Editor's Note: The 2008 hurricane season is underway—and off to an early start with Tropical Storm Arthur making a landfall along the northern coast of Belize on May 31. What kind of season are forecasters projecting for 2008? In this article, AIR Director of Research in Atmospheric Science, Dr. Pete Dailey discusses the factors that influence seasonal forecasts and what the current indicators—and the forecasters—are telling us.

By Peter S. Dailey, Director of Research in Atmospheric Science, AIR Worldwide Corporation

Introduction
Every hurricane season is unique, and it is the job of seasonal forecasters to anticipate how the upcoming season may deviate from long-term averages. Seasonal forecasters make their forecasts based on an examination and interpretation of the key climate factors that enhance or suppress activity on a seasonal time scale. For example, Figure 1 shows one measure of the primary fuel source for hurricanes, namely heat from the ocean surface. Sea surface temperatures, or SSTs, are measured over the Northern Atlantic Ocean and then, year to year, the temperature is compared to the long-term average to determine that year's SST anomaly. Values above zero indicate a warmer than average ocean surface, and values below zero indicate cool years. As is seen in Figure 1, the Atlantic has been consistently warmer than average since the mid-1990s and this has led to elevated tropical cyclone activity in the Atlantic during that period.

Figure 1. Atlantic Sea Surface Temperature Anomalies, 1950-2007

Most seasonal forecasts factor in projections for how warm or cool the ocean is expected to be in the coming season to estimate how much the ocean may influence activity. Over the last several years, above average SSTs explain, at least in part, why most seasonal forecasts have projected activity elevated well above the long-term average.

2008 Sea Surface Temperatures
The 2008 hurricane season is just underway, but forecasted SSTs for the busiest months ahead (July, August and September) are available. Figure 2 shows one such forecast from the International Research Institute for Climate and Society (IRI) out of Columbia University.

Figure 2 shows forecast SST anomalies for the months of July, August, and September. Yellows and reds indicate regions projected to be warmer than average, and greens indicate projected cool regions. These forecasts are derived through a combination of statistical analysis and through the use of physical—sometimes termed "dynamical"—general circulation models.

Figure 2. SSTs are expected to be near normal in the tropical latitudes of the Northern Atlantic during the 2008 hurricane season. (Source: IRI. Forecast for July, August and September.)

Focusing in on the Atlantic Ocean, it becomes apparent that the Northern Atlantic is expected to be warmer than average. However, in the tropical latitudes of the Northern Atlantic (red box)—where most hurricanes form and intensify—IRI expects it to be within a quarter degree of the long-term average, as indicated by the white color. While the mid- and polar- latitudes of the Northern Atlantic are projected to be quite warm, this may not have a great deal of influence on tropical cyclone activity.

Figure 3 shows observed ocean temperatures averaged over the four weeks of April for 2008 and for the previous four years. Many of the seasonal forecasting groups typically issue their first forecasts for the season ahead in April. Why are April SSTs relevant for the hurricane season that begins in June? Since water has a very high heat capacity, the oceans of the world are able to retain an enormous amount of heat. As a result, if the ocean is warm in April, it is very likely to remain warm into May and June, and potentially through much of the hurricane season.

Figure 3. Observed pre-season SST anomalies, 2004-2008.

Again, if one focuses in Figure 3 on the tropical Atlantic (red boxes), the 2008 SSTs are a mix of slightly cool and slightly warm temperatures. This is in sharp contrast to Atlantic SST anomalies for the same period for the previous four years, which were significantly warmer than average.

Will 2008 Be an El Niño or La Niña Year?
SSTs are the fuel source for hurricanes, but another key parameter in seasonal prediction is the level of wind shear anticipated in the Atlantic. Wind shear is a measure of how drastically the winds change with height. High levels of shear are generally destructive to tropical cyclones because shear disrupts the ability of a cyclone to convert the ocean's heat to wind energy.

Wind shear, like most weather variables, fluctuates dramatically within a season; but seasonal forecasters can take advantage of the relationship between wind shear and a more predictable climate signal known as ENSO, or the El Niño/La Niña cycle. Atlantic wind shear and ENSO are related through a complex cause-and-effect mechanism known as a "teleconnection," which in this case modifies Atlantic wind shear as SSTs off the coast of Peru change.

Figure 4 shows the mean level of wind shear in the Atlantic under typical El Niño (top) and La Niña (bottom) conditions. White colors indicate high levels of shear, which suppresses hurricane activity, and blue colors indicate low shear, which tends to enhance hurricane activity. As the top panel of Figure 4 indicates, during El Niño years, when SSTs in the Eastern Pacific are warm, Atlantic wind shear tends to increase—especially along the key tropical latitudes where storms often form. In contrast, the bottom panel shows that during La Niña years, when Pacific SSTs are cool, wind shear is suppressed in the Atlantic, allowing for more storms to form and intensify.

Figure 4. The impact of the ENSO cycle on wind shear in the tropical Atlantic (Source: IRI)

Notice that the region most impacted by ENSO lies between the Caribbean Islands and the coast of South America, at the far left of the red boxes in Figure 4. This is a key region because storms forming here have a relatively high probability of making landfall somewhere along the North American coastline.

Heading into the 2008 hurricane season, seasonal forecasters project ENSO because of its implications on the Atlantic season. When ENSO is projected to be in its El Niño phase, it should lead to enhanced shear and suppressed tropical cyclone activity in the Atlantic. During the La Niña phase, it should lead to suppressed shear and elevated Atlantic activity.

What are the projections for ENSO in 2008? Figure 5 shows a summary of ENSO observations and forecasts from a variety of global climate models. The x-axis shows time scale and the y-axis is the Pacific Ocean SST anomaly, with warm temperatures on the top half indicating an El Niño condition and the cool anomalies on the bottom indicating a La Niña (cool Pacific) condition.

Figure 5. ENSO forecast ensemble using a range of physical and statistical models (Source: IRI)

Note that the actual observed anomalies for the first four months of 2008 show that the Pacific was cool during this period (solid blue diamonds on the left of the chart)—a moderate La Niña event. However, the tendency seems to be for warming, indicating that the La Niña is weakening. The series of colored lines in the remainder of the plot show ENSO projections from a variety of statistical and physical models from now through well into 2009.

Figure 6 repeats Figure 5, but here the heavy red line represents the average of the physical models, the heavy blue line represents the average of the statistical models and the heavy green line represents the average of all models. A grey box is drawn around forecasts for the months that represent the 2008 Atlantic hurricane season (June through November).

Figure 6. ENSO forecasts predict relatively average Atlantic wind shear for 2008 hurricane season. (Source: IRI, AIR Worldwide)

Note, the consensus of the statistical models is for a weakening La Niña condition, and the consensus of the physical models is a return to near neutral conditions. The uncertainty is fairly considerable, however; some models actually project a maintained La Niña while others show the growth of a moderate El Niño event. Despite the uncertainties in these forecasts, the average for all models projects a very weak La Niña or nearly neutral condition for the coming Atlantic season.

Thus, when seasonal forecasters evaluate these projections, the results support a slightly low or roughly average level of wind shear in the Atlantic, and thus a slight tendency for enhanced tropical cyclone activity from the wind shear perspective.

Putting it All Together
To summarize the discussion thus far, the projected climate conditions for the 2008 hurricane season include SSTs that are expected to be warm for the Atlantic, but not especially warm in the tropical latitudes. (The current observed Atlantic anomalies are significantly cooler than in the last several years.) And second, wind shear conditions are expected to be slightly suppressed by a (weakening) La Niña, and this tends to support slightly elevated activity for 2008.

Of course, it is important to bear in mind what seasonal forecasters know all too well, namely that while SSTs and shear both play a key role in determining levels of tropical cyclone activity, there are other factors that can "make or break" a season. One such factor is the Saharan Air Layer (SAL), which episodically transports dust from the Saharan Desert over the Atlantic and can suppress storm formation. In addition, the general circulation of the atmosphere—including the strength and position of the Bermuda High—can influence where storms form and intensify over and, ultimately, what parts of the coastline are at risk for landfall. Finally, we all know that it takes just one major hurricane landfall along a densely populated part of the coastline to produce catastrophic damage and loss. So while the seasonal forecasts for basin activity are useful, those storms that actually make landfall are most critical to insurers.

Figure 7 shows the latest seasonal forecasts issued by various forecast operations, including Tropical Storm Risk (TSR) and Weather Services International (WSI), both commercial companies, and Colorado State (CSU) and North Carolina State (NC State) Universities. NOAA is the official U.S. team charged with forecasting Atlantic activity, and their forecast is released close to the beginning of the season. Some forecast operations provide a range, including TSR, NC State and NOAA. These are represented by the vertical bars.

Also shown in Figure 7 are the long-term average frequencies since 1950: 10.7 tropical storms per year and 6.2 hurricanes per year.

Figure 7. 2008 seasonal forecasts. (Source: AIR)

TSR's most recent forecast is for 14.4 (+/- 3.4) named storms and 7.7 (+/- 2.4) hurricanes this year. WSI's latest prediction is for 14 named storms, with 8 becoming hurricanes. Researchers at NC State predict between 13 and 15 named storms and between 6 and 8 hurricanes. CSU, which has been in the Atlantic seasonal forecasting business the longest, forecast 15 tropical storms and 8 hurricanes. Finally, NOAA predicts 12 to 16 named storms and 6 to 9 hurricanes. Note that the UKMet office, which last year developed a new physically-based model for seasonal forecasting, has not yet issued a forecast.

In Summary
In summary, this year's seasonal forecasts range from roughly average activity to above average activity for both tropical storms and hurricanes. What is noticeably absent from most seasonal forecasts is the number that will make landfall along the North American coastline. On average, only about 1 in 7 tropical storms reaches the U.S. coastline at hurricane strength.

The seasonal forecasters are well aware that their forecasts—which are primarily for basin, rather than landfall, activity—are largely dependent on predicting factors like SSTs and wind shear, which in themselves have uncertainty. Also, other factors like the Saharan Dust Layer can impact the season and are highly unpredictable. Further, projecting steering patterns—which ultimately dictate how close storms from the basin will approach the coastline—is highly uncertain months in advance. This explains the lack of landfall information in most forecasts.

While there is some skill in the models that predict the number of tropical storms that will form in a season, it is more difficult and therefore more uncertain how many of those named storms will intensify to hurricane strength. It is even more uncertain how many storms that do achieve hurricane status will reach the coastline having maintained that intensity. Since hurricane landfall is what insurers and reinsurers (and property owners) are most interested in, all these facets of predictive uncertainty are important to recognize and, indeed, they are an important element of any forecast of hurricane landfall risk.

Our position on the use of the two estimates of hurricane risk contained in the standard and WSST catalogs remains the same as in previous years: AIR will continue to provide both as part of the U.S. Hurricane Model, but the decision as to how to use these catalogs will depend on each clients' business and tolerance for risk.

Finally, both catalogs are based on credible, peer-reviewed techniques for estimating landfall risk. Indeed, the work leading up to this year's WSST catalog has been accepted for publication in a forthcoming issue of the American Meteorological Society's Journal of Applied Meteorology and Climatology. AIR's research has also been presented around the world at various technical and industry conferences.