SUMMARY OF 2006 ATLANTIC TROPICAL CYCLONE ACTIVITY AND VERIFICATION OF AUTHOR’S SEASONAL AND MONTHLY FORECASTS

The 2006 Atlantic basin hurricane season had activity at slightly less than average (1950-2000) levels. This activity was much less than predicted in our seasonal forecasts.

By Philip J. Klotzbach[1] and William M. Gray[2]

with special assistance from William Thorson[3]

This forecast as well as past forecasts and verifications are available via the World Wide Web at http://hurricane.atmos.colostate.edu/Forecasts

Emily Wilmsen, Colorado State University Media Representative, (970-491-6432) is available to answer various questions about this verification.

Department of Atmospheric Science

Colorado State University

Fort Collins, CO 80523

Email: amie@atmos.colostate.edu

17 November 2006

---

[1] Research Associate

[2] Professor of Atmospheric Science

[3] Research Associate

---

Acknowledgment

We are grateful to the National Science Foundation (NSF) and Lexington Insurance Company (a member of the American International Group (AIG)) for providing partial support for the research necessary to make these forecasts. We also thank the GeoGraphics Laboratory at Bridgewater State College (MA) for their assistance in developing the Landfalling Hurricane Probability Webpage (available online at http://www.e-transit.org/hurricane).

The second author gratefully acknowledges valuable input to his CSU research project over many years by former graduate students and now colleagues Chris Landsea, John Knaff and Eric Blake. We also thank Professors Paul Mielke and Ken Berry of Colorado State University for much statistical analysis and advice over many years.

---

Notice of Author Changes

By William Gray

The order of the authorship of these forecasts has been reversed from Gray and Klotzbach to Klotzbach and Gray. After 22 years (since 1984) of making these forecasts, it is appropriate that I step back and have Phil Klotzbach assume the primary responsibility for our project’s seasonal, monthly and landfall probability forecasts. Phil has been a member of my research project for the last six years and has been second author on these forecasts for the last five years. I have greatly profited and enjoyed our close personal and working relationships.

Phil is now devoting more time to the improvement of these forecasts than I am. I am now giving more of my efforts to the global warming issue and in synthesizing my projects’ many years of hurricane and typhoon studies.

Phil Klotzbach is an outstanding young scientist with a superb academic record. I have been amazed at how far he has come in his knowledge of hurricane prediction since joining my project six years ago. I foresee an outstanding future for him in the hurricane field. I expect he will make many new forecast innovations and skill improvements in the coming years. I plan to continue to be closely involved in the issuing of these forecasts for the next few years.

---

DEFINITIONS

---

ATLANTIC BASIN SEASONAL HURRICANE FORECASTS FOR 2006

Forecast Parameter and 1950-2000 Climatology (in parentheses)	6 Dec 2005	Update 4 April 2006	Update 31 May 2006	Update 3 Aug 2006	Update 1 Sept 2006	Update 3 Oct 2005	Observed 2006 Total
Named Storms (NS) (9.6)	17	17	17	15	13	11	9
Named Storm Days (NSD) (49.1)	85	85	85	75	50	58	50
Hurricanes (H) (5.9)	9	9	9	7	5	6	5
Hurricane Days (HD) (24.5)	45	45	45	35	13	23	20
Intense Hurricanes (IH) (2.3)	5	5	5	3	2	2	2
Intense Hurricane Days (IHD) (5.0)	13	13	13	8	4	3	3
Net Tropical Cyclone Activity (NTC)* (100%)	195	195	195	140	90	95	85

*NTC is a combined measure of the yearly mean of six indices (NS, NSD, H, HD, IH, IHD) of hurricane activity as a percent deviation from the 1950-2000 annual average.

Figure courtesy of Weather Underground (http://www.weatherunderground.com)

---

ABSTRACT

This report summarizes tropical cyclone (TC) activity which occurred in the Atlantic basin during 2006 and verifies the authors’ seasonal and monthly forecasts of this activity. A forecast was initially issued for the 2006 season on 6 December 2005 with updates on 4 April, 31 May, 3 August, 1 September and 3 October of this year. These forecasts also contained estimates of the probability of U.S. hurricane landfall during 2006. The 3 August forecast included forecasts of August-only, September-only and October-only tropical cyclone activity for 2006. Our 1 September forecast gave a seasonal summary to that date and included individual monthly predictions of September-only and October-only activity. Our 3 October forecast gave a seasonal summary to that date and included an October-November forecast. Our 2006 seasonal hurricane forecast was not successful. We anticipated a well above-average season, and the season had activity at slightly below-average levels. We did catch this downward trend beginning with our early August update We attribute a large portion of this forecast over-prediction to a late-developing El Niño and increased mid-level dryness in the tropical Atlantic.

Our August-only forecast was a bust. Our September-only forecast was quite successful, especially when evaluated against the Net Tropical Cyclone (NTC) activity metric. The October-only forecast also successfully called for activity at well below-average levels, and no tropical cyclone activity occurred after October 2. Our first forecast for the 2007 season will be issued on Friday, 8 December 2006.

---

1 Introduction

A variety of atmosphere-ocean conditions interact with each other to cause year-to-year and month-to-month hurricane variability. The interactive physical linkages between these many physical parameters and hurricane variability are complicated and cannot be well elucidated to the satisfaction of the typical forecaster making short range (1-5 days) predictions where changes in the momentum fields are the crucial factors. Seasonal and monthly forecasts, unfortunately, must deal with the much more complicated interaction of the energy-moisture fields with the momentum fields.

We find that there is a rather high (50-60 percent) degree of year-to-year hurricane forecast potential if one combines 4-5 semi-independent atmospheric-oceanic parameters together. The best predictors (out of a group of 4-5) do not necessarily have the best individual correlations with hurricane activity. The best forecast parameters are those that explain the portion of the variance of seasonal hurricane activity that is not associated with the other variables. It is possible for an important hurricane forecast parameter to show little direct relationship to a predictand by itself but to have an important influence when included with a set of 4-5 other predictors.

In a five-predictor empirical forecast model, the contribution of each predictor to the net forecast skill can only be determined by the separate elimination of each parameter from the full five predictor model while noting the hindcast skill degradation. When taken from the full set of predictors, one parameter may degrade the forecast skill by 25-30 percent, while another degrades the forecast skill by only 10-15 percent. An individual parameter that, through elimination from the forecast, degrades a forecast by as much as 25-30 percent may, in fact, by itself, show little direct correlation with the predictand. A direct correlation of a forecast parameter may not be the best measure of the importance of this predictor to the skill of a 4-5 parameter forecast model. This is the nature of the seasonal or climate forecast problem where one is dealing with a very complicated atmospheric-oceanic system that is highly non-linear. There is a maze of changing physical linkages between the many variables. These linkages can undergo unknown changes from weekly to decadal time scales. It is impossible to understand how all these processes interact with each other. It follows that any seasonal or climate forecast scheme showing significant hindcast skill must be empirically derived.

2 Tropical Cyclone Activity for 2006

Figure 1 and Table 1 summarize the Atlantic basin tropical cyclone activity which occurred in 2006. All the seasonal forecast parameters of NS, NSD, H, HD, IH, IHD and NTC were near their long-period averages. See page 4 for acronym definitions.

3 Individual 2006 Tropical Cyclone Characteristics

The following is a brief summary of each of the named tropical cyclones in the Atlantic basin for the 2006 season. See Fig. 1 for the tracks of these tropical cyclones, and see Table 1 for statistics of each of these tropical cyclones. Online entries from Wikipedia (http://www.wikipedia.org) were very helpful in putting together this tropical cyclone summary.

Figure 1: Tracks of 2006 Atlantic Basin tropical cyclones. Figure courtesy of Weather Underground (http://www.weatherunderground.com).

Table 1: Observed 2006 Atlantic basin tropical cyclone activity.

Highest Category	Name	Dates	Peak Sustained Winds (kts)/lowest SLP (mb)	NSD	HD	IHD	NTC
TS	Alberto	June 11-14	60 kt/995 mb	2.75			2.7
TS	Beryl	July 19-21	50 kt/1001 mb	2.75			2.7
TS	Chris	August 1-4	55 kt/1001 mb	3.25			2.8
TS	Debby	August 23-26	45 kt/1000 mb	3.25			2.8
H-1	Ernesto	August 25 – September 1	65 kt/988 mb	6.00	0.25		6.8
H-1	Florence	September 5-12	80 kt/972 mb	7.50	2.75		9.0
IH-3	Gordon	September 11-20	105 kt/955 mb	9.25	7.50	1.25	24.2
IH-3	Helene	September 14-24	110 kt/954 mb	10.75	7.50	1.75	26.4
H-1	Isaac	September 28-October 2	75 kt/985 mb	4.50	2.00		7.4
Totals	9			50.00	20.00	3.00	84.8

Tropical Storm Alberto: Alberto formed from an area of low pressure in the northwest Caribbean. It was upgraded to the first tropical storm of the 2006 season on June 11 based on aircraft reconnaissance measurements as well as a ship report. Alberto slowly intensified into a strong tropical storm reaching a maximum intensity of 60 knots on June 12. The system entrained dry air as it moved northeastward towards the Florida coastline and began to weaken. It weakened considerably before making landfall near Adams Beach, Florida on June 13 with maximum sustained winds at landfall estimated at 40 knots. Alberto became extratropical the following day. It was not responsible for any direct deaths, and property damage was reported to be minimal.

Tropical Storm Beryl: Beryl formed from an area of low pressure located about 250 miles southeast of the North Carolina coast on July 18. Beryl was upgraded to a tropical storm later on July 18 when aircraft reconnaissance measured 1000-foot flight level winds of 47 knots and a central pressure of 1007 mb. It tracked northward through a break in the sub-tropical ridge and intensified somewhat, reaching a maximum intensity of 50 knots on July 19. The system continued to track slowly northward and began to weaken as it moved over the cooler waters of the North Atlantic. A digging trough over the Great Lakes caused Beryl to accelerate northeastward, and it passed over Nantucket Island, MA early on July 21 with sustained winds at landfall estimated at 40 knots. It continued tracking northeastward and was declared extratropical later on July 21. Beryl was not responsible for any deaths, and it caused minimal damage.

Tropical Storm Chris: Chris formed from a tropical wave while near the Leeward Islands on July 31. It was upgraded to a tropical storm on August 1 due to its appearance in conventional and microwave satellite data. Chris tracked westward and gradually intensified due to favorable upper-tropospheric outflow channels. A ridge to the north of Chris continued to drive the system westward, and it reached its maximum intensity of 55 knots on August 2. It became rapidly disorganized early on August 3 as it encountered strong northerly shear and very dry air. It was downgraded to a tropical depression early on August 4, and it dissipated over Cuba on August 5.

Tropical Storm Debby: Debby formed from a very vigorous tropical wave that moved off the coast of Africa on August 20. It organized quite quickly and was upgraded to a tropical depression late on August 21 as it passed south of the Cape Verde Islands. Despite stable air, the system was upgraded to a tropical storm early on August 23 based on Dvorak satellite estimates as well as several Quikscat passes indicating winds of 30-35 knots near the center of the circulation. A mid-level high steered Debby towards the northwest, and it slowly intensified to its maximum intensity of 45 knots in an environment of light easterly shear. After undergoing a brief weakening trend likely due to dry air, it again strengthened to a 45-knot tropical storm. Southerly shear began to increase as the system moved northwestward, and it weakened to a minimal tropical storm on August 25 due to increasing shear as well as an environment of dry air. An upper-level low continued to impart southerly shear on Debby, and it was downgraded to a tropical depression on August 26. It dissipated later in the day on August 27.

Hurricane Ernesto: Ernesto formed from a tropical wave that was passing through the Windward Islands on August 24. It initially tracked westward due to a mid-level ridge of high pressure. Ernesto strengthened to a tropical storm the following day based on an aircraft reconnaissance report. An upper-level trough to the northwest of Ernesto imparted some southwesterly shear which prevented the system from intensify rapidly. The shear began to weaken as the upper low moved westward, and Ernesto strengthened to a hurricane on August 27. Even though the synoptic environment became much more favorable for intensification at this time, Ernesto slowed over the southwestern part of Haiti, and this interaction with land rapidly took its toll. It was downgraded to a tropical storm later on August 27. It drifted northward towards Cuba and weakened further to a minimal tropical storm before making landfall near Playa Cazonal on August 28. Ernesto moved back over water early on August 29 and began to move northward towards the Florida Peninsula as a shortwave trough displaced the subtropical ridge that was previously positioned over the southeastern United States. Even though thermodynamics were favorable for strengthening of Ernesto over the Florida Straits, the system did not strengthen, likely due to some easterly shear. Ernesto made landfall near Plantation Key, FL early on August 30 as a minimal tropical storm. It weakened to a tropical depression while tracking northeastward over the Florida Peninsula. Once Ernesto emerged back over water off the Florida coastline, it began to intensify over the warm waters of the Gulf Stream. It reached its secondary maximum intensity of 60 knots before making its second United States landfall near Long Beach, North Carolina. Ernesto rapidly dissipated over land on September 1. Over $100 million dollars in total damage was attributed to Ernesto in the United States. Two people died in Florida in traffic accidents caused by heavy rains from Ernesto. Ernesto was also responsible for five deaths in Haiti.

Hurricane Florence: Florence formed from a tropical wave about midway between Africa and the Lesser Antilles on September 3. Southwesterly shear inhibited intensification early in Florence’s life cycle. It became the season’s sixth named tropical cyclone on September 5 due to tropical storm-strength classifications from microwave and conventional satellite imagery. It initially tracked northwestward under a subtropical ridge and strengthened slightly over the next couple of days as it continued to battle southerly shear. Florence was a rather large system and continued to fight against southerly shear as it tracked along the periphery of a subtropical ridge. This shear eventually began to relax, and Florence intensified into the second hurricane of the 2006 season on September 10. The system continued to intensify over warm sea surface temperatures as it tracked towards Bermuda. After passing by Bermuda on September 11, the system began to weaken as it encountered cooler sea surface temperatures and strong upper-level westerlies. It underwent extratropical transition on September 12. Although hurricane-force winds were felt on Bermuda from Florence, there was only minimal damage reported on the island, and no fatalities were reported.

Intense Hurricane Gordon: Gordon formed from a tropical wave early on September 11 while located northeast of the Lesser Antilles. The system was initially under significant northerly shear generated by the upper-level anticyclone surrounding Hurricane Florence. It was upgraded to the seventh tropical storm of the year later on September 11 and began to intensify further as it moved into an area of light shear and warm sea surface temperatures. It turned northward while moving through a break in the subtropical ridge and was classified as a hurricane early on September 13. The system continued to deepen on September 13 and became the first major hurricane of the year early on September 14. Gordon began to turn towards the northeast as it became embedded in the westerlies. It began to weaken later on September 14, and it was only a minimal hurricane with 65 knot winds early on September 16. At this point, steering currents collapsed, and Gordon drifted slowly northeastward across the open Atlantic. Gordon then began to defy the odds and strengthened while traveling over somewhat cooler waters. It reached a secondary peak with 90 knot winds on September 19. Gordon then turned eastward as it became completely embedded in the westerlies and tracked towards the Azores Islands. It passed over the Azores as a Category One hurricane and weakened to a tropical storm on September 20. The system became extratropical later that day. Damage in the Azores from Gordon was minimal.

Intense Hurricane Helene: Helene formed in the far eastern tropical Atlantic from a tropical wave on September 12. It initially tracked westward under the subtropical ridge. Initial intensification was inhibited by a strong mid-level jet; however, it was able to escape from the jet’s influence and became classified as a tropical storm on September 14. Despite being in a low-shear environment, Helene struggled to develop further due to large amounts of dry air penetrating into the cyclonic circulation. It eventually began to enter a more favorable environment with more copious amounts of moisture. Helene became a hurricane on September 16. It began to move towards a break in the subtropical ridge and turned northwestward. At this point, Helene entered a very favorable environment for intensification with warm waters and low vertical wind shear, and it became the second major hurricane of the year on September 18. It maintained major hurricane status for the next couple of days while tracking more westward as a ridge temporarily built to the north of Helene. An upper-level trough soon broke down the ridge, and Helene began to track northward. Some dry air and increased levels of wind shear began to affect the cyclone, and it weakened to a Category 2 hurricane on September 19 and a Category 1 hurricane on September 21. By early on September 23, Helene was becoming less tropical in nature and was downgraded to a tropical storm. It was upgraded back to a hurricane later on September 23 due to satellite-measured hurricane-force winds. An upper-level trough was rapidly approaching Helene at this point, and it underwent extratropical transition on September 24.

Hurricane Isaac: Isaac developed from an area of low pressure in the central Atlantic on September 27. An upper-level low caused some southerly shear over Isaac which initially inhibited intensification. It was upgraded to a tropical storm on September 28 based on a Quikscat pass which indicated that the system had 35-knot winds. Isaac tracked northwestward as it interacted with the upper-level low and strengthened slowly while battling dry air entrainment and continued southerly shear. By later on September 29, the shear began to abate and the surrounding environment became moister, and Isaac strengthened. A strong trough began to curve Isaac more towards the north at this time. It became the fifth hurricane of the year on September 30 and reached its maximum intensity of 75 knots early on October 1. At this point, southwesterly shear and cooler sea surface temperatures began to impact Isaac, and it was downgraded to a tropical storm on October 2. Isaac began to accelerate towards the north and northeast, and it became extratropical later on October 2.

U.S. Landfall. Figure 2 shows the tracks of all 2006 tropical cyclones which impacted the United States. The U.S. was affected by three tropical storms this year: Tropical Storms Alberto, Beryl and Ernesto. Table 2 displays the estimated damage from these three storms. Alberto and Beryl did minimal damage, while Ernesto incurred approximately 100 million dollars in total damage in North Carolina and Virginia. Obviously, the damage in 2006 from tropical cyclones was very minor, especially when compared with the 50+ billion dollars in total damage incurred by the 2004 season and the 100+ billion dollars in total damage incurred by the 2005 season.

Figure 2: Tropical cyclones making U.S. landfall (TS Alberto, TS Beryl and TS Ernesto). Tropical depression strength is indicated by a dotted line while tropical storm strength is indicated by a dashed line.

Table 2: United States damage estimates for the three tropical storms that made U.S. landfall in 2006 (in millions of dollars). We assume that total damage is twice that of insured damage.

Storm Name	Insured Damage	Total Damage (Assumes Twice Insured Damage)
Alberto	Minimal	Minimal
Beryl	Minimal	Minimal
Ernesto	50	100
Total	50	100

4 Special Characteristics of the 2006 Hurricane Season

The 2006 hurricane season had the following special characteristics:

· Another early-starting season. Alberto formed on June 11. The climatological average date for the first named storm formation in the Atlantic, based on 1944-2005 data, is July 10.

· Nine named storms formed during the 2006 season. This is the fewest named storms to form in the Atlantic since 1997, when only seven named storms formed.

· Five hurricanes formed during the 2006 season. This is the fewest hurricanes to form in the Atlantic since 2002, when four hurricanes formed.

· Two major hurricanes formed during the 2006 season. 1997 was the most recent year to have fewer than two major hurricanes form (1 – Erika).

· 50 named storm days occurred in 2006. This is the lowest value of named storm days since 1997, when only 28.75 named storm days occurred.

· 20 hurricane days occurred in 2006. This is the lowest value of hurricane days since 2002, when 10.75 hurricane days were observed.

· 3 intense hurricane days occurred in 2006. This ties 2002 for the lowest value of intense hurricane days observed since 1997, when only 2.25 intense hurricane days occurred.

· Only one hurricane formed during August. This is the fewest hurricanes to form in August since 2002, when no hurricanes formed.

· September 2006’s NTC value was 66. This is the ninth straight September with NTC exceeding the climatological average of 48. The last September with below-average NTC was 1997, when only 28 NTC units were accrued.

· 18.25 hurricane days occurred in September 2006. This is more than were observed in September 2005 (16.75 hurricane days).

· No named storms formed in October. This is the first time that no named storms have formed in October since 2002. Prior to 2006, only eleven years since 1950 witnessed no named storm formations in October.

· Only two named storm days were observed in October (from Isaac which formed in late September). This is the fewest named storm days in October since 1994, when zero named storm days were observed.

· The season accumulated 85 NTC units. This is the lowest NTC value since the 2002 season which accrued 82 NTC units.

· No Category 4 or 5 hurricanes formed in the Atlantic basin this year. This is the first year with no Category 4-5 hurricanes in the Atlantic since 1997.

· Three named storms made United States landfall in 2006. This is the fewest number of named storms to make landfall in the United States since 2001 when three named storms (Allison, Barry and Gabrielle) made landfall.

· This is only the 11^th year since 1945 that no hurricanes have made United States landfall.

· From Alberto-Helene, each tropical cyclone lasted as long or longer than the cyclone preceding it. For example, Alberto and Beryl lasted 2.75 named storm days, Chris and Debby lasted 3.25 named storm days, Ernesto lasted 6 named storm days, etc.

· Both Gordon and Helene accumulated 7.5 hurricane days. These two storms accrued as many hurricane days as Wilma, which was the longest-lived hurricane of the 2005 season.

5 Verification of Individual 2006 Lead Time Forecasts

Table 3 is a comparison of our 2006 forecasts for six different lead times along with this year’s observations. Our seasonal forecasts for 2006 from early December 2005, early April 2006 and late May 2006 were a disappointment. We expected an active season, and the season actually had activity at slightly below-average levels. We did anticipate that these earlier seasonal forecasts were likely somewhat of an over-forecast in our early August and early September updates for the 2006 season. As we will discuss in detail later, we attribute this large over-forecast to a late-developing El Niño and copious amounts of dry air in the tropical Atlantic.

5.1 Preface: Aggregate Verification of our Last Eight Yearly Forecasts

Despite this year’s forecast bust, we are improving our skill in seasonal prediction with an improved level of understanding. This improved skill is demonstrated by the last eight years of our seasonal forecast verifications. Skillful extended range seasonal predictions are indeed possible. With more research, our understanding and skill should continue to improve. We define forecast skill as the degree to which we are able to improve the prediction of the variation of seasonal hurricane activity parameters above that specified by the long-term climatology. Forecast skill is expressed as the ratio of our forecast error to the observed difference from climatology or:

Forecast Error / Seasonal Difference From Climatology

Table 3: Verification of our 2006 seasonal hurricane predictions.

Forecast Parameter and 1950-2000 Climatology (in parentheses)	6 Dec 2005	Update 4 April 2006	Update 31 May 2006	Update 3 Aug 2006	Update 1 Sept 2006	Update 3 Oct 2005	Observed 2006 Total
Named Storms (NS) (9.6)	17	17	17	15	13	11	9
Named Storm Days (NSD) (49.1)	85	85	85	75	50	58	50
Hurricanes (H) (5.9)	9	9	9	7	5	6	5
Hurricane Days (HD) (24.5)	45	45	45	35	13	23	20
Intense Hurricanes (IH) (2.3)	5	5	5	3	2	2	2
Intense Hurricane Days (IHD) (5.0)	13	13	13	8	4	3	3
Net Tropical Cyclone Activity (NTC)* (100%)	195	195	195	140	90	95	85

For example, if there were a year with five more tropical storms than average and we had predicted two more storms than average, we would give ourselves a skill score of 2 over 5 or 40 percent. All predictands show skill in hindcast testing. Table 4 shows our average skill score based on 52 years of hindcasts from 1950-2001.

Table 4: Average variance explained by our hindcasts above that specified by climatology as a function of different forecast lead times (in percent) for the 52-year period of 1950-2001.

Tropical Cyclone Parameter	Early December	Early April	Early June And August
NS	31	31	31
NSD	29	38	39
H	35	36	36
HD	37	40	39
IH	41	40	36
IHD	29	34	35
NTC	44	47	41

Our early December forecasts of the last eight years have had only marginal skill. However, our hindcast skill is quite significant when evaluated over the 1950-2001 period (Table 4).

Another way to consider the skill of our forecasts is to evaluate whether the forecast for each parameter successfully forecast above- or below-average activity. Table 5 displays how frequently our forecasts have been on the right side of climatology for the past eight years. In general, our forecasts are successful at forecasting whether the season will be more or less active than normal by as early as December of the previous year with improving skill as the hurricane season approaches.

Table 5: The number of years that our tropical cyclone forecasts issued at various lead times have correctly predicted above- or below-average activity for each predictand over the past eight years (1999-2006)

Tropical Cyclone Parameter	Early December	Early April	Early June	Early August
NS	6/8	7/8	7/8	6/8
NSD	7/8	8/8	8/8	7/8
H	5/8	6/8	6/8	6/8
HD	5/8	6/8	6/8	7/8
IH	5/8	5/8	7/8	7/8
IHD	5/8	5/8	7/8	7/8
NTC	5/8	6/8	6/8	7/8
Total	38/56 (68%)	43/56 (77%)	47/56 (84%)	48/56 (86%)

Of course, there are significant amounts of unexplained variance in a number of the individual parameter forecasts. Even though the skill for some of these parameter forecasts is somewhat low, especially for the early December lead time, there is a great curiosity in having some objective measure as to how active the coming hurricane season is likely to be. Therefore, even a forecast that is only modestly skillful is likely of interest.

5.2 Predictions of Individual Monthly TC Activity

A new aspect of our climate research is the development of TC activity predictions for individual months. On average, August, September and October have about 26%, 48%, and 17% or 91% of the total Atlantic basin NTC activity. August-only monthly forecasts have now been made for the past seven seasons, and September-only forecasts have been made for the last five seasons. This is the fourth year that we have issued an October-only forecast. For the first time this year, we attempted to predict November activity and issued a joint October-November forecast with our 3 October update.

There are often monthly periods within active and inactive hurricane seasons which do not conform to the overall season. To this end, we have recently developed new schemes to forecast August-only, September-only and October-only Atlantic basin TC activity by the beginning of each of these three months. These efforts have been documented by Blake and Gray (2004) for the August-only forecast and Klotzbach and Gray (2003) for the September-only forecast – see citations and additional reading section.

Quite skillful August-only, September-only and October-only prediction schemes have been developed based on 51 years (1950-2000) of hindcast testing using a statistically independent jackknife approach. Predictors are derived from prior months, usually June and July (NCEP global reanalysis) data for all three (August-only, September-only and October-only) individual monthly forecasts and include August’s data for the early September forecast of September-only and October-only forecasts. We include data through September for our early October forecast. Table 6 gives an outline and timetable of the different forecasts and verifications we issue in early August, early September and early October.

Table 6: Timetable of the issuing of our after-July monthly forecasts (in early August, in early September, and early October), the times of their verification, and the dates of seasonal updates. Note that we make three separate October-only forecasts; two separate September-only forecasts, and one separate August-only forecast. Seasonal updates are issued in early September and early October.

Times of Forecast and Verification	Based on Data Through		Forecasts
Early August	July	August Forecast	September Forecast	October Forecast	Full Season Forecast
Early September	August	August Verification	September Forecast	October Forecast	Remainder of Season Forecast
Early October	September		September Verification	October Forecast	Remainder of Season Forecast

5.3 August-only 2006 Forecast

Our August 2006 forecast was a bust (see Table 7) and was not typical of our previous six August-only forecasts for 2000-2005 or our hindcasts of August-only activity as contained in our original developmental datasets over the period 1949-1999. Our developmental data sets showed considerable skill. Table 8 shows the skill of our prior six August-only forecasts for Net Tropical Cyclone (NTC) activity over the 2000-2005 period. Note that we have correctly predicted above- or below-average activity in five out of the prior six years.

Table 7: CSU forecast and verification of August-only hurricane activity made in early August.

Tropical Cyclone Parameters and 1950-2000 August Average (in parentheses)	August 2006 Statistical Forecast	Adjusted August 2006 Forecast	August 2006 Verification
Named Storms (NS) (2.8)	3.3	4	3
Named Storm Days (NSD) (11.8)	21.1	22	12
Hurricanes (H) (1.6)	2.9	3	1
Hurricane Days (HD) (5.7)	8.1	11	0.25
Intense Hurricanes (IH) (0.6)	0.7	1	0
Intense Hurricane Days (IHD) (1.2)	2.0	3	0
Net Tropical Cyclone Activity (NTC) (26.4)	53.6	50	12

Table 8: Predicted, observed, August-only 2006 forecast (bottom line) and climatological NTC for our six August-only forecasts of 2000-2005. Evaluation of skill with respect to average error and mean square error are also shown.

Year	Observed NTC	Predicted NTC	Climatological NTC
2000	42	33	26
2001	9	22	26
2002	7	18	26
2003	26	22	26
2004	89	35	26
2005	41	50	26
Average Error (2000-2005)		16.7	21.7
Mean Square Error (MSE) (2000-2005)		569	851
Skill of Prediction (relative to MSE)			0.33
2006	12	50	26

August 2006 had about average named storm activity, but the amount of hurricane and intense hurricane activity was well below average. Only one hurricane formed during August (Ernesto), and it lasted less than one day due to interaction with land. On average, about six hurricane days occur during August. Several features likely contributed to an inactive month.

There was considerable subsidence, dry air and dust (A. Evan 2006, personal communication) across the tropical Atlantic during the month of August. Subsidence inhibits the development and maintenance of strong thunderstorms which are necessary for the intensification of easterly waves into tropical cyclones. Two of the three storms that formed during August (Chris and Debby) never reached hurricane strength due partially to very dry air being ingested into their respective circulations. Figure 3 displays a measure of brightness temperature across the tropical Atlantic. Brightness temperatures can be considered a measure of mid-level moisture, with colder temperatures indicating more moisture. Note that brightness temperatures were well above average (i.e. less moisture) throughout most of the month of August.

Another factor that may have played a role in reducing Atlantic basin hurricane activity during August was the development of El Niño conditions in the tropical Pacific. Conditions rapidly trended towards El Niño during August. In general, associated with El Niño conditions, is a drier Caribbean and western tropical Atlantic and increased vertical wind shear across the entire tropical Atlantic/Caribbean area. See Section 7.1 for a more in-depth discussion of ENSO.

Figure 3: Water vapor brightness temperature across the tropical Atlantic from January-August. Note that brightness temperatures were warmer than average for most of August. Brightness temperatures are a proxy for mid-level moisture, with cooler temperatures indicating more moisture. Figure courtesy of the Cooperative Institute for Research in the Atmosphere (CIRA).

5.4 September-only 2006 Forecast

Our September 2006 forecast verified quite well (see Table 9). Even though conditions in August were not favorable for Atlantic basin tropical cyclone activity, we predicted that they would likely become more favorable for hurricane development in September, and this prediction verified very well. Dry air and African dust intrusions (A. Evan 2006, personal communication) continued to predominate across the tropical Atlantic in September; however, vertical wind shear was below average for most of the month. Four named storms (Florence, Gordon, Helene and Isaac) formed during September, and all four of these storms became hurricanes. Gordon and Helene became major hurricanes.

Table 9: Independent September-only forecasts for 2006 including the 3 August statistical forecast for September, the 3 August adjusted forecast for September, the 1 September statistical forecast for September and the 1 September adjusted forecast for September. Observed activity is in the far right-hand column.

Tropical Cyclone Parameters and 1950-2000 September Average (in parentheses)	3 Aug. Stat. Fcst. (for Sep.)	3 Aug. Adjusted Forecast	1 Sep. Stat. Fcst. (for Sep.)	1 Sep. Adjusted Forecast	Observed Sep. 2006 Activity
Named Storms (NS) (3.4)	4.1	5	3.4	5	4
Named Storm Days (NSD) (21.7)	20.8	25	17.2	20	30.50
Hurricanes (H) (2.4)	2.2	3	3.2	3	4
Hurricane Days (HD) (12.3)	7.2	12	5.5	10	18.25
Intense Hurricanes (IH) (1.3)	1.8	2	1.7	2	2
Intense Hurricane Days (IHD) (3.0)	1.5	5	2.5	4	3
Net Tropical Cyclone Activity (NTC) (48%)	48	60	45	59	66

September had above-average activity when evaluated by the NTC metric. This represents the ninth consecutive September that has had above normal NTC activity. September 2006 accrued 66 NTC units, which is somewhat more than the 1950-2000 average of 48. Although El Niño conditions continued to develop in the central and eastern Pacific, vertical wind shear in the tropical Atlantic was actually somewhat below average. Figure 4 shows vertical wind shear across the tropical Atlantic (0-20ºN, 20-60ºW) for 2006 in the tropical Atlantic from June-September. Note that values in September 2006 were generally below the long-term average. Atlantic basin sea surface temperatures also remained above average throughout the month. We think that the likely inhibiting factor that kept September from being a very active month was the continued predominance of dry air in the tropical Atlantic. Figure 5 displays water vapor brightness temperatures across the tropical Atlantic (0-20ºN, 20-60ºW). Note that brightness temperatures remained above average (i.e. less moisture) throughout most of the month of September, as they had in August. The continued dominance of subsidence across the tropical Atlantic may in part have been due to a shift in the Walker Circulation associated with the developing El Niño.

We consider our September monthly forecast to have been a success. We predicted that despite an inactive early season, we would see above-average activity in September, and this is what occurred. Our forecast predicted that three hurricanes and two major hurricanes would develop during September, and four hurricanes and two major hurricanes formed.

Figure 4: Vertical wind shear (850-200 mb) across the tropical Atlantic (0-20ºN, 20-60ºW) from June-September. Note that vertical wind shear values in September have generally been below the long-period average. Figure courtesy of the Cooperative Institute for Research in the Atmosphere (CIRA) from the Tropical Cyclone Formation Probability Product (DeMaria et al. 2001).

Figure 5: Water vapor brightness temperature across the tropical Atlantic (0-20ºN, 20-60ºW) from January-September. Note that brightness temperatures remained above average for most of September. Brightness temperatures are a proxy for mid-level moisture, with cooler temperatures indicating more moisture. Figure courtesy of the Cooperative Institute for Research in the Atmosphere (CIRA) from the Tropical Cyclone Formation Probability Product (DeMaria et al. 2001).

5.5 October-only 2006 Forecast

The October-only forecast successfully called for an inactive month; however, we did not expect it to be as inactive as it turned out to be. No named storms formed in the Atlantic after Isaac developed on September 28. The only activity that occurred during the month was early in October as Isaac dissipated over the open Atlantic. Since vertical wind shear tends to be heightened in El Niño years, and this year has seen a rapid transition towards El Niño in August-October, we expected to see a large amount of vertical wind shear across the tropical Atlantic this October (see Figure 6). Anomalously strong vertical wind shear tends to bring an early-season end to tropical cyclone activity in the tropical Atlantic, and this is what occurred this year. No new named storms formed in the Atlantic in October. In our early October update for the remainder of the season, we also added November to our October forecast, since tropical cyclone development in November in El Niño years is very rare. We have not seen any development of tropical cyclones in November. Table 10 displays the statistical and adjusted October-only forecasts issued on 3 August, 1 September, and 3 October respectively as well as the observed activity that occurred in October 2006. We are planning to revise our future October-only statistical forecasts to include a more explicit treatment of El Niño.

Figure 6: Vertical wind shear (850-200 mb) across the tropical Atlantic (0-20ºN, 20-60ºW) from January-October. Note that vertical wind shear values in October have generally been above the long-period average. Figure courtesy of the Cooperative Institute for Research in the Atmosphere (CIRA) from the Tropical Cyclone Formation Probability Product (DeMaria et al. 2001).

Table 10: Independent October-only forecasts for 2006 including the 3 August statistical forecast for October, the 3 August adjusted forecast for October, the 1 September statistical forecast for October, the 1 September adjusted forecast for September, the 3 October statistical forecast for October and the 3 October adjusted forecast for October-November. Observed activity is in the far right-hand column.

TC Parameters and 1950-2000 Oct. Clim. (in parentheses)	3 Aug. Stat. Fcst. (for Oct.)	3 Aug. Adjusted Forecast	1 Sep. Stat. Fcst. (for Oct.)	1 Sep. Adjusted Forecast	3 Oct. Stat. Fcst. (for Oct.)	3 Oct. Adjusted Forecast for Oct-Nov	Observed Oct-Nov 2006 Activity
NS (1.7)	1.4	2	2.6	2	3.2	2	0
NSD (9.0)	7.3	11	13.5	10	16.9	10	2
H (1.1)	0.9	1	1.7	1	2.1	1	0
HD (4.4)	3.6	4	6.6	3	8.3	4	1.5
IH (0.3)	0.2	0	0.5	0	0.6	0	0
IHD (0.8)	0.6	0	1.2	0	1.5	0	0
NTC (17%)	15	15	27	12	32	12	2

6 Verification of 2006 U.S. Landfall Probabilities

A new initiative in our research involves efforts to develop forecasts of the seasonal probability of hurricane landfall along the U.S. coastline. Whereas individual hurricane landfall events cannot be accurately forecast, the net seasonal probability of landfall (relative to climatology) can be forecast with statistical skill. With the premise that landfall is a function of varying climate conditions, a probability specification has been accomplished through a statistical analysis of all U.S. hurricane and named storm landfalls during a 100-year period (1900-1999). Specific landfall probabilities can be given for all tropical cyclone intensity classes for a set of distinct U.S. coastal regions. Net landfall probability is statistically related to the overall Atlantic basin Net Tropical Cyclone (NTC) activity and to climate trends linked to multi-decadal variations of the Atlantic Ocean thermohaline circulation (as measured by North Atlantic SSTA). Table 11 gives verifications of our landfall probability estimates for 2006.

Landfall probabilities for the 2006 hurricane season were estimated to be well above their climatological averages; however the season actually recorded below-average landfall activity. Three tropical storms made landfall this year (Alberto, Beryl and Ernesto). Fortunately, no hurricanes made landfall along the United States coastline this year. This is only the 11^th year since 1945 that no hurricanes made landfall along the U.S. coastline.

This is the first year that we have attempted to analyze landfall steering current patterns in an attempt to determine whether the Gulf Coast or East Coast was more likely to be targeted by tropical cyclones this year. Our analysis of steering current patterns correctly predicted that the Gulf Coast would likely not be targeted by many storms this year; however, we thought that the East Coast would be a likely target for tropical cyclone activity. We believed that there would tend to be a ridge along the East Coast this summer/fall, as there tends to be a moderate positive correlation between ridging in the northeast United States/eastern Canada in April-May and East Coast ridging in August-October. However, this relationship did not hold for this year. Ridging in the northeast United States and Canada in April-May of 2006 was replaced by a mid-level trough along the East Coast for most of the Atlantic basin hurricane season, and most storms recurved. Figure 7 displays mid-level (500 mb) height anomalies in April-May in the northeast United States and eastern Canada, while Figure 8 displays mid-level (500 mb) height anomalies in August-September 2006 along the East Coast of the United States from North Carolina northward. Note that the ridging along the East Coast in Figure 7 was displaced by a trough in Figure 8.

Figure 7: 500 mb geopotential height anomaly field for April-May in the northeast United States and eastern Canada.

Figure 8: 500 mb geopotential height field anomaly for August-September along the East Coast of the United States from North Carolina northward.

Active research continues on our landfall probability technique, and full documentation of the methodology for estimating hurricane landfall probability is being prepared. Landfall probabilities include specific forecasts of the probability of landfalling tropical storms (TS) and hurricanes of category 1-2 and 3-4-5 intensity for each of 11 units of the U.S. coastline (Figure 9). These 11 units are further subdivided into 55 subregions based on coastal population density, and these subregions are further subdivided into 205 coastal and near-coastal counties. The climatological and current-year probabilities are now available online via the United States Landfalling Hurricane Probability Webpage at http://www.e-transit.org/hurricane. Since the website went live on June 1, 2004, the webpage has received over half-a-million hits.

Figure 9: Location of the 11 coastal regions for which separate hurricane landfall probability estimates are made.

Table 11: Estimated forecast probability (percent) of one or more U.S. landfalling tropical storms (TS), category 1-2 hurricanes, and category 3-4-5 hurricanes, total hurricanes and named storms along the entire U.S. coastline, along the Gulf Coast (Regions 1-4), and along the Florida Peninsula and the East Coast (Regions 5-11) for 2006 at various lead times. The mean annual percentage of one or more landfalling systems during the 20^th century is given in parentheses in the 3 August forecast column. Table (a) is for the entire United States, Table (b) is for the U.S. Gulf Coast, and Table (c) is for the Florida Peninsula and the East Coast.

(a) The entire U.S. (Regions 1-11)
Forecast Date
	6 Dec.	4 Apr.	31 May	3 Aug.	Observed Number
TS	91%	91%	94%	85% (80%)	3
HUR (Cat 1-2)	88%	88%	90%	67% (68%)	0
HUR (Cat 3-4-5)	81%	81%	82%	73% (52%)	0
All HUR	98%	98%	95%	91% (84%)	0
Named Storms	99%	99%	99%	99% (97%)	3
(b) The Gulf Coast (Regions 1-4)
Forecast Date
	6 Dec.	4 Apr.	31 May	3 Aug.	Observed Number
TS	74%	74%	66%	57% (59%)	1
HUR (Cat 1-2)	61%	61%	44%	33% (42%)	0
HUR (Cat 3-4-5)	47%	47%	38%	26% (30%)	0
All HUR	79%	79%	62%	51% (61%)	0
Named Storms	95%	95%	86%	79% (83%)	1
(c) Florida Peninsula Plus the East Coast (Regions 5-11)
Forecast Date
	6 Dec.	4 Apr.	31 May	3 Aug.	Observed Number
TS	64%	64%	85%	64% (51%)	2
HUR (Cat 1-2)	69%	69%	83%	47% (45%)	0
HUR (Cat 3-4-5)	64%	64%	69%	64% (31%)	0
All HUR	89%	89%	87%	81% (62%)	0
Named Storms	96%	96%	94%	93% (81%)	2

7 Why Was the 2006 Atlantic Basin Season Over-Forecast?

As can be seen from the table outlining our predictions on page 5, we considerably over-forecast activity that occurred in the 2006 Atlantic basin hurricane season. We thought that the season would be very active, approximately in line with the average activity that we have experienced since the start of the most recent positive phase of the Atlantic Multi-decadal Oscillation (AMO) (1995-2005). Instead, this season ended up with activity at slightly below the 1950-2000 mean, with only nine named storms, five hurricanes and two major hurricanes forming. In the next few pages, we discuss some of the features that developed during the Atlantic basin season that likely caused the season to be much less active than we anticipated.

7.1 ENSO

One of the extraordinary features of the 2006 Atlantic basin hurricane season has been the rapid onset of El Niño conditions in the tropical Pacific. The warming of the eastern and central Pacific during August through October 2006 has been truly remarkable. Only 1997 witnessed a larger temperature increase in Nino 3 anomalies from June-July to August-September than did the 2006 season. But, in 1997, June-July Nino 3 anomalies (2.1ºC) were already well above average while 2006 June-July anomalies (0.1ºC) were not. This was by far the largest percentage warming of SST anomalies between June-July and August-September in the tropical Pacific for a year that had El Niño conditions in August-September. For this comparison, we define El Niño years as those with Nino 3 temperatures that averaged greater than or equal to 0.5ºC from August-September.

In 2006 sea surface temperatures in Nino 3 warmed by approximately 0.6ºC from their June-July values to their August-September values. This is the largest percentage increase in anomalies (700%) from June-July values to August-September values. The second largest percentage increase (500%) in SST anomalies in Nino 3 during this same time period was in 1979. Table 12 displays June-July and August-September Nino 3 values for the fifteen years with the warmest readings in Nino 3 during August-September.

Table 12: June-July Nino 3 temperatures, August-September Nino 3 temperatures and the percentage change of anomalies from June-July to August-September for the fifteen years that were classified as El Niño based on August-September Nino 3 anomalies >= 0.5ºC.

Year	June-July (JJ) Nino 3 (ºC)	Aug-Sep (AS) Nino 3 (ºC)	Anomaly Percentage Change (AS/JJ)
1997	2.1	3.0	143%
1972	1.3	1.7	131%
1987	1.4	1.7	121%
1982	1.0	1.5	150%
1965	1.0	1.2