Statistical Postprocessing Methods for Operational Tropical Cyclone Forecasting (Invited Presentation)

Shortly after the National Hurricane Center (NHC) began issuing 24 hr quantitative forecasts of tropical cyclone (TC) position in 1954, statistical post-processing methods were developed to provide guidance for those forecasts. The initial methods focused on track forecasting and evolved from simple regression methods with predictors from climatology and persistence to more sophisticated statistical-dynamical models that combined input from NWP models and TC parameters. By the early 1990s, the statistical-dynamical track models were phased out because the NWP model track forecasts became much more accurate. However, statistical post-processing is still a valuable tool for combining track forecasts from multiple NWP models and ensemble systems. Statistical post-processing for TC intensity forecasts is following a similar evolution, but lagged by about 4 decades. The simplest methods were introduced in the 1980s and have become increasingly complex since that time. We are now entering the era where NWP model intensity forecasts are more skillful than statistical models, so the emphasis is shifting to using post-processing to combine forecasts from multiple models. Post-processing is also still valuable for specialized predictions such as TC formation, rapid intensification and structure change. The history statistical post-processing for TC forecasting and the outlook for the future will be presented.