Statistical-dynamical intraseasonal prediction of tropical cyclogenesis in the western North Pacific

We have developed a statistical-dynamical prediction system for forecasting the probability of tropical cyclone (TC) formation at a weekly, 2.5° horizontal resolution across the western North Pacific (WNP) at intraseasonal lead times. We use five large scale environmental factors (LSEFs) to represent the favorability of the climate system to support tropical cyclogenesis. The LSEFs are: low level relative vorticity, sea surface temperature, vertical wind shear, planetary vorticity, and upper level divergence. We used logistic regression to develop a statistical model representing the probability of TC formation at each grid point based on the LSEFs. Data for the model development was obtained for: (a) the LSEFs from the NCEP reanalysis data sets; and (b) the corresponding TC activity from the Joint Typhoon Warning Center best track data set. Verification of independent zero-lead hindcasts shows the model has skill and potential value to risk adverse customers (e.g., positive Brier skill scores, skillful relative operating characteristic skill scores). We have used the fully coupled, one-tier NCEP Climate Forecast System to obtain forecasts of the LSEFs at lead times of 1-12 weeks, which we then used to force the regression model. We have conducted non-zero lead hindcasts for 1982-2008, and forecasts for the 2009 WNP TC season. Our assessments of the skill of these hindcasts and forecasts provide encouraging indications of the predictive potential of our statistical-dynamical system. We have also developed a corresponding system for intraseasonal forecasting of North Atlantic TC formations with similarly positive results.