Tropical Storm Ana made landfall in South Carolina on May 10, 2015, becoming the earliest tropical cyclone to make landfall in the U.S. (NASA)

Each spring, seasonal hurricane forecasts are issued, which attempt to predict what the overall activity will be in the months to come. These forecasts are issued by many groups, including forecasters at Colorado State, the National Weather Service, U.K.-based company Tropical Storm Risk and the U.K. Met Office. Now that hurricane season is ending, it’s time to take a look back at how they performed.

The Atlantic hurricane season, which officially comes to a close on Nov. 30, has been characterized by below-average activity punctuated with brief bursts of action. Eleven named storms, four hurricanes, and two major hurricanes major (Category 3 or stronger) formed this year, and the accumulated cyclone energy, which measures a season’s total activity, was lower than normal. In an average season, 12 named tropical storms, 6.5 hurricane and two major hurricanes form.

[Hurricane Joaquin tops the list as 2015 season comes to a close]

All four groups made similar predictions in the weeks leading up to the start of hurricane season, June 1, since everyone knew that this year’s El Niño was likely going to dominate the Atlantic basin for much of summer and fall.

Tropical Storm Risk ended up having the best accumulated cyclone energy prediction back in April; the group forecasted an ACE of 56, and the final was 62. The National Weather Service had the best forecast for ACE in May and June, and by August, both Tropical Storm Risk and the Weather Service were doing well in their ACE forecasts.

What was so interesting about this season is that nearly half of the ACE accrued in 2015 was generated by one storm — Hurricane Joaquin. Without Joaquin, the cyclone energy would have been much lower.


Seasonal forecasts are issued at three different times: April, May and June, and August. Each group’s predictions are listed at the different lead times when their forecasts are available. For groups that issue predictions as a range, the midpoint of the range was taken (e.g., a prediction of 7-13 named storms is shown as 10 named storms). (Phil Klotzbach)

The models that these groups use for seasonal forecasts were generally skilled in predicting a below-average season. Models are usually graded based on the long-term average from 1981 to 2010 — a skillful model does a better job than simply predicting an average year.

But all of the models showed even higher skill this year when compared against the short-term average of 2005 to 2014, which is extremely high thanks to gangbuster hurricane seasons like 2005 and 2010. This short-term average has been shown to be a tougher measuring stick when evaluating the skill of these seasonal forecasts over a longer period of time.

The biggest question going into this year’s hurricane season was how strong El Niño would be. All of the forecasting groups anticipated a moderate to strong El Niño event this year, and this prediction turned out to be correct.

The European model did an excellent job predicting this year’s El Niño event. While there is quite a spread in the forecasts that the model produced, “reality” fell solidly in the center of these forecasts. This is especially impressive given that, in general, there is large uncertainty with forecasts for El Niño issued during the spring.


Early April 2015 European model forecast for the Nino 3.4 region. The dotted line represents what actually happened. (ECMWF)

Probably the biggest surprise with this year’s hurricane season, and one of the reasons why the forecasts were slightly lower than what actually happened, was the significant warming that occurred across the tropical Atlantic over the course of the season. Sea surface temperatures were below-average across the primary region where most intense hurricanes form in May.


May sea surface temperature departure from normal. (NOAA)

But the tropical Atlantic warmed up considerably during the hurricane season. This was due to a combination of weak winds blowing across the tropical Atlantic, and increased incoming solar radiation due to less cloudiness caused by sinking motion driven by El Niño.

By October 2015, the tropical Atlantic had temperatures at near-record levels. The warmer tropical Atlantic was likely one of the reasons why several tropical cyclones formed in the eastern Atlantic this year, only to be torn apart by El Niño-fueled shear as they approached the Windward Islands.


October sea surface temperature departure from normal. (NOAA)

In general, seasonal forecasts did a good job anticipating a below-average Atlantic hurricane season in 2015 due to a strong El Niño event. Most seasonal forecasts predicted a bit less activity than was observed, due to a surprising warming of the tropical Atlantic during the peak of the hurricane season this year.

A little more about how each group makes their forecast

Colorado State has the longest history of continuous seasonal hurricane predictions, starting with the 1984 Atlantic hurricane season. The group at Colorado State utilizes a statistical methodology, that is, they relate large-scale conditions in the atmosphere and ocean (such as El Niño and Atlantic basin sea surface temperatures) and assess how well they predicted historical hurricane seasons.

csu_predictors
Current predictors utilized in the CSU statistical forecast issued in early August. (Phil Klotzbach)

NOAA, or the National Weather Service, has been issuing seasonal hurricane forecasts since 1999. These forecasts utilize a combination of statistical relationships between large-scale climate signals, such as the Atlantic multi-decadal oscillation along with predicted hurricane frequency from numerical model predictions to arrive at their final hurricane numbers.

noaa_climate_conditions
Large-scale features that NOAA looks for that are typically associated with active Atlantic hurricane seasons. (NOAA)

Tropical Storm Risk, based at the University College London in the U.K., has been issuing statistically-based seasonal hurricane forecasts since 1999. These forecasts are based on two primary predictors: 1) July-September low-level wind speed across the Caribbean and tropical Atlantic and 2) August-September tropical Atlantic sea surface temperatures. They utilize a statistical model to forecast these two parameters and then use forecast values of those two parameters to arrive at their forecast hurricane numbers.


Current predictors utilized in Tropical Storm Risk’s seasonal hurricane forecasts. The figure also displays the anomalies in August-September SST (color coded in °C) and low-level wind (arrowed) linked to active Atlantic hurricane years. (Tropical Storm Risk)

The U.K. Met Office has been issuing dynamically-based seasonal hurricane forecasts since 2007. These forecasts are based on a numerical model that simulates the upcoming year’s hurricane season and counts the number of hurricane-like systems that develop in the model.

The author is a member of the Colorado State University seasonal forecasting group.