WMO/CAS/WWW

FIFTH INTERNATIONAL WORKSHOP ON TROPICAL CYCLONES

Topic 4.5 Seasonal prediction of tropical cyclones


Rapporteur: B. Wang
Meteorology, HIG 367
University of Hawaii
2525 Correa Rd.
Honolulu, HI 96822

E-mail bwang@soest.hawaii.edu
FAX: 808.296.2877

Working Group: W. Gray, C. Landsea, R. Perez-Suarez, K.-T. Sohn


Abstract:

The activity of tropical cyclone (TC) varies interannualy worldwide, causing the deadliest and costliest disasters. This fact has led to an increasing interest by the research community into possible methods of seasonal prediction, especially in the Atlantic. Many authors have found relationships between the slowly varying background tropical circulation and interannual TC activity. Therefore the fundamental issue is the accurate prediction of such slowly varying circulation features and further to understand how they affect TC activity. One of the main difficulties of seasonal prediction lies in the accurate understanding of the interaction of several genesis-affecting factors. For example, in the Atlantic the El Nino-Southern Oscillation (ENSO), Sahelian rainfall, African easterly wave, decadal variations in sea surface temperature (SST) and the stratosphere quasi-biennial oscillation (QBO) all play their parts in the TC genesis potential. How these influences are teleconnected is largely unstudied and thus attempts at seasonal prediction show some skill but are not satisfactory. The current understanding of the physical processes governing variability of TC activity makes it possible to predict the numbers and to a lesser extent the location and maximum intensity of TCs one or two seasons in advance. Tropical cyclone activity in the Pacific and Indian Ocean hinges on accurate ENSO evolution prediction. Skill for TC seasonal prediction is largely determined by the persistence of global SST patterns, which to a large extent determines the large-scale atmospheric conditions.  The prediction of the global SST field has remained elusive. Future directions for seasonal prediction of TCs are suggested.


4.5.1 Introduction.

In the past two decades, the seasonal prediction of tropical cyclone(TC) activity has received increasing attention. Accurate prediction of TC activity weeks or months in advance would certainly be a powerful tool in disaster preparedness and prevention.

Seasonal prediction of TC activity requires accurate seasonal prediction of the environmental factors that either inhibit or enhance genesis potential. The genesis of TC’s is inherently a multistage process that begins with the existence of a suitable large scale environment that is capable of sustaining TC precursors such as the African easterly waves or mesoscale convective systems. Therefore the main focus of the seasonal prediction has been to relate large-scale atmospheric and oceanic circulation changes to changes in observed TC activity.

The effects of factors such as ENSO on TC activity vary from basin to basin and also exist with other influencing factors such as the stratospheric QBO, Madden-Julian Oscillation (MJO) and many others. A fundamental weakness in the literature is the lack of knowledge of how such factors are teleconnected. Until an accurate knowledge of how these factors inter-relate and combine is obtained, seasonal prediction will remain an art rather than a science.

This review contains a basin-to-basin discussion of the current knowledge of the factors affecting seasonal TC activity and progresses on seasonal prediction of TCs (section 2) and a summary on the major issues on TC predictability and grounds for further research (section 4.5.3).


4.5.2 Global seasonal TC prediction.

a) The Atlantic Basin

The annual number of TC in the Atlantic basin is highly variable ranging from 1 to 21 storms in the last century (Neumann et al. 1993). Researchers have identified a number of large-scale environmental influences on such variability. ENSO and the stratospheric QBO (Gray, 1984a), Atlantic sea level pressure (SLP) (Shapiro 1982,a,b, 1987; Gray et al 1993, 1994; Knaff 1997) and western Sahelian monsoon rainfall (Landsea and Gray 1992, Goldenberg et al. 1996) have all been shown to affect TC activity on seasonal time scales.

The effects these phenomena have on TC activity are all based on the change of the vertical shear induced in the so-called main development region (MDR)- 10-20N 20-80W. Landsea and Gray (1992) have established a very significant relationship between above (below) normal rainfall in western Africa and above (below) normal Atlantic TC activity. In dry years an observed anomalous zonal circulation with upper level westerlies and lower level easterlies acts to increase climatological vertical wind shear (Goldenberg and Shapiro 1996). Upper level anomalous westerlies over the south and southwest part of the MDR associated with the warm phase of ENSO have been shown to reduce TC numbers here (Goldenberg and Shapiro 1996). Figure 1 illustrates that much of the influence on major hurricane activity is linked to ENSO and Sahel rainfall variations although some other factors still contribute. The interannual variation of major hurricane activity in the MDR, explained by Sahel rainfall is 3 times that by eastern Pacific SST anomalies. It is thought that the reason for this is that the vertical shear variability induced by Sahel rainfall is less equatorially confined than ENSO variations.

Besides the shear anomalies induced by rainfall anomalies, the strength and number of African easterly waves (AEWs) may contribute to genesis potential. However, very little is known about the character of AEW’s leaving the African coast. Recent work by Thorncroft and Hodges (2001) suggested that the strength of the low-level vorticity (850mb) of AEW’s more effectively influence the number of TC’s in a given season than do total numbers. They also suggest that greater numbers of these well-organized waves may occur in wet Sahel years. This is an indication that as well as a vertical shear influence, a mechanism to produce more strong seeds as precursors for genesis may be at work. For accurate seasonal predictions more detailed studies of the character of AEW’s are needed as well as their linkage to the West African monsoon.

It is well established that SLP anomalies in the tropical Atlantic are related to seasonal TC activity. Shapiro (1982a,b) suggests that one sixth of TC activity can be explained by such variations in SLP. Knaff (1996) has suggested that an anomalously strong TUTT cell is associated with increased low level SLP. He hypothesizes that when SLP is high, subsidence is stronger and middle levels are drier resulting in stronger atmospheric cooling and increased baroclinicity and a stronger TUTT. The increased TUTT circulation leads to enhanced vertical shear over the MDR and therefore is inhibiting to TC genesis. Very recently Landsea et al. (2001) talked about a multidecadal oscillation in the tropical Atlantic SST and SLP related to a slowly varying thermohaline circulation and its relationship to decadal oscillations in major hurricane activity.

Gray et al (1992,93) have used many of these physical mechanisms mentioned along with the phase of the QBO in an attempt to provide forecasts several months before the onset of the Atlantic TC season. Gray et al. (1992) hypothesized that during the westerly phase of the QBO, while the stratospheric winds near the equator are westerly, those in the tropics are easterly. As a result a region of small vertical wind shear develops away from the equator over the tropics. Approximately 2 times as many major hurricanes are observed during the westerly phase of the QBO. Details of the model are discussed in Gray et al. (1992,1993) where forecasts are made in early December prior to the TC season and in early June and August. Table 1 shows the performance off the Gray et al. model by comparing forecast TC activity and the observed numbers. The model has some skill but evidently is no-where near satisfactory, probably because of a poor knowledge to what extent each factor contributes to TC variability and how the factors are teleconnected. Landsea et al. (1995) commented on the models under prediction of the extremely active season that year.

b) The western North Pacific (WNP)

In the WNP the interannual variation of TC numbers has been attributed to ENSO (Chan 1985,2000;Dong 1988;Lander 1993 1994;Wang and Chan 2002) and to the QBO (Chan 1995). Many authors noted an eastward shift of TC genesis during El Nino and some concluded that there was a reduced TC count in warm years. Lander (1993) however refuted this latter conclusion and suggested that although there is an eastward shift of TC’s in the warm phase of ENSO total TC counts were not significantly different.

Wang and Chan (2002) reveal the impact of strong warm and cold events on TC activity using a 35yr dataset (1965-99). During the warm events the TS activity increases markedly in the southeast quadrant (0-17N, 140-180E) and decreases in the northwest quadrant (17-30N,120-140E) (Fig. 2). During the fall of the strong warm events the number of TC’s that recurve is 2.5 times higher than the strong cold events and the lifetime of the systems is substantially enhanced (Fig. 2). After El Nino the early season (March-July) TC activity is suppressed. The enhanced TC formation in the SE quadrant is attributed to an increase in low-level shear vorticity due to enhanced westerlies there. The changes of the TC’s track were due to changes in the large-scale steering flows. Some predictive capability (e.g., the number of TC formation in the southeast and northwest quadrant, the life spans, the mean location of the TC formation) is noted based on the strength of El Nino. The authors also note a high level of predictability for TC activity in July-December using the preceding winter-spring Nino 3.4 SST anomaly and for March-July using the October-December Nino 3.4 anomaly in the previous year. The first predictive skill noted is due to the phase lock of ENSO to the annual cycle whereas the second is due to the persistence of Philippine sea wind anomalies induced by ENSO but maintained by local air-sea interaction (Wang et al. 2000).

The main limitation of these results for WNP prediction is that for neutral years or weak warm/cold events there appears to be less predictable of seasonal TC activity on the basis of ENSO. Chan (1995) found using spectral analysis techniques that the westerly phase of the QBO is more favorable for TC’s in the WNP due to similar mechanisms in the Atlantic. However Chan (1995) notes that the relationship is not as clear when ENSO events occur concluding that ENSO has the more powerful effect.

c) The eastern North Pacific (ENP)

The ENP has an average of 17 TC’s per year. ENSO warm events appear to slightly increase the number of TC’s as a whole but a more significant factor is more intense hurricanes occur. This appears logical due to warmer SST’s in the region during strong warm events. While the seasonal prediction of TC has not been well studied, there are many studies that attempt to elucidate genesis mechanisms based on which their potential for seasonal prediction can be discussed.

Molinari and Vollaro (2000) and Maloney and Maloney and Hartmann (2000) attempt to provide evidence of the interaction of the MJO and AEW’s in ENP tropical cyclone genesis (1991 hurricane season). The hypothesis is that the MJO provides a suitable environment in itself to enhance wave growth in the ENP and that such waves usually originate over Africa. Molinari and Vollaro (2000) stated, “Synoptic-scale waves reached the western Caribbean and eastern Pacific regularly from upstream, usually from Africa”. If this is so then the interannual variation of AEW’s and how many of them reach the ENP should be a critical factor when considering seasonal prediction of ENP TC’s.

A considerable basis for this study was gained from the work of Maloney and Hartmann (2000) who defined the phase of the MJO using a principal component analysis of the dominant EOF’s of the near equatorial 20-80 days filtered, 850mb zonal wind anomaly. Figure 3 shows the incidence of genesis over 9 defined phases of the MJO over the 16-year period of study. Phase 2 corresponds to the maximum westerly anomaly phase whereas phase 6 corresponds to the maximum easterly anomaly. Figure 3 illustrates that during the westerly phase of the MJO and its associated convective maximum in the ENP, tropical cyclogenesis is twice as likely as in the easterly phase. In addition the systems that formed were eventually stronger in the westerly phase. The enhanced background vorticity associated with the westerly phase of the MJO was argued to be responsible for the increased number of genesis events.

Molinari and Vollaro (2000) conclude that the most prominent mode of tropical cyclogenesis in the eastern north Pacific goes as follows: (i) convectively active phase of MJO moves in from west. (ii) meridional PV gradient becomes strong in the western Caribbean. (iii) upstream waves (most likely from Africa) grow in intensity. (iv) tropical depressions form in association with these growing waves. In identifying the meridional wind anomalies and tracking them across the Atlantic to the ENP, the effects of the orography cannot be neglected; the orography may play a vital role in inducing enhanced cyclonic vorticity to the lee (Zehnder et al. 1999).

In the eastern north Pacific (ENP) it is believed by many researchers that easterly waves that can be tracked back to Africa are responsible for most genesis events. So it could be presumed that the number of easterly waves traversing the Atlantic is related in some way to subsequent genesis in the ENP. However such suppositions are uncertain and the evidence for easterly wave’s setting of genesis in the ENP is poor--seasonal prediction here is hence in its infancy.

d) The Australian region

The pioneering work in the seasonal prediction field was by Nicholls (1979,1984,1985,1992) for TC’s in the 2 Australian basins. He demonstrated an association between the SOI during the Southern Hemisphere winter and the subsequent numbers of TC’s close to Australia (105-165E) during the subsequent cyclone season (October to April). The linear correlation coefficient for the 25yr year series is –0.68. He further demonstrated that TC activity around the NE Australian coast weakens while activity east of 107E increases during a warm event. The reverse was observed during La Nina years. The total number of cyclones is significantly larger during La Nina events.

Such an eastward shift of genesis appears to have similar physical mechanisms as that observed in the western north Pacific whereby the anomalous warming in the eastern Pacific results in an eastward extension of the monsoon trough. The monsoon trough being a favored genesis area results in a concurrent shift in genesis locations.

A problem arose with Nicholls work however. Nicholls (1992) detected a sudden decrease in cyclone numbers following the end of the 1985/86 season that has not been associated with a corresponding decrease in the SOI. He suggested that the apparent weakening of the relationship was due to an artificial long-term downward trend in TC numbers caused by differing techniques in the interpretation of satellite imagery.

e) South west Indian Ocean (SWIO)

The major part of the SWIO TC season extends from November through March. The most recent work in the seasonal prediction of TC’s in the southwest Indian ocean (SWIO) has been by Jury et al. (1991,1993,1998). In the SWIO El Nino years appear to reduce the number of TC’s in the December-March season. Jury at al. (1998) state that this is explained by strengthened subtropical upper-level westerly winds. Jury (1993) analyzed synoptic-scale influences on SWIO TC frequency. Increased TC are found to be associated with the east phase of the QBO. The SWIO has upper level westerlies as a climatological mean which are weakened by the east phase of the QBO. Jury does not make clear the statistical significance of this result. Jury el al. (1998) produce a multivariate model based on the discussed mechanisms to estimate SWIO TC days one season in advance. It is claimed that the model accounts for more than half of the interannual variance in the 1971-92 period.

An ocean upwelling region exists in the SWIO as a result of a cyclonic wind curl between equatorial westerlies and southeasterly trades which acts to raise the thermocline in the west. Xie and Annamalai (2002) offer an insight into the possible increase of TC activity in the SWIO during El Nino or cool Sumatra events. During either event anomalous easterlies appear in the equatorial Indian Ocean. This forces a downwelling westward propagating Rossby wave. In the SWIO the shallow thermocline allows the Rossby wave and its associated warm SST anomaly to interact with the atmosphere producing a positive precipitation anomaly and possibly an increase in TC numbers. In fact the difference achieves a maximum in the region near 15S, 60E where a comparison of TC days during deep and shallow thermocline years, reveals an average of 4 and 1 days respectively. This is a more dynamic view than has been generally accepted previously of the SWIO climate variability but potentially provides an improvement in SWIO climate variability and its TC’s seasonal predictability.

f) The North Indian Ocean

The north Indian Ocean typically has 5-6 TCs per year. Over the past 100 years the number has varied from 1 to 10. The seasonal distribution is bimodal with a primary maximum in November and a secondary maximum in May. It is typically thought that this is due to the intervening periods between the summer and winter monsoons when the vertical shear is smaller and the monsoon trough is located sufficiently far from land that TC’s have time to form.

Simple correlations of total Indian Ocean TC numbers and the SOI reveal essentially no correlation for the basin as a whole. However, Singh and Khan (2001) find that during the negative phase of the SOI the numbers of TC and depression in the Bay of Bengal and the Arabian sea diminish in May. The correlation coefficient they obtain between the SOI and the frequency of TC’s and depressions is +0.3 (significant at the 99% level). A similar picture is obtained for the Bay of Bengal for November. This study gives a first glimpse that ENSO may indeed influence TC numbers on a regional basis. Further studies to search for physical mechanisms are needed as well as how the intraseasonal oscillation affects TC numbers by its influence on the monsoon.

4.5.3 Roadblocks and future direction

a) Issues

The Atlantic Ocean is the most studied basin in the field of seasonal prediction but serious questions remain as to the nature of the teleconnection between various TC genesis/intensification factors and their combined effects on seasonal Atlantic TC activity. Our knowledge of the character of easterly waves is weak as is our knowledge of their connection with the West African monsoon. The supposition that AEW’s set of TC genesis in the ENP does not have much evidence to support it. If they do, then a better knowledge of the background state on which they propagate through the Atlantic is required before much seasonal predictive skill is achieved in this region. The WNP requires more detailed study of the interannual variation of TC’s in years that are not El Nino or La Nina years, as does the Australian region. Little solid information is known about the relationship of TC’s to ENSO over the north Indian Ocean. The interaction of the monsoon, intraseasonal oscillation and ENSO in relation to a bimodal TC distribution makes this area a complicated region for seasonal prediction. The SWIO shows a clearer ENSO signal. More concise studies of the evolution of ENSO in many years and subsequent TC activity are required.

The current understanding of the physical processes governing the variability of TC genesis and tracks makes it possible to predict the numbers and to a lesser extent the location and maximum intensity of TCs one or two seasons in advance.  Statistical methods have been developed to do such predictions for different basins. Attempts are underway to utilize dynamically determined wind field forecasts to aid in the seasonal prediction of TC formation (e.g. Thorncroft and Pytharoulis, 2001). Skill as well as the probability of success is largely determined by the persistence of global SST patterns, which to a large extent determine the large-scale wind fields.  The large-scale tropical SSTs (ENSO, Pacific-Indian Ocean inter-decadal, and Atlantic Decadal Oscillations (Enfield and Mesta-Nunez 1999), vary slowly enough that crude seasonal prediction is possible.  For instance the state of the ENSO phenomenon is critical to the forecasts in all of the tropical cyclone basins currently being forecasted on a seasonal basis (NW. Pacific, Australian Region, and the Atlantic), yet there is little skill associated with seasonal ENSO forecasts beyond one season (Landsea and Knaff 2000)

It is the prediction of the global SST field that has remained elusive. This is in part due to the complicated coupled nature of climate system and the limited oceanic datasets available for assimilation in state of the art ocean models. Uncertainties arise from both the models’ physical parameterization and coupling techniques. Statistical models and dynamic-statistical models remain necessary and valuable for making SST forecast.

Another issue is the quality of the best track datasets.  While efforts are under way to reanalyze the Atlantic best track, similar efforts are not underway in other basins.  The historical best track data is often the starting place for many seasonal formation studies, but the original intent of these datasets was for seasonal verification based upon the best knowledge at that time.  It is likely time that these datasets be systematically reanalyzed using our current knowledge on this subject.
 
b) Future Directions

It also seems that the thing we want to predict is the most difficult to estimate.  While information on seasonal activity pertaining to numbers of cyclones and mean intensity can be predicted crudely at one to two season lead time, the value of such information has limited utility.   The information content the world wants to know (e.g. landfalling intensity and location, timing and track, intensity and track) remain elusive.

1. Study of the physical basis for the climate prediction of the TC activity remains to be the fundamentals for further improvement and development of skillful forecast models (statistical, dynamical-statistical, and dynamical). Many of the issues identified in 3.1 provide impetus for further studies.

2. Development of skillful ocean forecasting models that incorporate the continual and systematic independent verification is an essential need.   

3. Develop dynamical-statistical models for TC formation prediction.   Dynamical model-based seasonal ensemble prediction of the circulation fields is becoming better with time and is nearing the point where they would be useful for this application.  Multi-dynamical model combined with statistical approaches, namely the super ensemble technique (Krishnamurti et al. 2000), has demonstrated its great potential as the state of art forecast tool.  Dynamic-statistical models such as the SHIPS intensity model that uses the fields from the model to infer intensity change would likely aid seasonal tropical cyclone forecasts.
 
4.  Emphasis should be placed on utility.  Landfall, track and timing information should be included and emphasized in future prediction methodologies.  (e.g. If 10 tropical cyclones form and 5 will be hurricanes, will these hurricanes affect land?  Where and how strong will these be at landfall?).

5.  Best track datasets should be reanalyzed using current understanding and techniques.  A consistent dataset is lacking even in our best-observed basins.

 

Bibliography

Chan,J.C.L.,1985:Tropical cyclone activity in the western north Pacific in relation to the El Nino/Southern oscillation phenomenen. Mon.Wea.Rev.,113,599-606.
Chan, J.C.L., 1995:Tropical cyclone activity in the western north Pacific in relation to the stratospheric QBO. Mon.Wea.Rev. 123,2567-2571.
Chan, J.C.L.,2000:Tropical cyclone activity over the western north Pacific associated with El Nino and La Nina events. J.Climate,13,997-1004.
Dong, K.,1988:El Nino and tropical cyclone frequency in the Australian region and northwest Pacific. Aust. Meteor. Mag., 36,219-225.
Goldenberg, S.B. and L.J. Shapiro,1996: Physical mechanisms for the association of El Nino and West African rainfall with Atlantic major hurricane activity, J. Climate. 9,1169-1187.
Goldenberg, S.B., C.W. Landsea., A.M. Mesta-Nunez and W.M.Gray. the recent increase in Atlantic hurricane activity: Causes and implications. Science. Vol 293.no5529 474-479.
Gray, W.M., 1984a: Atlantic seasonal hurricane variability. Part I: El Nino and 30mb QBO influences hurricanes. Mon. Wea. Rev., 112,1649 1668.
Gray, W.M., 1984b: Atlantic seasonal hurricane variability. Part II. Forecasting its variability.Mon.Wea.Rev.112,1669-1683.
Gray, W.M., 1990: Strong association between western Sahelian rainfall and U.S. landfall of intense hurricanes. Science, 279, 1251-1256.
Gray, W.M., C.W. Landsea, P. Mielke and K. Berry,1992: Predicting Atlantic seasonal hurricane activity 6-11 months in advance. Wea. and Forecasting, 7, 440-455.
Gray, W.M., C.W. Landsea, P. Mielke and K. Berry,1993: Predicting Atlantic basin seasonal tropical cyclone activity by August 1st. Wea. and Forecasting, 8, 73-86.
Jury,M.R.,B.Pathack,1991:A study of climate and weather variability over the tropical S.W Indian Ocean. Meteor. Atmos. Phys., 47, 37 48.
Jury,M.L.,1993:A preliminary study of climatological associations and characteristics of tropical cyclones in the S.W. Indian Ocean., Meteo. Atmos. Phys., 51, 101-115.
Jury, M. R., B. Pathack and B.Parker.,1998:Climatic Determinants and statistical prediction of tropical cyclone days in the southwest Indian Ocean,J.Climate,12,1738-1745.
Knaff, J.A.,1997:Implications of summertime sea level pressure anomalies in the tropical Atlantic region. J.Climate.,10,789-804
Lander,M.A.,1993:Comments on a ‘GCM simulation of the relationship between tropical storm formation and ENSO.’ Mon.Wea.Rev.,121,2137-2143.
Lander, M.A.,1994:An exploratory analysis of the relationship between tropical storm formation in the western north Pacific and ENSO. Mon.Wea.Rev.,122,636-651.
Landsea, C. W., G.D Bell, W.M. Gray and S.B Goldenberg .,1997: The
extremely active 1995 hurricane season: Environmental implications and verifications of seasonal forecasts. Mon. Wea. Rev., 126, 1174-1193.
Nicholls, N., 1979: A possible method for predicting seasonal tropical cyclone activity in the Australian region. Mon. Wea. Rev., 107, 1221-1224.
Nicholls, N., 1984: the Southern Oscillation, sea surface temperature, and interannual fluctuations in Australian tropical cyclone activity. J. Climatol., 4, 661-670.
Nicholls, N., 1985: Predictability of interannual variations of Australian seasonal tropical cyclone activity. Mon. Wea. Rev., 113, 1144-1149.
Nicholls, N., 1992: Recent performance of a method for seasonal forecasting of seasonal tropical cyclone activity. Aust. Meteor. Mag. 21, 105-110.
Neumann, C.J., B.R. Jarvinsen, C.J. McAdie and J.D. Elms, 1993: Tropical cyclones of the North Atlantic Ocean, 1871-1992. Historical Climatology Series 6-2, NOAA, Asheville, NC, 193pp.
Shapiro, L.J.,1982a:Hurricane climatic fluctuations: Part I: Patterns and cycles. Mon.Wea.Rev.,110,1007-1013.
Shapiro, L.J.,1982b:Hurricane climatic fluctuations: Part II: Relation to large scale circulation.Mon.Wea.Rev.,110,1014-1040.
Shapiro, L.J.,1987:Month to month variability in the Atlantic tropical circulation and its relationship to tropical cyclone formation. Mon.Wea.Rev.,115,2598-2614.
Singh, O.P.,T.M.A. Khan and M.S. Rahman,2001: Tropical cyclone frequency in the north Indian ocean in relation to the southern oscillation. Mausam., 52, 511-514.
Solow, A. and N. Nicholls, N.,1990:The relationship between the southern oscillation and tropical cyclone frequency in the Australian region. J. Climate. 3. 1097-1145.
Thorncroft, C.D. and K.I. Hodges,2001:African easterly wave variability and it relationship to Atlantic tropical cyclone activity, J. Climate, 14, 1166-1179
Thorncroft, C.D. and I. Pytharoulis,2001: A dynamical approach to seasonal prediction of Atlantic tropical cyclone activity. Weather and forecasting, December issue.
Wang, B., R. Wu and X. Fu, 2000: Pacific-East Asia teleconnection: How ENSO affects East Asian climate? J. Climate,
Wang, B. and J.C.L. Chan 2002: How strong ENSO events affect tropical storm activity over the western north Pacific, J. Climate., 1643 1657.
Xie, S. P., and H. Annamalai.,2002:Structure and mechanisms of South Indian Ocean Climate variability., J. Climate, 15,864-878.
Zehnder, J.A, D.M Powell, and D.L Ropp, 1999: The interaction of easterly waves, orography, and the ITCZ in the genesis of eastern Pacific tropical cyclones. Mon.Wea.Rev,127, 1566-1585.




























Table 1
: Real-time Gray et al. forecasts and results for the Atlantic tropical cyclone season during the years 1984-1996.



Year

Early DecemberForecast

Early JuneForecast

Early AugustForecast

Observed

Named Storms 1950 to 1990 Mean = 9.3

       
1984

-

10

10

12

1985

-

11

10

11

1986

-

8

7

6

1987

-

8

7

7

1988

-

11

11

12

1989

-

7

9

11

1990

-

11

11

14

1991

-

8

7

8

1992

8

8

8

6

1993

11

11

10

8

1994

10

9

7

7

1995

12

12

16

19

1996

8

10

11

13

r

+.038

+.270 **

+.558 ***

Persistence: +.023

Hurricanes 1950 to 1990 Mean = 5.8

       
1984

-

7

7

5

1985

-

8

7

7

1986

-

4

4

4

1987

-

5

4

3

1988

-

7

7

5

1989

-

4

4

7

1990

-

7

6

8

1991

-

4

3

4

1992

4

4

4

4

1993

6

7

6

4

1994

6

5

4

3

1995

8

8

9

11

1996

5

6

7

9

r

+.189

+.251 *

+.403 **

Persistence: +.150

Intense Hurricanes 1950 to 1990 Mean = 2.3

       
1990

-

3

2

1

1991

-

1

0

2

1992

1

1

1

1

1993

3

2

2

1

1994

2

1

1

0

1995

3

3

3

5

1996

2

2

3

6

r

+.107

+.115

+.239 *

Persistence: +.255

Hurricane Days 1950 to 1990 Mean = 23.7

       
1984

-

30

30

18

1985

-

35

30

21

1986

-

15

10

11

1987

-

20

15

5

1988

-

30

30

21

1989

-

15

15

32

1990

-

30

25

27

1991

-

15

8

8

1992

15

15

15

16

1993

25

25

25

10

1994

25

15

12

7

1995

35

35

30

60

1996

20

20

25

45

r

+.129

+.139

+.247 **

Persistence: +.175

Predictions and verifications that were for significantly above the climatological average (11-19 for named storms, 7-12 for hurricanes, 4-7 for intense hurricanes and 31-60 for hurricane days) are indicated by underlined numbers. Those which were below average (4-7 named storms, 2-4 hurricanes, 0-1 intense hurricanes and 4-17 hurricane days) are noted by boldfaced numbers. Skill of the forecasts and of predictions of year-to-year persistence is assessed by r, the agreement coefficient (Gray et al. 1992). Significance is for the r values is given by: ``-" not significant, "*" significant at the 0.10 level, "**" significant at the 0.05 level, and "***" significant at the 0.01 level.




Fig. 2 Tropical storm tracks during September 1 to October 31 for (a) six major El Niño years (65,72,82,87,91,97) and (b) six major La Nina years (70,73,75,88,98,99). The numbers of storm formed in the SE quadrant (dashed box) are respectively 23 for the El Niño years and 8 for the La Nina years. The numbers of storms that northward recurved across 35N are 29 in the El Niño years and 11 in the La Nina years, respectively. Both the differences are statistically significant at the 99% confidence level by the two-sample t-test.