JUNE 01 AIRCURRENTS: TRACKING THE 01 ATLANTIC HURRICANE SEASON USING THE CLIMATECAST U.S. HURRICANE RISK INDEX EDITOR S NOTE: This article by AIR s Director of Atmospheric Science, Dr. Peter Dailey, gives a brief preview of the 01 Hurricane Season now underway and explains how AIR s ClimateCast service provides an objective and up-to-date measure of U.S. hurricane risk. The 01 Atlantic Hurricane Season has officially begun. Forecasts for the season have been issued since as early as last December. (See the Preview of the 01 Atlantic Hurricane Season box below.) With early expectations suggesting that 01 would be fairly normal, mother nature raised up two named storms even before the season started the first time this has happened in over a century (not since 1908). Regardless of whether the 01 season turns out in the end to be above, below, or just plain normal, insurers, reinsurers, and other risk managers will be monitoring tropical cyclone activity continually. To assist in this task, increasingly sophisticated risk assessment tools are now available that can be used in real-time. THE ARTICLE: Describes the ClimateCast U.S. Hurricane Risk Index and explains the Index s operation and features as it tracked pre-01 Season Tropical Storm Beryl. HIGHLIGHTS: This new tool, which complements AIR s ClimateCast Atlantic Hurricane Conditions Advisory, provides real-time assessments of loss potential to insured properties in the United States from active tropical cyclones in the Atlantic. The Index has been expanded to include five at-risk regions and can be customized to evaluate real-time risk for individual portfolios. This article outlines AIR s ClimateCast service and demonstrates its use by way of Tropical Storm Beryl, which made landfall near Jacksonville Beach, Florida, just after midnight on May 8th. MONITORING TROPICAL CYCLONES IN THE ATLANTIC AIR s ClimateCast service provides two complementary web-based tools for monitoring tropical cyclones and assessing their potential risk. ClimateCast Atlantic Hurricane Conditions displays, every six hours, the environmental factors that influence the development, movement, and landfall potential of active tropical cyclones in the Atlantic basin. Figure 1 shows the conditions advisory on the eve of the 01 season: midnight UTC on Monday, May 8th about four hours before Tropical Storm Beryl made landfall. The top left panel of the advisory indicates that Beryl is active (its name is in red in the list of 01 storm names). Beryl s track as of midnight on the 8th is shown, as is its location at that time (indicated by the gray circle and B at Florida). The track of Tropical Storm Alberto, which preceded Beryl by about a week in almost the same location (and which dissipated without making landfall), is also shown. The bottom portion of the panel lists other information about the 01 season thus far. The other three panels show sea surface temperature conditions (top right), wind shear conditions (lower left), and atmospheric steering currents (lower right). Figure 1. The ClimateCast Atlantic Hurricane Conditions advisory as Beryl was about to make landfall in Florida on Monday, May 8th; reds and oranges indicate regions favorable to development and intensification, while blues indicate conditions unfavorable to tropical cyclones. (Source: AIR)
USING THE CLIMATECAST HURRICANE CONDITIONS ADVISORY The ClimateCast Atlantic Hurricane Conditions advisory places the National Hurricane Center s (NHC) official storm track forecast within the constantly shifting oceanic and atmospheric conditions of the wider Atlantic basin. It thus provides valuable contextual information by which a deeper understanding of a storm s likely progress and potential impact can be developed. For example, the upper right panel of the May 8 th 00:00 UTC Advisory for Tropical Storm Beryl (Figure 1) shows both the current ocean temperatures (the line-demarcated contours) and anomalies areas of above- or below-normal temperatures (the colored regions). South of the blue contour-line (.5 Celsius) is where hurricanes are most likely to develop; they are most likely PREVIEW OF THE 01 ATLANTIC HURRICANE SEASON The appearance of two tropical storms, Alberto and Beryl, so early in 01 even before the Atlantic hurricane season had officially begun was not only something that hadn t happened in over a century, but Beryl was the strongest pre-season Atlantic tropical storm to make landfall in the United States ever. Does this suggest that 01 will see an unusually active hurricane season? The graph shows that by mid-june, one expects about one named storm to have formed in the Atlantic on average. Generally, activity rises dramatically between the beginning of August and the end of October. The statistically average year produces 11.3 named tropical storms, a little more than half of which (55%, or. storms) will develop into hurricanes. Of those hurricanes, only a third,.3, will intensify to major hurricane status (Saffir-Simpson Category 3 or higher). An Average Season Not according to history. Even though the second named storm in a year typically wouldn t be expected until July, the record of storm activity in the Atlantic over the past 0 years shows that the correlation between activity in the very early part of a season and a season s activity as a whole is very weak. That is, a strong start seldom means a strong season. It is also past experience that determines whether a season is active, normal, or below normal. Figure below shows the statistical average Atlantic hurricane season as derived from the historical record. Atlantic Basin Storm Count 1 11 10 9 8 7 5 3 1 TROPICAL STORMS HURRICANES MAJOR HURRICANES May Jun Jul Aug Sep Oct Nov Dec Jan Atlantic Hurricane Season Figure. An average Atlantic hurricane season. (Source: AIR) 11.3. (55%).3 (0%) Figure 3. Atlantic sea surface temperatures just before Tropical Storm Beryl s landfall. Influencing Factors An actual season, however, may diverge from the statistical average for a host of reasons, many of which are captured in the ClimateCast Hurricane Conditions advisory. Figure 3 shows the Sea Surface Temperatures panel of the conditions advisory for May 8 th. Immediately noticeable are the generally neutral or marginally cooler-than-average temperatures across the tropical Atlantic. In fact, the temperatures in this region on May 8 th were approximately the third-coolest observed since the current active hurricane cycle in the Atlantic basin began, in 1995. Additionally, temperatures in the pivotal Main Development
Region (which stretches south from Cuba to the northern tip of South America and east to Africa) were just marginally higher than average (about 0.35 C) and those temperatures were located mainly in the Caribbean. Even though temperatures may be ever so slightly elevated in the Caribbean, strong upper-level winds accompanying the subtropical jet stream have been situated over the region and expected to stay put at least through June. Such winds tend to create wind shear, which tears tropical storms apart before they can develop fully. 01 Atlantic Seasonal Forecasts With these conditions as background and the state of other longer-term meteorological and climate conditions (such as the El Niño/La Niña [ ENSO ] phenomenon in the Pacific) taken into account, the forecasts of several prominent academic, government, and private forecasters are shown in Figure. The solid horizontal line on the graphs indicates the long-term average number of storms. About half of the forecasters, including the National Oceanic and Atmospheric Administration (NOAA), have stated a range rather than a single number, drawing attention to the uncertainty that accompanies these forecasts. 0 18 1 1 1 10 8 TROPICAL STORMS long-term average = 11.3 HURRICANES long-term average =. <Dec Jan Feb Mar Apr May Jun <Dec Jan Feb Mar Apr May Jun Month of Forecast Release Figure. 01 pre-season Atlantic Hurricane forecasts. While NOAA s Administrator said the agency was predicting a less active season compared to recent years, the agency s official characterization was that the 01 season would be near normal. NOAA has forecast a 70 percent chance of 9 to 15 named storms forming, to 8 becoming hurricanes, and of those, 1 to 3 developing into major hurricanes. 1 13 1 11 10 9 8 7 5 3 TSR WSI CSU ACCU WRC NOAA NC State FSU Met Office to undergo rapid intensification when they move over warm eddies (indicated by the orange and red regions). As shown in this advisory, after landfall Beryl was expected to move back over the ocean and briefly encounter warmer waters near the Outer Banks. (In fact, Beryl dissipated completely over land and the NHC stopped monitoring the storm on May 30.) The bottom left panel of Figure 1 shows that wind shear along the forecast track would be mildly favorable to intensification orange and red colors in this panel indicate low wind shear while the panel on the bottom right, which shows steering currents, indicates that Beryl would be influenced to move toward the open and ultimately cooler ocean. THE CLIMATECAST U.S. HURRICANE RISK INDEX Figure 5 shows the AIR ClimateCast U.S. Hurricane Risk Index analysis for the same time period shown in Figure 1. The Risk Index provides an assessment of the risk to U.S. insured exposures. The assessment is based on the mean value of the full distribution of probable insured losses based on an ensemble of 500 potential track and intensity scenarios generated from operational forecast models and stochastic simulation techniques. The value of the index is expressed as insured loss indexed against the AIR U.S. Hurricane Model s exceedance probability (EP) curve. The scale is logarithmic; hence a value of 5.0 indicates an expected (or mean) loss corresponding to an annual exceedance probability for the U.S. of 10% (a 10-year return period loss), while a value of 10.0 indicates a 1% exceedance probability loss (100-year return period loss). The Risk Index is displayed (in red in the top left panel) to one decimal point-precision: for a weakening Tropical Storm Beryl four hours before landfall as shown here, the value was a relatively low 0.5. In addition to the value of the Risk Index, the top left panel provides a -Hour Trend indicator represented by an arrow or horizontal line indicating that the value of the index over the past day is trending upward (as it was for Beryl at that moment), downward, or not trending at all. 3
LANDFALL PROBABILITY AND RISK HISTORIES The top right panel in Figure 5 above is a landfall probability map. It displays the likelihood of a landfall occurring on any given 50-nautical mile segment of the U.S. coastline. Segments not at risk are shown in gray. Non-zero levels of landfall probability are depicted by segments colored according to the legend provided. Purple segments indicate the highest probability of landfall which is defined as at least 10% of the storm scenarios reaching a coastal segment with (sustained) damaging winds of at least 0 mph. The map above shows the segment of northern Florida coastline, where Tropical Storm Beryl would make landfall four hours later in purple. Figure 5. The ClimateCast Risk Index when Beryl was nearing U.S. landfall on Monday, May 8th; also shown are landfall probabilities, the evolution of the index over the season and over the last two weeks, both for the U.S. as a whole and by region. (Source: AIR) Immediately below the Risk Index is a graphic that indicates the -Hour Coefficient of Variation (COV), which reflects the swings, or volatility, in the index value over the previous four -hourly forecast cycles. These swings are represented by five vertical bars indicating very low, low, medium, high, or very high variation. The value of the COV for Beryl at the time stamp shown in Figure 5 was very low. Finally, this upper left-hand summary panel also displays the set of 500 simulated scenarios. Tracks for each scenario are color-coded according to their intensity (as indicated by a legend in the upper right of the panel). When multiple storms threaten the U.S., only the scenario-set for the storm having the highest index value is displayed. The two bottom panels of the ClimateCast U.S. Hurricane Risk Index display the evolution of the index value over time. The full record since May 1st is displayed in the left panel (in the lower half of which are also listed the storm names designated for the current season, active storms appearing in red), while a more visibly detailed two-week history is displayed in the right panel (in the lower half of which the record of the Risk Index is displayed at a more granular level; that is, it reflects risk to more narrowly defined geographic regions, specifically: Florida, Gulf (TX, LA, MS, AL), Southeast (GA, SC, NC), Northeast (CT, ME, MA, NH, NY, RI, VT, VA, DE, MD, PA, NJ), and Inland (OK, AK, TN, KY, WV, OH, IL, IN, MO). Finally, the gray band around the blue line indicating the index value represents the 50 th to 95 th percentile values associated with all 500 storm scenarios, and thus reflects the ensemble s fluctuating range of potential losses over time. As such, when this uncertainty CALCULATING THE HURRICANE RISK INDEX The AIR U.S. Hurricane Risk Index is derived as a log-scale of loss distributions estimated by the AIR U.S. Hurricane Model stochastic scenarios from an ensemble of 500 storm track and intensity projections. The scenarios describe a complete spectrum of damage and potential loss. Using the AIR U.S. Hurricane Model exceedance probability curve, loss is translated to a ten-step 0 to 10 risk index as shown in Figure. The benchmark values are 0 (no significant threat); 5.0, which corresponds to a 10-year return period loss, or 10% exceedance probability; and 10.0, which corresponds to a 100- year return period, or 1% exceedance probability. Selected historical storms have been placed along the index to indicate benchmarks of damage and loss were those storms to recur today. Full documentation of the ClimateCast Risk Index and how it is calculated can be found at http://alert. air-worldwide.com/climatecast_risk_desc.pdf. 10 100.0 1.0% 9 3.1 1.% 8 39.8.5% 7 5.1.0% 15.8.3% 5 10.0 10.0%.3 15.9% 3.0 5.0%.5 0.0% 1 1..5% Miami Hurricane (19) Hurricane Andrew (199) Hurricane Katrina (005) Hurricane Wilma (005) Hurricane Charley (00) Figure. The ClimateCast Risk Index scale and benchmark loss-causing U.S. landfalls. (Source: AIR)
band narrows, it indicates that confidence with respect to the risk estimate is increasing. For a low-intensity storm like Beryl, the Risk Index remained low, with an uncertainty band showing relatively minor variation. MONITORING CHANGING RISK The ClimateCast U.S. Hurricane Risk Index provides a timely measure of contemporaneous, forward-looking threat1. By tracking the Index, an analyst can assess how changes in a forecast are influencing the potential for damage and loss. For example, Figure 7 shows the full risk index histories of both Tropical Storm Beryl (top) and 011 s Hurricane Irene (bottom). Tropical Storm Beryl appears as a low plateau followed by two blips on the right of the two-week window (Tropical Storm Alberto is the plateau on the left), indicating that as a relatively weak storm it quickly peaked at a low level of risk, remained at that level the gray uncertainty band showing little variation and quickly returned to negligible risk after landfall on the 8 th. Beryl in fact caused little damage and dissipated soon after making landfall. Hurricane Irene was a different story entirely. From August th to August 8 th the period prior to landfall in North Carolina Irene s forecast track was fairly consistent, but the forecast intensity along the track was in flux, as indicated by the broad width and significant peaks and valleys of the gray uncertainty band. For two days, starting on August th, risk began trending upward as it looked like Irene could impact much of the east coast at hurricane strength and then, from the th onward, it began trending downward (as indicated by the superimposed arrows). FURTHER EXPLORATION: REGIONAL RISK Catastrophe risk management often concerns local risks and niche portfolios. To better service these interests, the ClimateCast Risk Index has been extended to include several regions of the United States, including Florida, the Gulf coast, the coastal states of the southeast and northeast, and the inland Atlantic states. Figure 8 shows the full regional record of Beryl s potential risk the right-most rises in the graphs at the end of May. In Florida and the southeast, the Risk Index can be seen to be both slightly elevated and to have some uncertainty associated with it. This reflects the fact that the ensemble of models forecast Beryl to make landfall in Florida as it did but possibly also in Georgia; potential losses were uncertain because of the uncertain landfall location. Perhaps surprisingly, the Gulf region also exhibits some risk, showing as two low but peaked rises. This expression reflects the fact that the ensemble scenarios produced some outlier projections that had Beryl crossing over Florida and passing into the Gulf states instead of turning north and east, as it eventually did. Over the warm Gulf waters, Beryl possibly might have strengthened and gone on to a Gulf landfall. The northeastern and inland states, however, were never under serious threat. Because Hurricane Irene was expected to brush the entire east coast from North Carolina to New England, even in this instance of a well-predicted track, changes in forecast intensity strongly influenced expected loss because high value exposures along the storm s path were not concentrated uniformly. Figure 8. Risk of insured losses threatened by Tropical Storm Beryl to specific regions of the United States between the 1 th and 8 th of May, 01, as recorded by AIR s ClimateCast Risk Index; the Risk Index can be customized to a particular region, portfolio, or even high-value location such as an industrial facility. (Source: AIR) CUSTOMIZATION How individual companies come to use the capabilities AIR s ClimateCast provides will differ greatly according to each company s specific needs: possibly in logistics planning for claims management, possibly in asset management or short-term risk exchange (such as ILWs), or possibly in briefings to senior management. AIR can customize the Risk Index calculation to satisfy an organization s specific requirements. By using a client s own portfolio, for example, instead of the U.S. Industry Exposure Database (the Risk Index s default), a customized Index can quantify risk to a specific book of business. Figure 7. The stark contrast in risk profile between a relatively weak and short-lived Tropical Storm Beryl (top) and the Category 3 Hurricane Irene (bottom). (Source: AIR) 5
CLOSING THOUGHTS Successful risk management is not a matter of simply managing expected levels of risk. Rather, expected risk levels also must be informed by an assessment of the degree of confidence that can be placed in those estimates. ClimateCast can help managers in this task by providing guidance such as mean and median values concerning both the expected levels of risk and the current degree of confidence (based on a full distribution of potential losses) in those risk levels. With respect to real-time hurricane risk, both expectations and uncertainty levels can change quickly and in unanticipated ways. Following the ClimateCast U.S. Hurricane Risk Index closely and continuously can support real-time risk management efforts and provide an edge over competitors. The ClimateCast U.S. Hurricane Risk Index provides an objective and quantitative means by which changes in expected risk and the corresponding level of uncertainty can be assessed as events unfold. Editor s Note: If you have questions about ClimateCast, please contact your AIR representative or write to climatecast@airworldwide.com. 1 AT ANY GIVEN POINT IN TIME, THE RISK INDEX REFLECTS LOSS POTENTIAL GOING FORWARD. WHEN A STORM HAS ENTIRELY DISSIPATED, THE VALUE OF THE INDEX WILL BE ZERO, EVEN IF IT CAUSED DAMAGE EARLIER IN ITS LIFE CYCLE. AFTER DISSIPATION, A REALIZED INDEX VALUE IS PROVIDED THAT REFLECTS THE CUMULATIVE IMPACT OF THE STORM. ABOUT AIR WORLDWIDE AIR Worldwide (AIR) is the scientific leader and most respected provider of risk modeling software and consulting services. AIR founded the catastrophe modeling industry in 1987 and today models the risk from natural catastrophes and terrorism in more than 90 countries. More than 00 insurance, reinsurance, financial, corporate, and government clients rely on AIR software and services for catastrophe risk management, insurance-linked securities, detailed site-specific wind and seismic engineering analyses, and agricultural risk management. AIR is a member of the Verisk Insurance Solutions group at Verisk Analytics (Nasdaq:VRSK) and is headquartered in Boston with additional offices in North America, Europe, and Asia. For more information, please visit www. air-worldwide.com. 01 AIR WORLDWIDE. ALL RIGHTS RESERVED.