• Skill

    The accuracy of a particular prediction will vary depending on how long before the start of the season it was generated. Generally, the closer to the start of the hurricane season a prediction is produced, the higher its accuracy will be. For example, a prediction issued in August will have, in general, higher accuracy than a prediction issued in April.

    Below, the real-time skill of seasonal forecast schemes used by different forecasters for over one decade is shown. Forecasts displayed here are issued by the Colorado State University (CSU), the National Oceanic and Atmospheric Administration (NOAA) and Tropical Storm Risk (TSR). The period compared in this analysis is 2003-2015, which is a time period when all three forecast agencies were utilizing relatively similar techniques for issuing these predictions.

    Select an image

    Mean Squared Skill Scores for three different forecast agencies for a 50-year climatology mean (1951-2000) and for the previous 10-year mean. Note the improvement in the skill as the peak of the Atlantic hurricane season approaches. A score of 100 represents a perfect prediction while a score of 0 or less means that the forecasting technique offers no improvement with respect to a climatological forecast.


    Rank correlations for three different forecasting agencies related to four different parameters: accumulated cyclone energy, major hurricanes, hurricanes and named storms. Note that the skill of the models applied by each of the agencies improves as the peak of the Atlantic hurricane season approaches. The Tropical Storm Risk group shows the best real-time from 2003-2015 for accumulated cyclone energy.