North Atlantic Seasonal Hurricane Prediction: Underlying Science And An Evaluation Of Statistical Models

Statistically based seasonal hurricane outlooks for the North Atlantic were initiated by Colorado State University (CSU) in 1984, and have been issued every year since that time by CSU. The National Oceanic and Atmospheric Administration (NOAA) Climate Prediction Center and the UK-based Tropical Storm Risk (TSR) have the next longest records (1998-present) of continuous outlooks. This chapter describes how these three forecasts have evolved with time, and documents the approaches, the environmental fields, and the lead times which underpin the models' operation. Some of the environmental parameters used in early seasonal outlooks are no longer employed, but new predictive fields have been found that appear to be more important for seasonal hurricane prediction. An assessment is made of the deterministic skill of the seasonal hurricane outlooks issued in real time by CSU, NOAA, and TSR between 2003 and 2014. All methods show moderate-to-good skill for early August outlooks (prior to the most active portion of the hurricane season), low-to-moderate skill for outlooks issued in early June, and lesser skill for outlooks issued in early April. Overall, the TSR model has the most skillful predictions of Accumulated Cyclone Energy (ACE), while NOAA has the best named storm predictions issued in early August.