Skill of NMC Operational Dynamical Models in Prediction of Explosive Cyclogenesis

Abstract

The skill of the Nested-Grid Model (NGM) and the global spectral model (GLBL) at the National Meterological Center in the prediction of explosive cyclogenesis was evaluated for the period 1 September 1986?30 April 1987. Manual analyses covering the eastern North Pacific, North America and the North Atlantic eastward to 20°W were used as ground truth. The criterion for a bomb event in the analyses of the forcasts was a deepening of the center of at least 24 mb at 60°N, normalized geostrophically at other latitudes, in a period of 24 h, beginning at 0000 or 1200 UTC. Both models displayed skill out to 48 h for the NGM and 60 h for the GLBL. The NGM performed notably better in the innermost fine-grid area than in the surrounding area of overlap with a more coarse grid. For the GLBL in the Atlantic and North America, similar skill was seen through 36 h; skill was very small in the Pacific region. 12-h deepening beginning 12 h after initialization was compared with analyzed deepening for both models. Correlations ranged from 0.72 for the NGM in the inner grid over the Atlantic and North America to 0.03 in the Pacific. The GLBL values were intermediate, again better in the Atlantic than in the Pacific. All samples showed an average shortfall of predicted deepening from 12?24 h after initialization, ranging from 1 mb for the inner NGM grid to 7 mb for the overlap area, with the GLBL intermediate; again, it was much better in the Atlantic than in the Pacific. Growth of skill over the past few years is attributable to improved analyses, better model resolution and better treatment of bounndary-layer fluxes. Initial data limitations are now the most important factor, both in models and in verifying analyses. These results alter the nature of the problem of research on explosive cyclogenesis from one of discovering a missing ingredient to one of improving the performance and extending the range of predictability.