Forecast of 2017 Atlantic Hurricane Activity

August 8, 2017


CFAN’s seasonal forecasts for Atlantic tropical cyclone activity are based on a breakthrough in understanding of the impact of global climate dynamics on Atlantic hurricane activity.  


We continue to predict an active 2017 Atlantic hurricane season:

•  ACE:  128 (average value 103 since 1980)

•  U.S. landfalling hurricanes: 1.5 (average value 1.5 since 1995)


To date in the 2017 Atlantic hurricane season, we have seen 5 named storms, but none have become hurricanes. Historically, 95% of hurricanes have developed after August 1.


Forecasters have increasingly been predicting an active hurricane season.   Traditional early season signals for an active hurricane signals seen in 2017 include:

  • Early season tropical cyclones developing from African easterly waves

  • Greater than average rainfall in west Africa


Other signals of a generally favorable set up for Atlantic hurricane activity in 2017 based on data from May-June-July (MJJ) include (see Figure 1):

  • Trend (MJJ) of cooling sea surface temperatures (SST) in the equatorial eastern Pacific, indicating a trend away from El Nino conditions towards solidly neutral ENSO conditions

  • Warm sea surface temperatures in the western Atlantic, and warming trend during MJJ

  • Evidence of strong subtropical anticyclone in Atlantic, with positive trending sea level pressure (SLP) for MJJ in the region

  • Decreasing SLP in the Carribbean for MJJ

Figure 1:  Global maps of:  average sea surface temperature anomalies for the period May through June (upper left); average sea level pressure anomalies for the period May through June (upper right);  sea surface temperature trend from May through June (lower left); sea level pressure trend from May through June (lower right)

North Atlantic hurricanes vary from year to year with basin-scale sea surface temperatures (SSTs), Pacific ENSO conditions and other large-scale patterns of ocean-atmosphere anomalies. Moreover, recent research shows that mechanisms favorable to greater hurricane activity also tend to inhibit U.S. landfalls, a further indication that multiple factors, in combination, contribute to overall seasonal storm behavior. As the 2017 North Atlantic hurricane season begins, we identify several regional and global anomaly patterns that have historically led to high-amplitude seasonal anomalies in both hurricane activity (i.e. North Atlantic ACE) and the total number of landfalling U.S. hurricanes. 


The 2017 forecast is very challenging since recent May-June-July (MJJ) conditions show evidence of conflicting hurricane influences. We anticipate that these influences will combine to produce near-average anomalies of both North Atlantic ACE and U.S. hurricane landfalls relative to the average for the period since 1995, which has overall been an active period for Atlantic hurricanes.


Hindcasts and predictions using CFAN’s predictors for Atlantic ACE and U.S. landfalls are described below.   

Accumulated Cyclone Energy (ACE) forecast

ACE is an integrative metric of tropical cyclone duration and intensity based on the square of maximum sustained surface winds during Named Storms. To date, the North Atlantic ACE is 4, about half of the normal year-to-date values.  Historically, 90% of the ACE has been after August 1.


CFAN’s June ACE forecast model used March-April-May observations of SST anomalies, SST tendencies, SLP lower-stratospheric temperatures. Our June ACE forecast was for 134 (+/- 35). 


With closer proximity to period of peak hurricane activity, our August forecast relies primarily on atmospheric (troposheric) predictors. We use both global and North Atlantic tropospheric predictors for May • June • July (MJJ):  geopotential heights, winds, divergence, and trends in velocity potential.


Higher ACE totals are typically promoted by cool La Nina-like surface conditions in the tropical eastern Pacific.  La Nina tends to inhibit wind shear above the low-latitude North Atlantic, allowing tropical disturbances to mature into hurricane-intensity storm systems. Additionally cool eastern Pacific surface conditions favor greater atmospheric subsidence and higher sea-level pressure, enhancing low-level divergence of winds and moisture, and greater convergence hurricane activity in the adjacent Atlantic.  While the eastern Pacific has cooled in recent months, SST anomalies remain in a near-neutral state, suggesting limited hurricane impacts. However, we identify stronger La Niña-like tendencies in global 3-dimensional patterns of tropospheric velocity potential (VP), a metric of atmospheric divergence that we identify as a more consistent indication of atmospheric hurricane forcing and high ACE totals than cool eastern Pacific SSTs. VP indices, based on historical relationships, suggest a 2017 ACE total of 170 to 190, considerably higher than the 1995-2016 mean of 128. 

However, other global-scale aspects of the 2017 MJJ atmosphere show strong historical connections to negative ACE anomalies and reduced hurricane activity.  Global 500 hPa geopotential height (Z500) anomalies suggest a 2017 ACE value below 100, based on 2017 MJJ Z500 anomaly patterns that contrast with historical patterns of positive ACE correlation (Fig. 3).  Persistent negative Z500 anomalies during MJJ appear over the Asian sector of the Arctic and the NE Atlantic; similar anomalies tend to precede reduced ACE totals.

Figure 4: ACE model estimates using 4 highest-correlated predictors (colored lines), compared with observed values (black line), for predictors in the North Atlantic (left) and global predictors (right).

U.S. Landfalling Hurricanes Forecast 

To date, there have been no U.S. landfalling tropical cyclones, although Tropical Storms Cindy and Emily have made U.S. landfall.


CFAN’s June forecast for the # of U.S. landfalls was based on an Arctic Index that characterized circulation patterns at the surface and in the stratosphere.  CFAN’s June forecast for the # of U.S. landfalls was 3 +/- 1.


There is a general correlation between ACE and the number of U.S. landfalls since 1995:

  • Of the 7 years with 0 landfalls, all had ACE values below 120, with the exception of 2010 (ACE of 164).

  • Of the 5 years with 3 or more landfalls, all had ACE values exceeding 140


However, the years with intermediate values of ACE (between 110 and 150) had landfall values between 0 and 3.


Positive and negative hurricane tendencies in MJJ 2017 atmospheric circulation suggest a quantitatively ‘average’ season for U.S. landfalls, despite the prediction of an active season in terms of ACE.


Comparison of the predicted predicted number of U.S. landfalls (for each of 4 predictors with the highest correlations) with the observed number of landfalls is shown in Figure 6, for global predictors and North Atlantic predictors. 

Figure 2:  Correlations of North Atlantic ACE with MJJ 200 hPa velocity potential tendencies (dVP200, left) and 2017 MJJ anomalies (right).

Figure 3:  MJJ 500 hPa geopotential height (Z500) correlations with North Atlantic ACE (left), compared with normalized 2017 MJJ anomalies (right).  Circled features highlight 2017 anomalies that contrast with historical ACE correlations and may tend to limit ACE during 2017.

To develop our ACE forecasts, we considered a total of 40 different predictors, from both global patterns and patterns in the North Atlantic.  We examined patterns in geopotential heights, wind velocity, divergence, and velocity potential trends at each of 5 levels in the troposphere.


Comparison of the predicted ACE with observed ACE is shown in Figure 4, for the 4 predictors with highest correlations among the global predictors and North Atlantic predictors. 

The individual predictors with correlations of 0.8 or higher provide the following range of ACE forecasts for 2017:

  • North Atlantic: mean 120;  range 89-140

  • Global: mean 136;  range 74-190


The overall spread among the different predictors for 2017 ranks among the highest for the years considered.  There is a wider spread and a bimodal distribution in the global estimates – higher values from the velocity potential indices, and lower values from  tropospheric geopotential heights.  The lowest estimates relate to the positions and shapes of geopotential height anomalies at 700 and 500 hPa in the South Indian Ocean and around North America, that are somewhat opposite to the 2004/2005 teleconnection patterns.


Our final 2017 ACE forecast is 128, with average error of 30.

Figure 5: Model estimates of the number of U.S. landfalls using 4 highest-correlated predictors (colored lines), compared with observed values (black line), for predictors in the North Atlantic (left) and global predictors (right).

The individual predictors provide the following range of forecasts for 2017 U.S. landfalls:

  • North Atlantic: mean 1.5;  range 1.2 to 2.3

  • Global: mean 1.5;  range 1.4 to 1.6


The most likely outcome is 1-2 landfalls, with the possibility of zero landfalls indicated by a stratospheric height pattern that is analogous to 2010 (high ACE year with zero landfalls). We don’t see any evidence in the current forecast to support 3 or more landfalls in 2017.  Our June forecast of 3 U.S. hurricane landfalls now seems to be too high.  We will be revisiting our June forecast model before the June 2018 forecast.


30 day forecast


CFAN produces twice weekly outlooks out to 32 days for track anomalies, using the ECMWF Ensemble Forecast System.  Figure 7 shows the most recent two forecasts. 


We are starting to see a regular train of African easterly waves, which traditionally have been the source of 60% of Atlantic hurricanes and 85% of major hurricanes.


The forecast initialized 8/3/17 shows positive track anomalies that are slightly southward of climatology as the waves traverse the Atlantic.  The dominant track shows recurvature in the Atlantic.  The forecast initialized 8/7 shows similar features, with possible landfalls along the Atlantic coast.

Figure 6:  32 day forecast of track density anomalies. Top:  forecast initialized 8/3/17.  Bottom:  forecast initialized 8/7/17.

Further information about CFAN’s forecasts


Further information about CFAN’s tropical forecast products – TropiCast – can be found at


CFAN publishes seasonal hurricane forecasts for the Atlantic on the following schedule (typically around the 8th of the month):

  • June

  • August

  • December

  • April


The June and August forecast summaries and reports will be issued publicly, although the technical forecast reports will be available only to subscribers. Options for subscription to CFAN’s seasonal hurricane forecast reports:

  • Subscribers: $250/year. December, April forecast summaries

  • Sponsors:  $4K/year.  Forecasts summaries; technical reports; consultation.


About CFAN


Climate Forecast Applications Network (CFAN) develops innovative forecast tools that give longer and more accurate warnings of extreme weather events, so clients can better prepare and recover.  CFAN’s staff applies the latest research to a wide range of customer challenges, helping businesses and government around the world. Our advanced prediction tools provide clients with the confidence to make complex and difficult decisions about weather risks.

CFAN was founded in 2006 by Judith Curry and Peter Webster and launched under the Enterprise Innovation Institute’s VentureLab program at Georgia Tech. Its research has been assisted by grants from NOAA, NASA, and the Department of Energy.


Contact Information


Dr. James Johnstone, Senior Scientist


Dr. Judith Curry, President
(404) 803-2012