Post Mortem 2016

The 2016 Atlantic basin hurricane season officially ended on November 30. The season was characterized by overall slightly above-average hurricane activity with a total of 15 named storms, 7 hurricanes, 3 major hurricanes and an Accumulated Cyclone Energy of 135. Final ACE values may change slightly in post-season reanalysis of tropical cyclone intensity by the National Hurricane Center. The ensemble mean forecast called for 15 named storms (-), 8 hurricanes (+1), 3 major hurricanes (-) and an ACE of 108 (-20%). For comparison, the 1981-2010 median values of these quantities are 12 named storms, 6.5 hurricanes, 2 major hurricanes and an ACE of 92.

The aggregate forecast was relatively successful, accurately predicting the total number of storms and major hurricanes, but slightly overestimating the number of hurricanes while at the same time underestimating the ACE. The average forecast outperformed a forecast based on climatology.

Seventeen groups submitted forecasts between early March and early August. It should be noted that not all forecast groups issue predictions for all four tropical cyclone intensity metrics. The number of forecasts that was submitted for each metric is given in the table below.

Fewer forecast groups made a prediction for ACE, but those who did tended to be somewhat more conservative in their overall predictions for 2016. This could explain the difference between higher hurricane forecasts (which almost every group forecasts) and lower ACE forecasts.

The average values predicted by all groups issuing forecasts in March/April were 14 named storms, 7 hurricanes, 3 major hurricanes and an ACE of 107. These numbers increased in May/June to 14 named storms, 8 hurricanes, 3 major hurricanes and an ACE of 117. Predictions issued in July/August were generally lower than in the previous period, with 15 named storms, 7 hurricanes, 3 major hurricanes and an ACE of 97. The number of forecasts for each period was 10 for March-April, 10 for May-June and 6 for July-August. Since the number of forecasts and the composition of the ensemble changes for each 2 month period, it is difficult to analyze how the skill of the average forecast changes as a function of forecast time.

The figure below shows changes in the forecast error of an ensemble constructed using only the individual forecasts which were updated at least once during the months leading up to the hurricane season.

The tendency to increase the predictions for May-June (compared to March-April) and to revise the forecasts downward at the start of the season is clearly visible. And while one would generally expect an increase in skill with forecast time, except for forecasts of named storms, this is not observed in this case. That being said, forecasts of named storms and hurricanes produced in July-August performed better than those produced in March-April and May-June.

Physical mechanism

The 2016 Atlantic hurricane season was generally fairly quiet until late September, when long-lived major hurricane Matthew formed. Matthew was responsible for over 40% of all ACE generated during this season. 2016 was the 3rd year in a row where more ACE was generated in October than in September. On average, about three times as much ACE is generated in the Atlantic in September compared with October.

The 2016 Atlantic hurricane season was generally characterized by cool neutral El Niño-Southern Oscillation (ENSO) conditions. Contrary to the past two years, this season was characterized by below-average vertical wind shear when averaged from August-October across most of the tropical Atlantic. Slightly above-average vertical wind shear prevailed over the Caribbean. While the tropical Atlantic was somewhat warmer-than-normal, the mid-levels of the atmosphere were quite dry this hurricane season. Dry mid-levels enhance downdrafts and suppress the deep convection necessary for supporting hurricanes.

A thorough discussion of the 2016 Atlantic hurricane season is available from Colorado State University.