Open AccessAssessment of Western North Pacific 96- and 120-h Track Guidance and Present Forecastability
Author(s) -
Kathryn Anne Payne,
Russell L. Elsberry,
Mark A. Boothe
Publication year - 2007
Publication title -
weather and forecasting
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.393
H-Index - 106
eISSN - 1520-0434
pISSN - 0882-8156
DOI - 10.1175/waf1030.1
Subject(s) - forecast error , meteorology , middle latitudes , typhoon , climatology , global forecast system , environmental science , forecast verification , spurious relationship , consensus forecast , tropical cyclone , track (disk drive) , tropical cyclone forecast model , computer science , forecast skill , numerical weather prediction , geology , econometrics , geography , mathematics , machine learning , operating system
Because the Joint Typhoon Warning Center (JTWC) has only four dynamical models for guidance in making 96- and 120-h track forecasts, an opportunity exists for improving the consensus forecast by the proper removal of a likely erroneous forecast to form a selective consensus (SCON). Forecast fields from all four models [the U.S. Navy Operational Global Atmospheric Prediction System (NOGAPS), the U.S. Navy version of the Geophysical Fluid Dynamics Laboratory model (GFDN), the Met Office (UKMO) model, and the Global Forecast System (GFS)] were available during the 2005 western North Pacific season to evaluate for the first time the error mechanisms leading to large track errors. As shown previously for the NOGAPS and GFDN models during the 2004 season, error sources related to the midlatitude circulations accounted for about 90% of all large 120-h track errors by all four models during the 2005 season. This dominance of midlatitude-related error source is a major shift from the 72-h errors, which include more errors related to tropical circulations. In the GFS model, 95% of the large errors occurred because of an incorrect depiction of the vertical structure of the tropical cyclone. A systematic error in the GFDN model was identified in which a false anticyclogenesis was predicted downstream of the Tibetan Plateau, which accounted for over 50% of the large GFDN track errors. The consensus spread versus consensus error relationship is examined to isolate those 20%–25% of cases with large spreads and large errors that are candidates for forming an SCON. If the model tracks that contributed to the large errors are eliminated, the average improvement of the SCON forecasts relative to the nonselective consensus is 222 (239) n mi during 2005 (2004), and the corresponding average improvement relative to the JTWC forecasts is 382 (203) n mi. This application of SCON is considered the potential “forecastability” in that it represents the optimum use of the present numerical guidance for consensus forecasting.