An Analysis of Sea-Level Cyclone Errors in NMC's Nested Grid Model (NGM) During the 1987-88 Winter Season

Abstract

A study of sea-level cyclone errors which occurred in 24- and 48-h forecasts of the National Meteorological Center's nested grid model (NGM) is performed for the 1987–88 winter season (1 December 1987–31 March 1988). All available 0000 UTC and 1200 UTC forecast cycles are analyzed for North America and adjacent ocean regions. Errors in forecasted central pressure and position are computed. NGM forecasts of cyclone central pressure average 0.6 mb too deep at 24 hours and 0.3 mb too deep at 48 hours. The root-mean-square (RMS) errors for central pressure are 5.7 mb at 24 hours and 7.9 mb at 48 hours. The mean systematic displacement errors are 29 km at 24 hours and 51 km at 48 hours, and are directed towards the west at both times. The mean absolute displacement errors are much larger, 268 km at 24 hours and 393 km at 48 hours. Cyclone movement is forecasted too slow more frequently than too fast. Large variability in the skill of successive runs characterizes NGM cyclone forecasts though, with ∼70%–∼75% of the temporal variance being associated with fluctuations having periods shorter than seven days. The results for the 1987–88 NGM are compared to those for the limited-area fine-mesh model (LFM) for the 1978–79 winter season. The size of the systematic pressure error decreased 50%–75% over the past decade. The absolute displacement error only decreased 10%–15%. Forecast variability, as measured by the RMS error of central pressure, is ∼15% smaller at 48 hours but remains essentially the same at 24 hours. The comparison of the NGM and LFM results suggests that the nature of the short-range forecast problem for wintertime extratropical cyclones has changed somewhat over the past decade. It now appears that the problem is no longer one of primarily reducing the systematic error but also involves minimizing the impact of variability among individual forecasts.