This report summarizes tropical cyclone (TC) activity which occurred in the Atlantic basin during 2000 and verifies the authors' seasonal forecasts of this activity which were initially issued on 8 December 1999, with updates on 7 April, 4 June and 4 August of this year. The 4 August forecast also contained our first attempt at forecasting August-only tropical cyclone activity. All of these forecasts verified well. The 2000 hurricane season was characterized by enhanced levels of tropical cyclone activity but with no U.S. hurricane landfalls. A total of 14 named storms (average is 9.3) and 8 hurricanes (average is 5.8) occurred and persisted for a total of 32 hurricane days (average is 24). There were 3 major hurricanes of Saffir/Simpson intensity category 3-4-5 (average is 2.3) with 5.25 intense hurricane days (average is 4.7). The seasonal total of named storm days was 66 which is 141 percent of the long-term average. Net tropical cyclone (NTC) activity was 134 percent of the 1950-1990 average and 212 and 178 percent of average for the recent periods between 1990-94 and 1970-94, respectively. All of our forecast parameters were close to what occurred. We consider this to have been a successful forecast year.

1  Introduction

      The Atlantic basin (including the Atlantic Ocean, Caribbean Sea, and Gulf of Mexico) experiences more year-to-year hurricane variability than occurs in any of the other global tropical cyclone basins. The number of Atlantic basin hurricanes per season in recent years has ranged as high as 12 (as in 1969), 11 (as in 1950 and 1995), 10 (as in 1998), and as low as 2 (as in 1982) and 3 (as in 1997, 1994, 1987, 1983, 1972, 1962, 1957). Until the mid 1980s there were no objective methods for predicting whether forthcoming hurricane seasons were likely to be active, inactive, or near normal. Recent ongoing research by the authors (see Gray, 1984a, 1984b, 1990; Landsea, 1991; Gray et al., 1992, 1993a, 1994) indicates that there are surprisingly skillful 3-to-11 month (in advance) predictive signals for Atlantic basin seasonal hurricane activity. This research now allows us to issue extended-range forecasts in early December for next year's Atlantic basin hurricane activity with updates in early April, early June, and early August. The purpose of this end-of-season report is to compare these forecasts with actual observed hurricane activity during the 2000 hurricane season.

2  Factors Known to be Associated With Atlantic Seasonal Hurricane Variability

      Our forecasts, which are issued at several lead times prior to each hurricane season, are based on the current values of indices which are derived from various global and regional predictive factors which the authors have shown to be related to subsequent seasonal variations of Atlantic basin hurricane activity. Figures 1-3 provide a summary of the geographic locations from which the various forecast parameters are obtained. Our forecast methodology emphasizes the analysis of prior oceanic and atmospheric precursor conditions which are observed to be associated with the amount of hurricane activity during the following season. These predictors include the following:

(a) El Niño-Southern Oscillation (ENSO): El Niño events are characterized by anomously warm sea surface temperature anomalies in the eastern equatorial Pacific areas termed Nino 1-2, 3, 3.4 and 4 (Fig. 1), a negative value of the Tahiti minus Darwin surface pressure gradient and enhanced equatorial deep convection near the Dateline. These conditions alter the global atmospheric circulation fields contributing to anomalous upper-level westerly winds over the Atlantic basin which typically reduce Atlantic basin hurricane activity. Conversely, during La Niña seasons, anomalously cold sea surface temperatures are present together with high values of Tahiti minus Darwin surface pressure difference and reduced deep equatorial convection near the Dateline are associated with enhanced Atlantic basin hurricane activity.

(b) The stratospheric Quasi-Biennial Oscillation (QBO). The QBO refers to the variable east-west oscillating stratospheric winds which encircle the globe near the equator. Other factors being equal, more intense (category 3-4-5) Atlantic basin hurricane activity occurs during seasons when equatorial stratospheric winds at 30 mb and 50 mb (23 and 20 km altitude, respectively) are from a westerly (versus easterly) direction.

(c) African Rainfall (AR): The incidence of intense Atlantic hurricane activity is enhanced when rainfall during years when prior August-September Western Sahel region is above average and when August-November Gulf of Guinea region during the prior year is also above average. The June-July rainfall is also a predictor for August through October hurricane activity. Other factors being equal, hurricane activity is typically suppressed if the rainfall in the prior year (or season) in these two regions is below average.

(d) Prior Year October-November and March northeast Atlantic Subtropical Ridge Strength
(ONR). When this pressure ridge is anomalously weak during the prior autumn and spring periods, eastern Atlantic trade winds are weaker. A weak ridge condition is associated with decreased mid-latitude cold water upwelling and advection off the northwest African coast, as well as to decreased evaporative cooling rates in this area of the Atlantic. In this way, a weak ridge leads to warmer sea surface temperatures which typically persist into the following summer period and contribute (other factors being constant) to greater seasonal hurricane activity. Conversely, less hurricane activity occurs when the October-November and spring pressure ridge is anomalously strong.

(e) Atlantic Sea Surface Temperature Anomalies (SSTA) in the three regions [(MATL; 30-50^°N,
10-30^°N and TATL; 6-22^°N, 18-60^°W) during April through June] and [NATL; 50-60^°N, 10-50^°W
and TATL during January through March]: [See Fig. 3 (bottom) for the location of these areas]. Warmer SSTAs in these areas enhance deep oceanic convection and, other factors aside, provide conditions more conducive for Atlantic tropical cyclone activity; cold water temperatures the reverse.

(f) Caribbean Basin Sea Level Pressure Anomaly (SLPA) and upper tropospheric (12 km)
Zonal Wind Anomaly (ZWA): Spring and early summer SLPA and ZWA have moderate predictive potentials for hurricane activity occurring during the following August through October months (Fig. 3). Negative anomalies (i.e., low pressure and easterly zonal wind anomalies) imply enhanced seasonal hurricane activity (easterly 200 mb) while positive values imply suppressed hurricane activity (westerly 200 mb shear).

(g) Influence of West Africa west-to-east surface pressure and temperature gradients (D PT):
Anomalously strong west-to-east surface pressure and temperature gradients across West Africa between February and May are typically correlated with the hurricane activity which follows later in the year.

Our various lead-time forecast schemes are created by maximizing the pre-season forecast skill from a combination of the above predictors, for the period 1950-1997. We also use an analog methodology whereby we look for those years with specific precursor climate signals strongly similar to the current forecast year whereby, the recurrence of similar TC activity is likely.

3  Statistical Summary of 2000 Atlantic Tropical Cyclone Activity

      The 2000 Atlantic hurricane season officially ends on 30 November. To date, there have been eight hurricanes and 32 hurricane days during the 2000 season. The total named storms (i.e., the number of hurricanes plus tropical storms) was 14, yielding 66 named storm days. There were three intense or major (category 3-4-5) hurricanes this season. All designated tropical cyclone activity parameters exceeded the long-term average. Figure 4 and Table 1 show the tracks and give statistical summaries, respectively, for the 2000 season. Table 2 characterizes 2000 seasonal tropical cyclone activity in terms of long-term average annual percentages for the 1950-1990, 1970-94 and 1990-94 periods. Note that 2000 hurricane activity was much above the typical seasonal averages of 1970-1994 and of the more recent 1990-1994. The biggest changes have occurred for the most intense cyclones. Figure 5 shows the U.S. landfalling named storms of 2000.

By all measures, 2000 was a very active hurricane year. Only the number of U.S. hurricane landfall was below the long period average.

4  Individual 2000 Tropical Cyclone Characteristics

Hurricane Alberto: Hurricane Alberto formed in the far eastern Atlantic on 3 August. Alberto briefly attained hurricane status on 5 August as it was moving west-northwestward. It weakened below hurricane strength for two days before turning to the north. Alberto reached a maximum intensity of 125 mph unusually far to the north (poleward of 35N) and took an eastward course across the North Atlantic, weakening to a tropical storm. It then unexpectedly executed a large clockwise loop between Bermuda and the Azores, re-strengthening into a hurricane before recurving out to sea. Alberto maintained tropical characteristics north of 52N, another unusual event. It will be remembered most for its longevity, becoming the longest-lived August named storm on record. It was a tropical storm for over 19 days, becoming the third longest-lasting tropical storm on record for any month in the Atlantic basin.

Tropical Storm Beryl: Tropical Storm Beryl formed in the southwestern Gulf of Mexico on the 13th of August. It was a large system with multiple centers, which may have inhibited its intensification. It drifted westward and made landfall in northeast Mexico on the 15th, reaching a peak intensity of 50 mph.

Tropical Storm Chris: Tropical Storm Chris developed from a tropical wave about 600 miles east of the Lesser Antilles on the 17th of August, briefly attaining tropical storm status on the 18th before dissipating the next day.

Hurricane Debby: Hurricane Debby formed from a large tropical wave on the 18th of August about eleven hundred miles east of the Windward Islands. It became a tropical storm the next day and a hurricane on the 21st before impacting the Lesser Antilles. The storm passed through the Northeast Caribbean Islands on the 22nd, reaching a maximum intensity of 85 mph. It then moved north of Puerto Rico and turned to the west, skirting the coast of the Dominican Republic while weakening back to a tropical storm. Debby dissipated after moving south of Eastern Cuba on the 24th. Debby was at one time predicted to reach category 4 status. It was temporarily viewed as a great threat to South Florida. Fortunately this did not materialize.

Tropical Storm Ernesto: Tropical Storm Ernesto developed from a tropical wave about 850 miles east of the Lesser Antilles on the 1st of September, attaining tropical storm intensity for about a day on the 2nd before dissipating a day later.

Hurricane Florence: Hurricane Florence formed from a mid-latitude baroclinic system about 375 miles south-southeast of Cape Hatteras on the 11th of September. It then moved very little over the next three days, briefly reaching hurricane status on the 12th. Florence then moved to the northeast and re-intensified into a hurricane on the 15th, bringing tropical storm force winds to Bermuda. The hurricane merged with an extratropical low on the 17th over the North Atlantic.

Hurricane Gordon: Hurricane Gordon was the first tropical cyclone of the year to affect the U.S. It formed just north of the Yucatan Channel from a tropical wave on the 14th of September. It drifted northward and became a hurricane about 300 miles south of Apalachicola on the 19th. Strong vertical wind shear then weakened Gordon into a tropical storm before it made landfall near Cedar Key, Florida on the evening of the 17th. It then merged with a cold front and dissipated shortly after landfall.

Tropical Storm Helene: Tropical Storm Helene was first observed as a tropical depression about 475 miles east of the Leeward Islands on the 15th of September. It then weakened into a tropical wave the next day and continued moving westward across the eastern Caribbean Sea. It then reformed as a tropical depression on the 19th just northeast of Grand Cayman Island. It then moved northwest toward the Gulf of Mexico and strengthened into a tropical storm shortly after exiting Cuba. Helene reached a maximum intensity of 70 mph on the 22nd. Helene weakened significantly before landfall as a minimal tropical storm near Fort Walton Beach, Florida on the morning of the 23rd. The remnants of Helene emerged into the Atlantic Ocean two days later, re-strengthening into a strong tropical storm before becoming extratropical in the North Atlantic.

Hurricane Isaac: Hurricane Isaac was the second major hurricane of the season. It developed from a tropical wave that was a couple hundred miles south of the Cape Verde Islands. It reached tropical storm strength on the 22nd of September as it was moving to the west-northwest. It rapidly increasing to major hurricane strength the next day. Isaac's peak intensity was 140 mph on the 28th as it was moving northward over the open ocean. Isaac was quite large as it recurved well east of Bermuda as a classic North Atlantic hurricane, transitioning into a strong extratropical storm that affected the British Isles.

Hurricane Joyce: Hurricane Joyce formed southwest of the Cape Verde Islands on the 25th of September. It became a hurricane a couple of days later, reaching a peak intensity of 90 mph on the 28th. Joyce weakened steadily as it progressed westward at a low latitude, affecting portions of the southern Windward Islands as a minimal tropical storm. The storm dissipated on the 2nd of October just north of the South American coast.

Hurricane Keith: Hurricane Keith developed over the northwest Caribbean Sea on the 28th of September and slowly moved northwest toward the Yucatan Peninsula, becoming a tropical storm the next day. The system rapidly intensified on the 30th into a major hurricane with sustained winds of 140 mph. Keith battered the coastal islands of Belize and the Yucatan as it slowly drifted westward into Central America. It weakened to a tropical depression while crossing the Yucatan Peninsula, re-strengthening into a tropical storm in the Bay of Campeche on the 4th of October. Keith reached an intensity of 90 mph shortly before making landfall just north of Tampico, Mexico. It then dissipated over the higher terrain of northeastern Mexico the next day.

Tropical Storm Leslie: Tropical Storm Leslie formed just off the east coast of Florida on the 4th of October from a subtropical depression. The precursor disturbance to Leslie caused extremely heavy rain over South Florida with flood damage estimated at 700 million dollars. Leslie gained tropical characteristics on the 5th but remained only a minimal tropical storm. It then became extratropical on the 7th while northwest of Bermuda.

Hurricane Michael: Hurricane Michael developed from a baroclinic low pressure system northeast of Bahamas. After the system moved northward for a few days, it quickly acquired tropical characteristics to become a hurricane on the 17th while racing north-northeast. It reached a maximum intensity of 100 mph east of Sable Island, Nova Scotia. Michael weakened slightly to a minimal hurricane before landfall along the southern coast of Newfoundland on the 19th and then was absorbed into an extratropical low pressure system.

Tropical Storm Nadine: Tropical Storm Nadine formed from a tropical wave a few hundred miles southeast of Bermuda on the 19th of October. It became a tropical storm the next day, reaching a peak intensity of 60 mph. Nadine then accelerated toward the northeast and merged with a mid-latitude system over the open ocean.

5  Variation of 2000 Forecast Parameters

      Specific trends in the factors which we know to be associated with seasonal variation of hurricane activity during 2000 include the following:

a) ENSO Conditions. Equatorial Pacific SSTAs (in ^°C) in Niño-1-2, 3, 3.4 and 4 (see Fig. 1 for locations) are shown in Table 3. Cool water conditions were present throughout the season. In addition, the Tahiti minus Darwin surface pressure difference (the Southern Oscillation Index, SOI) was generally positive (as typical of cool ENSO conditions) while equatorial Outgoing Longwave Radiation (OLR) values near the Dateline were high, indicating diminished deep convection in that area. The forgoing were all favorable ENSO-linked conditions for the enhancement of this year's hurricane activity. Our ENSO forecast of early December 1999 for no El Niño during 2000 verified.

b) Stratospheric QBO Winds

Table 4 shows both the absolute and relative (i.e., anomaly) values of 30 mb (23 km) and 50 mb (20 km) stratospheric QBO zonal winds near 12^°N during March through October 2000. We had projected both 30 and 50 mb QBO winds to be from a relative easterly direction and, as such, would be an inhibiting feature for this year's activity. At the height of the 2000 season (in September), 50 mb QBO wind anomalies were westerly but were easterly at 30 mb, this increased the wind shear between the two levels. We judge this shear to have been a weak negative influence on this year's activity while the westerly winds at 50 mb were a positive influence, the net QBO influence likely neutral.

c) Sea-Level Pressure Anomaly (SLPA)

Table 5 gives information on regional Caribbean basin and Gulf of Mexico SLPA during the 2000 season. Caribbean SLPA was near neutral during the August through September period. The absence of high pressure prevented surface pressure from being an inhibiting influence on this season's conditions. Knaff's (1997) Atlantic SLPA forecast scheme predicted slightly below average SLPA for 2000. The forecast did not verify as well as last year. The base period of these pressure anomolies is 1950-1985.

d) Zonal Wind Anomalies (ZWA)

Table 6 shows that the average upper tropospheric (12 km or 200 mb) ZWA during the critical months of August and September were slightly negative. Above average ZWA were present in October. The negative average August-September ZWA values reduce regional tropospheric vertical wind shear and were a factor in explaining the active 2000 season. Negative ZWA conditions allowed the westward moving easterly waves from Africa to experience less vertical wind shear and become better organized.

e) African Western Sahel Rainfall in 2000

Summer rainfall in the Western Sahel region of Africa turned out to be significantly below average during June through September (-0.75 SD) despite the generally active hurricane season. Watching daily satellite loops of infrared (deep convective storms) imagery, one tended to conclude that the western Sahel ITCZ cloudiness and rainfall were not below average. But, as recently pointed out by S. Nicholson (1999), satellite estimates of rainfall appear to overestimate rainfall (wet bias) as compared with the available rain gauge data (such as it is!).

We have, so far, been unable to explain this lack of positive Sahel rainfall and major hurricane association since 1995. Our previous research has shown (Gray 1990, Landsea 1991, Landsea and Gray 1992, Landsea et al. 1992) a strong relationship between western Sahel rainfall and Atlantic basin major hurricane activity. Clearly, this topic requires further study.

6  Special Characteristics of the 2000 Season

      The 2000 season included the following special features:

1. Three major hurricanes. This year saw the persistence of the recent strong upturn in major hurricane activity which began in 1995. The years 1995-2000 are the most active six-year period on record and 1998-2000 constitute the most active three-year period in terms of total hurricanes (26). It must be pointed out, however, that pre-World War II records are less reliable for weaker cyclone systems and those in the Central Atlantic.
2. An exceptional outbreak of tropical cyclone activity occurred during September. Six tropical cyclones formed within a 19-day period during this month.
3. No U.S. hurricane landfall events occurred. Only two tropical storms (Gordon and Helene) made landfall on the U.S. coast (Fig. 5). This was the first year since 1994 when a hurricane has not crossed the U.S. coastline.
4. This season saw no named storms form during June, July or November.
5. Seven named storms formed in September. Only four prior months have had this many named storms in a month and 1995 and 1933 (for August) and 1998 and 1949 (for September).
6. There was quite an enhancing high latitude hurricane activity in the central Atlantic (Alberto, Florence, Isaac, Leslie, Michael and Nadine).
7. Several African origin named storms that progressed westward at low latitudes did not intensify into significant hurricanes. This is not typical. Those weakening and dying in the low latitudes over tropical Atlantic included Chris, Debbie, Ernesto and Joyce. The weakening of these low latitude storms was due to the persistence of a strong Tropical Upper Tropospheric Trough (TUTT) in the West Atlantic.
8. The Madden-Julian Oscillation (MJO) appeared to exercise a stronger than usual intra-seasonal influence on this season's activity. This can be inferred from the concentration of named storm formation in the seven days between August 13-19 (four systems), in the 19 days between 9-28 September (six systems), and the three days of 17-19 October (two systems). This approximate monthly time-scale, up-and-down variation can also be found in the monthly variation in SLPA and ZWA (Tables 5 and 6).

7  Verification of Individual 2000 Lead Time Forecasts

      Table 7 shows our forecasts for 2000 at four different lead times with this year's observed numbers. Note that we consistently forecast an active 2000 hurricane season. Beginning on 4 December 1999, we held to this forecast in our subsequent 7 April, 4 June and 6 August updates. We consider this year to have been one of our better Atlantic basin seasonal forecasts.

Our forecast of an above average probability of U.S. hurricane landfall did not materialize however. Landfall probability is a different type of forecast, which is not expected to verify well in individual years. It must be judged over periods of 4-5 years. Whereby, there are almost always higher numbers of landfalling hurricanes during 4-5 active seasons in comparison with 4-5 inactive seasons.

The following are quotes from our initial and updated 2000 seasonal forecasts, all of which verified.

From our initial 8 December 1999 forecast:

``we are predicting that there will be no El Niño event next year (i.e., 2000). Rather, the current La Niña, or cool surface temperatures in the eastern equatorial Pacific should continue through next hurricane season, though possibly in a diminished state from the very cold conditions presently observed ....... We predict yet another year of above average hurricane activity though less active than the recent very busy years of 1995, 1996, 1998 and 1999."

Our 7 April 2000 updated forecast

Same statements as our 8 December 1999

Our 7 June 2000 updated forecast

Same statements as the two earlier forecasts:

Our 4 August 2000 updated forecast:

``Information obtained through July 2000 indicates that the Atlantic hurricane season in 2000 is likely to be less active than the four recent very busy years of 1995, 1996, 1998 and 1999. However, total activity is expected to exceed the long term average and is anticipated to be considerably more active than the mean for the recent period of 1970 through 1994 ......"

``The observation of no named storms through the 4th of August 2000 is judged to have little or no bearing on whether we will have an overall active or inactive hurricane season."

``We expect August to have hurricane activity above that (a net NTC of 33/26.1 or 126 percent) of the long period August mean. In round numbers the August forecast is for three named storms, two hurricanes, and one intense or major hurricane."

[Our August only verification values were 4 NS, 2 H, 1 IH - very close to what we observed.]

We consider our Atlantic basin seasonal forecast for 2000 to have been a success. Our probability of 88 percent of a U.S. landfalling hurricane and 72 percent chance of a landfalling major hurricane did (fortunately) not verify. A continuation of the fortuitous trend (or luck) where the U.S. has been spared an increase in major landfall activity despite the strong up turn of major hurricane activity since 1995 should not be expected. We have experienced only three U.S. landfalling major hurricanes the last six years, when by the standards of last century ratio of U.S. landfall to Atlantic basin major hurricanes we should have experienced seven or eight events between 1995-2000.

8  Initial Prediction and Verification of Individual Monthly Hurricane Activity

       Background. Periods within variously active or inactive Atlantic basin hurricane seasons do not conform to the overall trend of the season as a whole. For example, although 1961 was a very active hurricane season there was no tropical cyclone activity of any kind during the entire month of August. In 1995, 19 named storms formed in the Atlantic but only one new named storm developed during the 30-day period spanning the statistical peak of the hurricane season between 27 August and 26 September. Conversely, the inactive season of 1941 had only six named storms (average 9.3) but four of these storms developed during September (average September activity is 3.4). During the inactive hurricane season of 1968, three of the eight named storms that year formed during June (average is 0.5).

We are studying how well various sub-season and/or individual monthly trends can be forecast in efforts being spearheaded by Eric Blake of our project.

It is, in general, more difficult to predict shorter periods of hurricane activity than to predict the entire yearly activity. [signals vary more from climatology on a monthly than on a seasonal basis.] Despite these inherent difficulties, we have, nevertheless devised a quite skillful forecast scheme (as determined by 51 years of hindcast testing using a seasonal independent jackknife approach) for the prediction of August-only activity. This technique involves searching June and July global reanalysis data for potential predictors associated with active versus inactive August periods. We predict the same activity parameters (NS, NSD, H, HD, etc.) as in our seasonal scheme. This monthly forecast methodology will be fully documented in a forthcoming paper.

Verification. Table 8 summarizes our forecast of TC activity for August 2000, along with a jackknife estimate of hindcast skill for the 51-year period of 1949-1999, long period August mean values, and final adjusted August 2000 forecast numbers (Table 8).

Table 9 shows the best July analog years for 2000 which were used as an aid in fine tuning the final August 2000 forecast values. These six analog years averaged slightly less activity than the forecast. Table 9 also shows the August verification. Note that the above average 4 August forecast was made despite a total absence of tropical cyclone activity prior to issuing the forecast.

August was characterized by Caribbean basin ZWA and SLPA values which were slightly unfavorable. But other global June-July circulation features were favorable for August activity. Four named tropical storms formed with two reaching hurricane strength and one reaching major hurricane status. It is notable that three of the named storms were very short-lived, largely due to the higher than average local vertical shear. In fact, two of the storms dissipated completely over the open ocean due to strong shear in the deep tropics, an unusual occurrence for the deep tropics in an otherwise active hurricane season.

Alberto was the one storm that thrived during August. It persisted as a hurricane for a long period. However, Alberto did not intensify greatly until it reached higher latitudes, where the ZWA was actually negative due to the presence of a strong TUTT. Alberto became an intense hurricane north of 35N, a rare occurrence. As noted earlier, it was the longest-lasting tropical cyclone on record in August, which caused the extremely large number of named storm and hurricane days.

This was our first monthly forecast. We will be working to develop monthly forecasts for other individual months and for the crucial 30-day period of mid-August to mid-September.

9  Seasonal Forecast Methodology

       Table 10 provides a comparison of the statistical forecast values versus our final adjusted forecasts at different lead times. Note that all of the statistical forecasts consistently underestimated 2000 year cyclone activity and that we made an upward adjustment of our actual forecast. Our statistical regression based predictions have been consistently less skillful than our actual forecasts since 1995, when we entered what we believe will be a new multi-decadal era of altered global atmosphere and ocean circulation features which enhance Atlantic basin hurricane activity. Our previous statistical forecast methodology was based on training data from 1950 through to 1990 and, as such, has proven less robust for predictions in the period since 1995. We do not know the specific reason(s) for the sudden fall-off of skill. However, the unexpected break down of the association between major hurricane activity and West African rainfall may be implicated. The ENSO-hurricane relationship appears to have been stronger the last six years.

In consideration of this fall off in statistical regression reliability we chose to base our 2000 seasonal forecast on analog prediction methodology which involves identifying past years with atmosphere and ocean conditions similar to those of the year being forecast.

Since 1949, there were five years with 3 to 12 month precursor ocean and atmosphere conditions fairly similar to 2000: 1949, 1956, 1981, 1989 and 1996. Table 11 lists these precursor properties for these analog years. Properties considered in selecting these years include the following:

1. The QBO relative wind conditions at 30 mb are from the east.
2. An active multi-decadal period for hurricanes which is characterized by a comparatively strong Atlantic thermohaline circulation as gauged by measurements of SSTA in the North Atlantic (50-60^°N, 10-50^°W) during the prior 11 to 19-month period. Warm SSTA's in this location indicate a stronger than average conveyor; cold anomalies a below average Atlantic Ocean conveyor. A century of evidence indicates that hurricane activity (particulaly intense or major hurricane activity) is enhanced when a strong Atlantic conveyor circulation is present and suppressed when a weaker circulation is present. All 1999 and 2000 early SSTA information indicated a strong conveyor for 2000.
3. Projected below average SSTA for August-October in the Pacific Nino 3.4 area which occurred.
4. Projected negative Caribbean basin 200 mb ZWA conditions for August through October. This occurred during the average of the two critical months of August and September but not for October.
5. Projected negative Caribbean basin SLPA conditions for August through October. These values were neutral for August and September. October had relatively high pressure however.
6. Projected above average Western Sahel for June through September 2000 did not occur wherein actual rainfall amounts were -0.75 SD below average.

Table 12 compares the seasonal hurricane activity during the five analog years versus observed 2000 conditions. Note that the observed seasonal tropical cyclone parameters for 2000 were generally similar to the average of the five analog years excepting western African rainfall. Clearly, the methodology of consulting prior years wherein similar global atmosphere and ocean conditions occur (ie - the analog approach) appears quite promising. We hope to spend the next few years in further development of this forecast methodology.

Note in Table 12 that our forecast was very similar to the average of the five analog years. Given the similarity of 2000 precursor conditions to the average of the precursor conditions of the analog years we were wise to have issued a seasonal forecast similar to the average of the hurricane activity that occurred in these years.

This analog methodology does not require detailed understanding of all the complex processes behind such associations and is a great advantage over initial value numerical modeling approaches which depends on accurately simulating most of the complex physical processes of the atmosphere-ocean system.

10  Comparison of Statistical and Analog Forecasts

       We are presently trying to provide an explanation for the recent (since 1995) decrease in the skill of our seasonal statistical regression forecasts. The best measure of forecast skill is for NTC activity which is a percentage composite measure of the six seasonal activity parameters (NS, NSD, H, HD, IH, IHD). Comparisons of our various lead time forecasts of NTC for 2000 are given in Table 13. Note that our analog forecast technique predicted NTC one-and-a-half to two times greater than that of our statistical scheme (column a). Knowing that our recent year statistical forecasts were underestimating hurricane activity we opted to use our analog techniques. This proved to be a wise decision. Column (b) of Table 13 shows that the ratio of the 2000 observed value of NTC (134) to our analog forecast of NTC was quite close. Observed values of NTC were a very good match for our analog forecasts. Table 14 shows the ratio of each of our four lead time adjusted forecasts to that of our statistical forecasts (column c). Note that our four lead time forecasts were about 1.8 times higher than our statistical regression forecasts. By contrast, the observed NTC to our forecast NTC (column d) was very close.

This same relationship of analog to statistical regression forecasts occurred for our forecast of the active Atlantic hurricane seasons during 1998 and 1999. Table 15 is similar to Tables 13 and 14. They show the average of the four lead time forecasts for each of the last three seasons. All three years of analog forecasts were superior to our statistical forecasts - compare columns (1) and (2). The average three-year verification errors were only seven percent by the analog forecast technique whereas there was a 80 percent underforecast by our statistical regression method (column 2). The average of our actual final forecasts during these three years were somewhat less skillful than our three-year average analog forecasts - underforecast of 28 percent (column 4).

Our actual 1998 NTC forecast was a significant underestimate of that year's actual NTC activity (compare columns 1 versus 4). We would have been closer if we had relied only on our analog information. Our forecasting for both the 1999 and 2000 seasons used the analog approach and were quite close to the actual NTC values which occurred.

10.1  Conjecture on Possible Causes for the Recent Failure of Statistical Regression Scheme

       Since 1995 we have had a major change in the basic global atmospheric and oceanic circulation features including (among others):

1. Atlantic Sea Surface Temperature (SST) and surface pressure patterns
2. Trend in the North Atlantic Oscillation
3. Alteration in Pacific SST and Circulation Features

The global atmosphere-ocean system has been functioning differently since mid-1995 and, it appears, that our statistical schemes have missed the influence of these altered physical factors which affect Atlantic tropical cyclone activity. As our statistical schemes were developed on training data from the 1950-1990 period, since 1995 the atmosphere-ocean is functioning differently than it had in the earlier period, it is possible that cause of the break down of our statistical regression scheme may have the same origin as the recent changes in the western Sahel and major hurricane relationship and the appearing stronger relationship between ENSO and Atlantic hurricane activity.

10.2  Current Superiority of Analog Forecast Methods

      At present the analog forecast technique, when property applied, appears to have distinct advantages over the typical statistical regression schemes that we have used for many years.

Typically, there are at least a few prior years wherein a number of similar global precursor relationships such as the El Niño Southern Oscillation (ENSO), the stratospheric QBO, the global arrangement of surface pressure anomaly (SLPA) and sea surface temperature anomaly (SSTA), etc. are generally similar to the current year's pre-season conditions. Selecting analogs allow us to immediately focus on the few prior years likely to be more representative of current conditions.

The atmosphere and ocean, in combination, have a strong and long multi-month to multi-season memory. It is reasonable to infer that global precursor signals of past years that are associated with specific levels of hurricane activity would be similarly indicative of the same level of hurricane activity for a coming season. As no prior year can ever be expected to be a perfect analog to current year conditions, it is likely that an ensemble average of the four or five closely similar multi-parameter precursor analog years may be the best estimate of the forthcoming season. This is an area which we plan to conduct further research.

11  Increased Level of Atlantic Basin Hurricane Activity (Including U.S. Landfall) During the Last Six Years

       A major rearrangement of Atlantic Ocean SST features began in mid-1995 and has continued through October 2000 (Fig. 6). This change is well associated with increased Atlantic basin intense or major (cat. 3-4-5) hurricane activity during the last six years. We hypothesize that these strong, broadscale SST changes are due to basic changes in the strength of various components of the Atlantic Ocean thermohaline (``conveyor belt") circulation. This interpretation is consistent with changes in a long list of global atmospheric circulation features during the last six years which conform to a prominent shift towards a stronger Atlantic Ocean thermohaline circulation and Atlantic SSTAs (see Fig. 6). This change occurred between 1994 and 1995. Such changes in Atlantic multi-decadal thermohaline circulation shifts appear to occur on periods of 25-50 years. If this interpretation is correct, then increased Atlantic basin intense (category 3-4-5) hurricane activity may be expected to persist through the early decades of the 21st century, which will be in contrast with the greatly diminished activity during the later decades of the 20th century.

Despite El Niño-linked reduced hurricane activity of 1997, the last six years (1995-2000) are together the most active six consecutive hurricane years on record. Table 16 lists the total number of named storms (79), hurricanes (49), major hurricanes (category 3-4-5) (23), major hurricane days (56.25) and Net Tropical Cyclone (976) which occurred during 1995-2000. Note that despite the inactive 1997 season, the annual average NS, H, IH, IHD and NTC during these six years was 146, 163, 239, 329, 331 and 214 percent of the NS, H, HD, IH, IHD, and NTC hurricane activity of the prior (1989-94) six-year period. Note also that NS, H, HD, IH, IHD and NTC during the last six years are 153, 165, 247, 250, 373 and 217 percent of the average for the prior 25-year (1970-1994) period; the greatest increase having occurred for IH and IHD activity. Figure 7 portrays differences in H and IH tracks during these periods. These trends to increased hurricane activity give strong support to the suggestion that we have indeed entered a new era of greatly increased major hurricane activity. Despite the large El Niño-linked reductions in NTC during 1997 (55), NTC activity of the six-year period of 1995-2000 has averaged 165 or 165 percent of the 1950-1999 average. Excluding 1997 average NTC for the five years of 1995, 1996, 1998, 1999, and 2000 have had a mean NTC of 187.

For some years we have suggested that the era of greatly reduced intense Atlantic category 3-4-5 hurricane activity between the late 1960s to early 1990s would end and that the U.S. and Caribbean coastal regions should be expected to see an increase in landfalling major hurricanes (Gray 1990). This outlook is ominous because of the increases in U.S. ssoutheastern population and the new realization that when hurricane destruction is normalized for coastal population, inflation, and wealth per capita [see Pielke and Landsea (1998)] that major hurricanes (on a statistical basis) cause about 85 percent of all U.S. tropical cyclone linked destruction.

Recent Upswing of U.S. Landfalling Tropical Cyclones but Decrease in Major Hurricane
Landfall. In comparison to Atlantic basin major hurricane activity Table 17 lists the number of U.S. landfall named storms during the two most recent six-year periods, 1989-1994 versus 1995-2000. Note the increase from 8 to 13 in U.S. hurricane landfall during the last six years from the previous six-year period. This increase is consistent with the overall increase in net Atlantic basin hurricane activity as shown in Table 16. Table 18 gives ratios of 1995-00 versus 1989-94 incidence of landfalling tropical cyclones and two categories of hurricanes. Overall, hurricane landfall activity increased 163 percent between the earlier and later period.

However, the foregoing is deceiving, particularly in comparison to the differences in major landfall events between these two periods. Official records indicate that over the last century (1900-1999) there have been 218 major hurricanes in the Atlantic basin and of these category 3-4-5 storms, about one-third (73) have come ashore along the U.S. coastline. In the last six years (1995-2000) there have been 23 major hurricanes within the Atlantic basin but only three (Opal, 1995; Fran, 1996; and Bret, 1999) have come ashore. If the typical one out of three ratio of major hurricane landfall events of the last six years had taken place, then we should have experienced 7-8 major hurricane landfall events, not just three that did come ashore.

We have been fortunate that an upper-air trough has been located along the U.S. East Coast during a high percentage of time during the last six hurricane seasons. The fortuitous frequent location of this upper-level East Coast trough has caused a large portion of otherwise northwest moving major hurricanes to be recurved to the north before they reach the U.S. coastline. Thus, we have been lucky. But this luck can not be expected to continue. Very few residents of the southeastern U.S. coastline are likely aware of how fortunate they have been over the last 3-4 decades.

Given the U.S. major hurricane landfall numbers of the last century, our luck at beating climatology has now extended about four decades. For example, in the 30-year period of 1971-2000, the U.S. experienced 15 major landfall events, or 0.50 per year. This is only 62 percent the annual incidence of major hurricane landfall events which occurred in the previous 72 years of 1900-1971.

With regard to the Florida Peninsula and the U.S. East Coast, the situation is even more skewed. In the last 40 years (1961-2000), there have been only six landfalling major hurricanes (average 0.15 per year) along the Florida Peninsula and U.S. East Coast. Between 1900-1960 there were 31 major landfall events along this same coastline (or 0.51 per year). The first six decades of the 20th century had 3.4 times more landfall major hurricanes along the Florida Peninsula and East Coast than occurred in the last four decades. This long downturn in U.S. major hurricane landfall events along the Florida Peninsula and East Coast is unlikely to continue. Climatology will eventually right itself and we must expect a great increase in landfalling major hurricanes in the coming few decades.

12  Landfall Probabilities for 2000

       A new aspect of our research involves efforts to develop forecasts of the probability of hurricane landfall along the U.S. coastline. Whereas individual hurricane landfall events can not be accurately forecast for an individual year, the net yearly probability of landfall can be forecast with statistical skill. With the premise that landfall is a function of varying climate signals, a probability specification has been accomplished through a statistical analysis of all U.S. hurricane landfalls of named storms during the last 100 years (1900-1999). Specific landfall probabilities can be given for all 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 Activity (NTC) and to climate trends linked to multi-decadal variations of the Atlantic Ocean thermohaline circulation (as measured by recent past years of North Atlantic SSTA*). Table 19 gives verification of our landfall predictions for 2000. These landfall probabilities did not materialize.

Active research is in progress on this technique. Full documentation of the methodology for estimating hurricane landfall probability study is being prepared and will, hopefully, be available in the next few months. Landfall probabilities include specific forecast of the probability for landfalling tropical storms (TS) and hurricanes of category 1, 2, 3, and 4-5 is being developed for each of 11 units of the U.S. coastline (Fig. 8). These 11 units are further being subdivided by coastal population into 96 regions based on coastal population. Statistics are being developed for each 100 km (65 mile) segment of the entire U.S. coastline.

Figure 9 gives a general outline of this methodology. These forecast probabilities will be supplemented with probability values for each 100 km coastal segment receiving gale force winds ( ³ 40 mph), sustained hurricane force winds ( ³ 75 mph), and major hurricane (category 3-4-5) winds ( ³ 115 mph). There will also be a discussion of potential tropical cyclone spawned hurricane destruction within each of the 96 different U.S. coastal locations.

13  The 1995-2000 Upswing in Atlantic Hurricanes and Global Warming

       Some may interpret the recent large upswing in Atlantic hurricane activity (since 1995) as being in some way related to increased man-made greenhouse gases such as carbon dioxide (CO₂). There is no scientifically reasonable way that such an interpretation of this recent upward shift can be made. Anthropogenic greenhouse gas warming, even if a physically valid hypothesis, is a very slow and gradual process that, at best, might only be expected to bring about small changes in global circulation over periods of 50 to 100 years and could not cause the abrupt and dramatic upturn in hurricane activity as occurred between 1994 and 1995. Also, the large downturn in Atlantic basin major hurricane activity between 1970-1994 would need to be reconciled with proposed global warming scenarios during this period. Atlantic intense (or category 3-4-5) hurricane activity showed a substantial decrease during 1970-1994 to levels about 40 percent of the amount which occurred during the 1950-1969 or the 1995-2000 periods. There were 78 Atlantic basin major hurricanes in the 26 years of 1950-1969, 1995-2000 versus 38 in the 25 years of 1970-1994. This is an annual ratio differences of two to one. And, even if man induced greenhouse increases were shown to be causing global temperature increases over the last 25 years, there is no way to relate such a small global temperature increase to more hurricane activity.

In contrast with the large increase in Atlantic basin major hurricane activity during the last five years, total hurricane and typhoon activity in the (East and West) North Pacific region during the period 1995-2000 has decreased. When we combine Atlantic and North Pacific tropical cyclone activity, we see a net downward trend for the recent 1995-2000 period (Table 20). Hence, we should not interpret the recent enhancement of major hurricanes in the Atlantic as indicative of the changes of hurricane activity around the globe. It is only in the Atlantic where hurricane activity has shown a sharp rise and this rise is in conformity with the changes in Atlantic sea surface temperature patterns and the diagnosed increase in the thermohaline circulation. Such up and down multi-decadal changes in Atlantic intense sea surface temperature and tropical cyclone activity have been observed to take place many times in the past and are considered to be naturally occurring modes of multi-decadal variability.

14  Forthcoming Early December Forecasts for 2001 Hurricane Activity

      We will be issuing a seasonal forecast for 2001 Atlantic basin hurricane activity on
8 December 2000. A separate forecast will also be attempted for the month of August 2001. These forecasts will be based on data available to us through November 2000. These forecasts will be disseminated on the World Wide Web. Updates to the 2001 seasonal forecast will be issued in early April, early June, and early August 2001 as will a separate forecast for August 2001 with our late update.

15  Acknowledgments

       John Sheaffer, John Knaff, Todd Kimberlain, Eric Blake, and William Thorson have made many important contributions to the conceptual and scientific background for these forecasts. The authors are indebted to a number of meteorological experts who have furnished us with the data necessary to make this forecast or who have given us valuable assessments of the current state of global atmospheric and oceanic conditions. We are particularly grateful to Arthur Douglas, Richard Larsen, David Masonis, Vern Kousky, Ray Zehr and Mark DeMaria for very valuable climate discussions and input data. We thank Colin McAdie and Jiann-Gwo Jiing who have furnished data necessary to make this forecast and to Gerry Bell, James Angell, and Stan Goldenberg for input data and helpful discussions. Richard Taft has provided valuable data development and computer assistance. We wish to thank Tom Ross of NCDC and Wassila Thiao of the African Desk of CPC who provided us with West African and other meteorological information. In addition, Barbara Brumit and Amie Hedstrom have provided excellent manuscript and data analysis assistance. We have profited over the years from many indepth discussions with most of the current NHC hurricane forecasters. These include Lixion Avila, Miles Lawrence, Richard Pasch, Jack Beven and James Franklin. The first author would further like to acknowledge the encouragement he has received for this type of forecasting research applications from Neil Frank, Robert Sheets, Robert Burpee, Jerry Jarrell, former directors of the National Hurricane Center (NHC), and from the current director, Max Mayfield.

       The financial backing for the issuing and verification of these forecasts has, in part, been supported by the National Science Foundation. But this NSF support is insufficient. Recently, the Research Foundation of the United Services Automobile Association (USAA) and State Farm insurance companies have made contributions to the first author's project. It is this support which is allowing our seasonal predictions to continue.

16 Additional Reading

APPENDIX A: Verification of Past Seasonal Forecasts