasts the intensity of oceanic cyclones. Overforecasting of continental cyclones occurs to the lee of major mountain barriers and eastward of 95°W to the Great Lakes with the exception of the southwestern United States. Corresponding model 1000–500 mb thickness forecasts are too cold over the ocean basins and the western and southeastern United States. A warm thickness error is observed over the remainder of the continent, particularly northern Canada. Finally, model cyclones are forecast too slow in October and November and too fast in March and April.
Examples are shown which illustrate both typical and atypical error patterns and the possible consequences of initial analysis errors in the Pacific on forecast quality. While a useful signal appears to be present in our findings, caution is advised in their interpretation over the western United States where Pacific analysis uncertainties are undoubtedly a source of contamination.
