Storm Trackers

Why are hurricane seasons so dissimilar? How precise are pre-season forecasts? Key climatological data can help explain the observed variability and therefore aid in assessing the consequent optimal risk assessment approach.

THREE DIFFERENT SEASONS

The Atlantic hurricane seasons of 2004, 2005 and 2006 are a clear example of the striking risk volatility of natural hazards on an annual timescale (See Chart 1, page 36). Faced with such variability, markets are frequently influenced by last year’s results and pre-season hurricane forecasts. But to what extent does this make sense? The risk and climatological data presented below explains the observed seasonal dissimilarities in terms of frequency, intensity and track, highlighting important considerations for risk assessment.

FORECASTING HURRICANE SEASONS

After two intense Atlantic hurricane seasons in 2004 and 2005, forecasts in June 2006 perhaps not surprisingly predicted above-average Atlantic storm frequency for the 2006 season (which runs from June to November each year). This year, the same prediction has been made for the 2007 hurricane season.

But what in fact happened in 2006 was an average start in terms of tropical storm frequency. The first hurricane formed only in mid-September — about a month later than the long-term average (See Figure 1). All told, the 2006 season produced 10 tropical storms, five hurricanes, and two major hurricanes. All these numbers are below the June 2006 forecasts, and below the respective long-term averages of 11 tropical storms, six hurricanes and three major (Category 3-5) hurricanes.

Stepping back, the June 2004 and 2005 forecasts anticipated approximately average and slightly above-average storm frequency. Both underestimated the activity. The 2004 storm season got off to a slow start; the first tropical storm didn’t form until the end of July. But the season quickly reached an above-average activity of 15 tropical storms, nine hurricanes and six major hurricanes. The 2005 season was exceptionally active, with a total of 28 tropical storms, 15 hurricanes and seven major hurricanes. The season started unusually early and lasted longer than usual, with the last tropical storm forming on Dec. 30, 2005.

The fact is, seasonal forecasting of tropical cyclone activity is an extremely challenging task: It is still almost impossible to predict the evolution of the climate and weather patterns that influence the activity in advance of a season.

FREQUENCY, INTENSITY AND TRACK

Tropical cyclone activity can be linked to key determining climatological phenomena such as sea surface temperature, El Nio and weather systems including the ‘Bermuda high.’

Sea Surface Temperature

A minimum sea surface temperature (SST) — i.e. a minimum ocean heat energy content — is required for tropical cyclone formation. The mean SST over the whole of the North Atlantic basin (a measure known as the Multidecadal Oscillation, or AMO) correlates well with observed storm activity. Since 1995, the AMO has been higher than in the preceding period of 1965 to 1995 (See Chart 2, Page 38) and there have been more hurricanes.

The scientific community, however, is still debating whether we are observing a natural cycle of warm and cold phases, or whether other factors, such as climate change, are at least partially responsible for the current upward trend of the AMO.

The SST also influences tropical cyclone activity at a smaller scale. The SST in the mid-Atlantic, where most hurricanes form, was above average in June and July of 2005. In contrast, in 2004 and 2006, the corresponding SSTs were closer to the average. This explains the early and intense start to the 2005 season compared to 2004 and 2006.

Vertical wind shear and El Nio

Tropical cyclone activity is also influenced by vertical wind shear in the atmosphere (i.e. the difference in wind speed between the sea surface and the top of the atmosphere). High shear, a large vertical wind speed gradient, inhibits the development of a tropical cyclone and can also reduce the intensity of an existing storm by tipping over the storm tower. Vertical shear over the Atlantic usually exhibits a regular-time oscillation, with high shear every other season (hence the climate signal name of Quasibiannual Oscillation, or QBO).

El Nio events have in the past been observed in combination with reduced Atlantic hurricane activity (See Chart 3, page 38). This is because El Nio brings dry air and adds wind shear, both of which inhibit tropical cyclone formation. The appearance of El Nio in 2006 is regarded as a contributing factor to the low hurricane frequency of that season. However, in contrast to the regular QBO oscillation described above, El Nio events are irregular and far more difficult to predict. El Nio in 2006, for example, could only be officially declared in mid-September, when the hurricane season was mostly already over.

Track: Large-scale weather patterns

In addition to sea surface temperature and wind shear, storm track, which affects the number of tropical cyclones that make landfall, is strongly influenced by large-scale weather patterns. These are predictable only on a very short-term basis.

In 2004, 2005 and 2006, tropical cyclones displayed very different track patterns (See Chart 4, Page 38). In 2004, storms either headed towards the U.S. coast or remained out in the Atlantic. In contrast, 2005 tracks displayed no distinct pattern and were widely distributed across the Atlantic area. In 2006, only a small number of storms reached or came close to the U.S. coast; all other storms turned north/northeast and away from land.

The reason for the split in 2004 was the high-pressure system between the U.S. coast and Bermuda, the so-called ‘Bermuda high,’ which prevented storms from passing through a region just east of Bermuda. In 2005, there was no distinct and persistent large-scale weather pattern to direct tracks in one direction or another. In 2006, a persistent low-pressure system over the eastern U.S. pushed storm tracks away from the U.S. coast, explaining the low number of landfalling tropical cyclones.

REACHING CANADA

Atlantic hurricanes are relatively uncommon in Canada, but there are several historical occurrences of hurricanes hitting the Canadian Atlantic coast (Hurricane Bob in 1991, for example, and Hurricane Hortense in 1996). Some major hurricanes have travelled further inland and reached Canada. Hurricane Hazel in 1954 is the strongest recorded hurricane to strike so far north, killing 81 people in Toronto. Similarly, the Great New England Hurricane of 1938 caused damage throughout New England and Massa-chusetts before eventually reaching Canada.

2007 FORECASTS

All forecasts are predicting an active 2007 Atlantic hurricane season. Although the actual numbers can be misleading, the various underlying climate conditions point towards an above-average activity. These are:

* a warmer-than-average Atlantic sea surface temperature, especially in the Caribbean; and

* signs of the formation of a weak La Nia (cooler-than-average sea surface temperatures in the equatorial Pacific). This would cause a reduction in wind shear over the Atlantic and thus have the opposite effect as an El Nio event.

IMPACT OF SEASONAL VARIABILITY ON RISK ASSESSMENT

Ideally, risk assessment should be based on the long-term climatology of tropical cyclones, while at the same time including the newest scientific data and findings.

The reliability of seasonal forecasts is not yet high enough for them to be considered in the assessment of hurricane risk. Rather, if there is a consensus of being in a phase of heightened or reduced activity in comparison to the long-term average, this may be taken into account as a pricing adjustment. This is happening now in the pricing of reinsurance treaties due to the c urrent phase of high activity, which scientists expect to last for at least another five to 10 years.

Footnotes

1National Oceanic and Atmospheric Administration (NOAA).

2The Atlantic Hurricane Database Reanalysis Project is led by the Hurricane Research Division of the Atlantic Oceanographic and Meteorological Laboratory (AOML). AOML is a part of the NOAA.

3The Climate Diagnostics Center (CDC) is part of the NOAA.

4The National Centers for Environmental Prediction (NCEP) is part of the NOAA.