inﬂuence of the global warming on tropical …...merous studies have been conducted. emanuel...

Journal of the Meteorological Society of Japan, Vol. 80, No. 2, pp. 249--272, 2002 249

Influence of the Global Warming on Tropical Cyclone Climatology:

An Experiment with the JMA Global Model

Masato SUGI, Akira NODA

Meteorological Research Institute, Tsukuba, Japan

and

Nobuo SATO

Japan Meteorological Agency, Tokyo, Japan

(Manuscript received 26 December 2000, in revised form 26 December 2001)

Abstract

The influence of the global warming on tropical cyclones has been examined using a high resolutionAGCM. Two ten-year integrations were performed with the JMA global model at T106 horizontal resolu-tion. For the control experiment, the observed SST for the period 1979–1988 is prescribed, while for thedoubling CO2 (2 � CO2) experiment, SST anomaly due to the global warming estimated from a coupledmodel transient CO2 experiment (Tokioka et al. 1995) is added to the SST used in the control experiment.

The results of experiments show that a significant reduction in the frequency of tropical cyclones ispossible in response to the greenhouse gas-induced global warming. The most significant decrease is in-dicated over the North Pacific. On the other hand, a considerable increase in tropical cyclone frequency isindicated for the North Atlantic. As for the maximum intensity of tropical cyclones, no significant changehas been noted.

It has been found that the regional change in tropical cyclone frequency is closely related to the dis-tribution of the SST anomaly, and the change in convective activity associated with it. The results of theexperiment indicate that the change in tropical cyclogenesis is strongly controlled by dynamical factorsassociated with the change in SST distribution, rather than the thermodynamical factors associated withthe change in absolute value of local SST.

On the other hand, for the decrease in the global total number of tropical cyclones on doubling CO2, aweakening of tropical circulation associated with the stabilization of the atmosphere (the increase in drystatic stability), seems to be responsible. It is found that the rate of increase in the tropical precipitationdue to the global warming is much less than the rate of increase in the atmospheric moisture. With thislittle increase in precipitation (convective heating), a considerable increase in the dry static stability ofthe atmosphere leads to a weakening of the tropical circulation.

1. Introduction

The influence on the tropical cyclone clima-tology of the global warming, due to the in-

crease of carbon dioxide and other greenhousegases, is not only a scientifically interestingproblem but also of particular importance fromthe viewpoint of societal concern. From a sim-ple thought, it is expected that the increase inatmospheric moisture due to the global warm-ing may lead to an increase in the intensity,and frequency of tropical cyclones. Is this sim-ple thought true? To answer this question, nu-

Corresponding author: Masato Sugi, Meteorologi-cal Research Institute, 1-1 Nagamine, Tsukuba305-0052, Japan.E-mail: [email protected]( 2002, Meteorological Society of Japan

merous studies have been conducted. Emanuel(1987) suggested, based on his simple theoreti-cal model, that the possible maximum intensityof a tropical cyclone might be increased by theglobal warming. Ryan (1992) and Royer et al.(1998) applied an empirical formula by Gray(1975) to the results of global warming experi-ments using GCMs. They found a significantincrease in tropical cyclone frequency due tothe global warming, but noted a limitation ofapplying an empirical formula based on thepresent climate to the double CO2 climate thatis substantially different from the present cli-mate.

An alternative approach to the problem is toutilize a GCM, although grid sizes of the mostGCMs are too coarse to simulate tropical cyclo-nes realistically. Broccoli and Manabe (1992)first conducted such an experiment using theGFDL model with two different resolutions (R15and R30, corresponding grid sizes are 5� � 7:5�

and 2.5� � 3:75�, respectively). In their experi-ment, they employed two different treatmentsof clouds: fixed and variable. The differenttreatment of clouds, regardless of the resolu-tion of the model, gave an opposite answer tothe problem: the model with fixed cloud hasshown an increase, while the model with vari-able cloud has shown a decrease in the simu-lated tropical cyclone frequency on doublingCO2. Haarsma et al. (1993) has conducted asimilar experiment using the UKMO GCM with2.5� � 3:75� latitude-longitude grid. Their ex-periment has shown a significant increase intropical cyclone frequency and intensity on dou-bling CO2.

There have been criticisms on the GCM ex-periments of the influence of the global warm-ing on tropical cyclones (Evans 1992; Lighthillet al. 1994). They argue that the resolutionof the GCMs is too coarse to simulate tropicalcyclones properly, and the results of experi-ments are not physically acceptable. Recently,Bengtsson et al. (1996) reported the results ofan experiment with a high resolution (T106,1.1� � 1:1� grid) ECHAM model. In contrast toHaarsma et al. (1993), they found a significantreduction in the number of tropical cyclones ondoubling CO2. Sugi et al. (1997) also conductedan experiment using a high resolution (T106,1.1� � 1:1� grid) JMA model. They also found asignificant decrease in the tropical cyclone fre-

quency on doubling CO2. More recently, Yoshi-mura et al. (1999), May and Andersen (1999)and McDonald (1999) conducted experimentsusing high resolution models: JMA model,ECHAM model and UKMO model, respectively,to study the influence of the global warmingon tropical cyclone climatology. An experimentwith a high resolution regional climate model(Walsh and Ryan 2000), and an experimentusing a very high resolution hurricane model(Knutson et al. 1998), were performed to studythe possible change in tropical cyclone intensitydue to the global warming.

In the present study, the influence of theglobal warming on tropical cyclones has beenexamined using a high resolution AGCM. Aftera brief description of the model and experi-ments in section 2, the results of the experi-ment briefly reported in Sugi et al. (1997) willbe presented more in detail in section 3. Theinterpretation of the experimental results: thepossible reasons for the reduction in the tropi-cal cyclone frequency due to the global warmingis discussed in section 4, followed by a sum-mary in section 5.

2. Model and experiments

2.1 ModelThe model used for this study is the T106

GCM version of the Japan MeteorologicalAgency (JMA) global model JMA-GSM8911(JMA 1993; Sugi et al. 1990). The model is aspectral model with triangular truncation attotal wave number 106. It has unevenly spaced21 levels, the top of which is placed at 10 hPa.The model includes comprehensive physicalprocesses: radiation (Sugi et al. 1990; Lacis andHansen 1974), cumulus convection (Kuo 1974),large scale condensation, planetary boundarylayer processes (Louis et al. 1982), simplifiedbiosphere (SiB) (Sellers et al. 1986; Sato et al.1989), and orographic gravity wave drag (Iwa-saki et al. 1989).

The simulated climate with the T42 versionof the JMA global model is reported by Sugiet al. (1995a, b). Generally, the model simu-lates large-scale features of the climate reason-ably well. One major deficiency of the modelthat is most relevant to the present study is thesimulation of tropical precipitation. The modeltends to underestimate precipitation over thewestern North Pacific in the summer season.

Journal of the Meteorological Society of Japan250 Vol. 80, No. 2

The sensitivity of the simulated climate tohorizontal resolution, and cumulus parameter-ization scheme has been examined. It was foundthat large-scale features of the simulated cli-mate generally is not very sensitive to horizon-tal resolution of the model, whereas the tropicalprecipitation of the model is much more sensi-tive to a cumulus parameterization scheme,than to horizontal resolution (Sugi et al. 1995c).

2.2 ExperimentsA preliminary one-year integration of the

model was first conducted to examine the abil-ity of the model in simulating tropical cyclones.The integration was conducted for the periodfrom April 1988 to March 1989, using the ob-served SST for the same period. Several realis-tic typhoons were simulated in this integration

(Fig. 1). The lowest central pressure of thetyphoons in the model is 980 hPa, maximumwind is 30 m s�1, temperature anomaly of thewarm core is 5�C, and the radius of the maxi-mum wind is about 200 km. Such structure ofthe simulated typhoon is not sufficient, butmuch better than the simulation with lowerresolution models (Sugi et al. 1994).

To simulate present tropical cyclone clima-tology, a ten-year integration of the model (con-trol experiment) has been conducted using theAMIP observed SST for the period from 1979 to1988. To examine the influence of global warm-ing, another ten-year integration correspondingto the 2 � CO2 climate (2 � CO2 experiment)has been performed. For this purpose, the re-sults of the transient CO2 experiment withMRI-CGCM (Tokioka et al. 1995) have been

Fig. 1. A typhoon simulated with the T106 JMA global model. (a) A weak tropical depression wasborn on the day 195 of the simulation at around 9�N, 150�E. (b) The tropical depression movedwest-north-westward and developed slowly. (c) It reached to the east of the Philippines on the day201 and turned its direction to north-eastward, and further developed. (d) The simulated typhoonapproached south of Japan on the day 204.

M. SUGI, A. NODA and N. SATO 251April 2002

used. First, the linear trend of the annual meanSST is computed to remove the natural decadalscale variation of SST in the transient CO2 ex-periment. Then, the increase of annual meanSST is computed for the period of 60–70 yearsof the transient CO2 experiment that corre-sponds to the period of doubling CO2. The SSTincrease due to the global warming thus com-puted (Fig. 2), together with the mean seasonalvariation of SST increase during the period of60–70 years, is added to the SST used in thecontrol experiment. Using this SST and thedoubled CO2 concentration in the atmosphere,the second integration has been conducted fromthe same atmospheric initial condition as thecontrol experiment. A climatological sea icedistribution is used for both the control and2 � CO2 experiments. We believe that the in-fluence of sea ice change due to the globalwarming may be negligible on the tropical cy-clone climatology.

2.3 Criteria for selecting tropical cyclonesBecause of the insufficient resolution of the

model for simulating tropical cyclones, thestructure and intensity of simulated tropicalcyclones are different from the real tropical cy-clones. Therefore, the criteria for selecting si-mulated tropical cyclones should be different

from that of real tropical cyclones. Broccoli andManabe (1992), Haarsma et al. (1993), Bengts-son (1995), Tsusui and Kasahara (1995) em-ployed different criteria. We followed Bengts-son et al. (1995):

1) Sea level pressure: Candidate of tropical cy-clone center is a grid point at which the sealevel pressure takes local minimum, and thevalue is less than 1020 hPa.

2) Vorticity: Near the cyclone center, 850 hPavorticity z850b 35 � 10�6 sec�1.

3) Maximum wind: Maximum wind speed at850 hPa near the cyclone center Vmaxb15 m sec�1.

4) Warm core: The average temperature dif-ference from the area mean of surroundingregion at 300 hPa, 500 hPa, 700 hPa and850 hPa exceeds 3�C.

5) Upper level wind speed: Maximum windspeed at 300 hPa is less than the maximumwind speed at 850 hPa near the cyclone cen-ter.

6) Duration : Tropical cyclones should continueat least 2 days.

The criterion 5) effectively removes the ex-tratropical cyclones (Bengtsson et al. 1995),and we do not need any prescribed geographi-cal restriction for selecting tropical cyclones in

Fig. 2. The sea surface temperature change (�C) due to the global warming estimated by the lineartrend of a transient CO2 experiment with the MRI-CGCM.


the model. The results of the simulation werestored every 24 hours. An automatic trackingprogram was developed to check the above cri-teria, and search a tropical cyclone near thelocation of the tropical cyclone found on theprevious day.

3. Results

3.1 Control experimentFigures 3(a) and 3(b) show the tracks of ob-

served and simulated tropical cyclones for the10-year period from 1979 to 1988. The total

Fig. 3. Tropical cyclone tracks for a ten-year period. (a) Observed tracks for 1979–1988 based on theUS Navy Best Track Dataset. (b) Tracks in the control experiment. (c) Tracks in the 2 � CO2 ex-periment.


number of simulated tropical cyclones is 881 forthe 10-year period, which is a little more thanthe observed number 812 in the same period(see Table 1). Overall geographical distributionof tropical cyclone is simulated well, althoughthere are some discrepancies. In the model,tropical cyclones are simulated over the south-east coast of Brazil where no tropical cyclonesare observed in reality. Tropical cyclones arealso simulated but not observed around theMalay peninsula. From Fig. 3(a) and 3(b), wesee that the density of simulated tracks is lessthan the observation over the North Pacific,and North Atlantic. In the North Pacific, thenumber of simulated tropical cyclones is con-siderably less than the observed (see Table 1).Moreover, the simulated tracks are generallyshorter than those of the observed. The averagelifetime of the simulated tropical cyclones is 3.4days, which is considerably shorter than 5.2days, the average observed lifetime of typhoons.

Table 1 shows the number of observed (col-umn OBS), and simulated (column CNTL),

tropical cyclones in each ocean basin. Thenumber of simulated tropical cyclones in theNorth West Pacific are considerably less thanthe observed number, while those in the NorthIndian Ocean and South West Pacific Oceanare much larger than those observed. About18% less tropical cyclones are simulated thanthe observation in the Northern Hemisphere,while significantly larger number of tropicalcyclones are simulated than the observation inthe Southern Hemisphere. Note that the differ-ence in the numbers of simulated, and observedtropical cyclones in some ocean basin is largerthan the standard deviation (shown in the pa-renthesis) of the interannual variation of thenumber of simulated or observed tropical cyclo-nes, and therefore the difference is statisticallysignificant. The difference in the global totalnumbers of simulated and observed tropical cy-clones, is less than the corresponding standarddeviations.

The seasonal variation of the number of si-mulated and observed tropical cyclones in each

Table 1. Number of tropical cyclones by ocean basin. Numbers in parenthesis show the standarddeviations. Italic letter indicates that the value is not statistically significant at 95% confidencelevel.


ocean basin is plotted in Fig. 4. As noted above,the total number of simulated tropical cyclonesis much less than those observed in the North-ern Hemisphere, while it is much larger thanthose observed in the Southern Hemisphere.

The model fails to simulate the late summerpeak in the Northern Hemisphere, while it sim-ulates too many tropical cyclones in the South-ern Hemisphere winter and spring seasons. Inthe North West Pacific, the maximum frequen-cies of tropical cyclones are simulated in Octo-ber and November, two months later than theobservation. The value of the peak frequencyis significantly less in the North West Pacific,North Atlantic and North Eastern Pacific.Therefore, the amplitude of seasonal variationof simulated tropical cyclone frequency in theNorthern Hemisphere, is much less than thatobserved. The amplitude of seasonal variationof the simulated tropical cyclone is also lessthan the observation in the Southern Hemi-sphere, because too many tropical cyclones aresimulated from April to December in the SouthWest Pacific. In the North Indian Ocean, thedouble peaks in May and November are simu-lated well, although the number of simulatedtropical cyclones is considerably larger than theobserved number.

Figure 5 shows the inter-annual variationof the global total number of simulated andobserved tropical cyclones during the 10-yearperiod. The standard deviation of the inter-annual variation of simulated tropical cyclonefrequency is 10.4, and comparable to the ob-served standard deviation 12.3 (see Table 1). Itshould be noted that the similar magnitude ofinter-annual variation of tropical cyclone fre-quency has been simulated with a GCM usinga climatological SST as a lower boundary con-dition (Bengtesson et al. 1995; Tsutsui et al.1996). This suggests that the major portion ofinter-annual variation of tropical cyclone fre-quency is not forced by inter-annual variationof SST. Therefore, it is not so unreasonable thatthe correlation between the simulated andobserved inter-annual variation of tropical cy-clone frequency is weak (0.15), although anobserved SST is used as a lower boundary con-dition in the present study.

Figure 6 shows the frequency distributionsof simulated, and observed tropical cyclones asa function of maximum wind speed for each

one. The frequency distribution of simulatedtropical cyclones exhibits a sharp peak at 20–25 m s�1. The largest maximum wind speed ofsimulated tropical cyclones is 40–50 m s�1. Onthe other hand, the frequency distribution ofobserved tropical cyclones has a weak peak at20–25 m s�1, and extends over a broad range ofmaximum wind speed up to 80–85 m s�1. It isclear from the figure that the model fails tosimulate very intense storms. This indicatesthat the horizontal resolution of the model isstill too coarse to simulate tropical cyclonesvery realistically. To summarize the results ofcontrol experiment; it may be said that themodel generally simulates the observed tropi-cal cyclone climatology well, although thereare some deficiencies in the simulation of geo-graphical distribution, seasonal variation andintensities of tropical cyclones. The reason forthe discrepancy between the observed and si-mulated tropical cyclone climatology may be animportant subject. Generally speaking, it maybe related to deficiencies of the model in simu-lating tropical convection and tropical cyclonedevelopment. The analysis of more specific rea-son for each discrepancy is left for a futurestudy.

3.2 2 � CO2 experimentFigure 3(c) shows the tracks of tropical cyclo-

nes simulated in the 2 � CO2 experiment. Com-pared with the result of the control experimentshown in Fig 3(b), we see a significant reduc-tion in the number of simulated tropical cyclo-nes over the North West Pacific and North EastPacific Oceans. In contrast, over the North At-lantic Ocean, we can see more tropical cyclonessimulated in the 2 � CO2 experiment thanthe control experiment. The number of tropicalcyclones simulated in the 2 � CO2 experi-ment is shown in the third column of Table 1.In the Table 1, also shown are the difference(2 � CO2 � control) and the ratio (2 � CO2/con-trol) of the numbers of tropical cyclones in thetwo experiments. Most significant reduction isseen over the North West Pacific, where thenumber of simulated tropical cyclones in the2 � CO2 experiment is 7.5 per year, which isone-third of the number for the control experi-ment (21.9). A similar reduction rate is seenover the North East Pacific, but it is not statis-tically significant because of the large inter-


Fig. 4. Seasonal variation of tropical cyclone frequency by ocean basin. Dotted lines indicate the ob-served frequency. (a) Northern Hemisphere, (b) North Indian Ocean, (c) North Western Pacific, (d)North Atlantic and North East Pacific, (e) Southern Hemisphere, (f ) South Indian Ocean, (g) SouthWest Pacific, and (h) South Atlantic and South East Pacific (see Table 1 for the definition of re-gions). Solid and dashed lines show the frequencies of simulated tropical cyclones in the controlexperiment and the 2 � CO2 experiment, respectively.


annual variation over this region in the controlexperiment. The South Indian Ocean also showsa substantial reduction in the tropical cyclonefrequency on doubling CO2. On the other hand,over the North Atlantic Ocean, we see a signif-icant increase in the number of tropical cyclo-nes, from 7.4 per year in the control experimentto 11.4 per year in the 2 � CO2 experiment,which is more than 60% increase. Over theNorth Indian Ocean, there is a slight increase,although it is not statistically significant. FromTable 1, we note that a considerable reductionin the tropical cyclone frequency over the SouthPacific Ocean on doubling CO2. Figure 3 showsthat this reduction mostly occurs in the north-ern part of the South Pacific Ocean. In thesouthern part of the South Pacific Ocean to theeast of Australia, we can see more tropical cy-clones simulated in the 2 � CO2 experimentthan in the control experiment. This seems tobe associated with the southward shift of theregion of tropical cyclogenesis (probably SPCZ).

In Fig. 4, the seasonal variation of tropicalcyclone frequency over each ocean basin simu-lated in the 2 � CO2 experiment is shown bydashed curves. In the North West Pacific, thereduction in the tropical cyclone frequency inthe 2 � CO2 experiment is seen in all seasons.In the North Atlantic and North East Pacific,

more tropical cyclones are simulated in the2 � CO2 experiment than the control experi-ment in late autumn (October and November),while less tropical cyclones are simulated inearly summer (June and July). In the NorthIndian Ocean, the peak in May is more promi-nent in the 2 � CO2 experiment.

The inter-annual variation of the number oftropical cyclones simulated in the 2 � CO2 ex-periment is shown by long-dashed curve in Fig.5. We can see a clear separation of this curvefrom the solid curve (control experiment), andthe short-dashed curve (observed), although themaximum number occurred in the year 1985 ofthe 2 � CO2 experiment is slightly larger thanthe minimum number occurred in 1986 of thecontrol experiment. The global total number ofsimulated tropical cyclone in the 2 � CO2 ex-

Fig. 5. Inter-annual variation of tropicalcyclone frequency. Dotted curve in-dicates the observed frequency based onthe US Navy Best Track Deataset.Solid and dashed lines show the fre-quencies of simulated tropical cyclonesin the control experiment and the2 � CO2 experiment, respectively.

Fig. 6. Tropical cyclone frequency distri-bution as a function of maximum windspeed of each cyclone. (a) Raw fre-quency during the ten year period. (b)Normalized frequency. Dotted line in-dicates the observed frequency. Solidand dashed lines show the frequenciesof simulated tropical cyclones in thecontrol experiment and the 2 � CO2 ex-periment, respectively.


periment is 58.5 per year, which is about 34%less than the control experiment (see Table 1).The standard deviation of the inter-annualvariation of the number of tropical cyclonessimulated in the 2 � CO2 experiment is 7.7,which is a little less than that of the controlexperiment. The difference in the global totalnumber of tropical cyclones simulated in thetwo experiments is 29.6 per year, which is threetimes larger as the standard deviation, andtherefore, the difference is highly statisticallysignificant.

In Fig. 6, the frequency distribution as afunction of maximum wind speed for the2 � CO2 experiment is plotted by a dottedcurve. From Fig. 6(a), we can see the overallreduction in the frequency of tropical cycloneson doubling CO2. From the normalized fre-quency distribution curves in Fig. 6(b), it isnoted that the relative frequency distributionsare very similar for the two experiments. Thereis a slight increase in the normalized frequencyon doubling CO2 at the region of maximumwind speed exceeding 40 m s�1, but it is notstatistically significant. From Fig. 6, we mayconclude that there is no significant change inthe intensity of tropical cyclones on doublingCO2.

From Fig. 6, we note that the simulated trop-ical cyclones are much weaker than the ob-served, and this raises a concern about the se-lection criteria for simulated tropical cyclones.If we employ different criteria, would the re-sults be different? Figure 7 shows the frequencydistribution of tropical disturbances as a func-tion of intensity of the vorticity at 850 hPa.Here, the center of a ‘‘tropical disturbance’’is defined as a grid point where the 850 hPavorticity takes local maximum value on the2.5� � 2:5� latitude-longitude grid between 30�Nand 30�S. For the entire tropical belt and an-nual average case, the frequency distribution isplotted in Fig. 7(a). In this case, the frequencyof tropical disturbances, with vorticity intensityexceeding 23 � 10�6 sec�1, is less in the 2 � CO2experiment than in the control experiment.Note that the selection criteria used in thisstudy employs 35 � 10�6 sec�1 as a thresholdvalue of 850 hPa vorticity intensity. From Fig.7(a) we can see that a change in this thresholdvalue would not affect the conclusion that thenumber of tropical cyclone simulated in the

2 � CO2 experiment, is significantly less thanthe control experiment.

Figures 7(b) and Fig. 7(c) show the frequencydistributions of tropical disturbances over theNorth West Pacific Ocean, and the North At-

Fig. 7. Frequency distribution of tropicaldisturbances as a function of vorticity.(a) Annual frequency for the entiretropics. (b) September–November fre-quency for the North West Pacific. (c)September–November frequency for theNorth Atlantic. Line with open circleindicates the control experiment, withsolid circle 2 � CO2 experiment, andopen square the difference of the two,respectively.


lantic Ocean, in the autumn season (Septem-ber–November). In the Pacific Ocean case, wesee a significant reduction on doubling CO2 inthe frequency of tropical disturbances with vor-ticity larger than 30 � 10�6 sec�1. In contrast,in the Atlantic Ocean case, an increase in thefrequency on doubling CO2 is seen over a widerange of vorticity intensity. Thus, the changein frequency distribution of vorticity over theNorth West Pacific, and North Atlantic, is inagreement with the change in tropical cyclonefrequency over these regions shown in Fig. 4and Table 1. It may be said, therefore, that theregional changes in tropical cyclone frequencyalso would not be much affected by the value ofvorticity criteria for tropical cyclones.

4. Discussion

The results of experiments indicate that thenumber of tropical cyclones may significantlybe reduced due to the global warming. On theother hand, there is not a clear evidence show-ing significant change in the intensity of tropi-cal cyclones. In this section, we will examinepossible reasons for the reduction in tropicalcyclone frequencies on doubling CO2. First, wediscuss the regional change, and then the globalchange in the tropical cyclone frequencies.

4.1 Regional changeFrom Fig. 3, we see the most significant re-

duction in the number of tropical cyclones overthe North Pacific Ocean, while we see a consid-erable increase over the North Atlantic Ocean.From Fig. 2, we note that the SST anomalygiven to the 2 � CO2 experiment is relativelysmall over the central part of the Pacific Oceancompared with the other regions of tropicalocean. Also we note that the SST anomaly islarge over the Atlantic Ocean. We can see thatthe regions where the reductions in the tropicalcyclone frequency occur in Fig. 3 agree with theregions where the SST anomalies are less than1�C (lightly shaded regions). On the otherhand, over the regions where the SST anoma-lies are larger than 1�C (medium dark areas),more tropical cyclones are simulated in the2 � CO2 experiment than the control experi-ment. The agreement between the SST anom-aly pattern and the tropical cyclone frequencychange pattern indicates that the anomaloustropical circulation associated with the SST

anomaly play an important role in the tropicalcyclogenesis. It should be noted that the SSTanomaly over the Pacific Ocean is positive,though it is less than the other regions. There-fore, if the absolute value of SST is a dominantfactor for the tropical cyclogenesis, more tropi-cal cyclones should be simulated in the 2 � CO2experiment than the control experiment evenover the North Pacific Ocean. The result of ex-periment indicates that it is not the absolutevalue of SST that dominates the tropical cyclo-genesis. Rather, a tropical circulation pat-tern associated with the SST distribution is adominant factor for the regional tropical cyclo-genesis.

Figure 8(a) shows the difference between theannual mean precipitation rate of the controlexperiment, and that of the 2 � CO2 experi-ment, while Fig. 8(b) shows the SST anomaly(minus 1�C) given to the 2 � CO2 experiment.From Fig. 8(a) and 8(b), we can see a goodagreement between the pattern of precipitationdifference, and the pattern of SST anomaly. Aswe have noted above, the pattern of tropicalcyclone frequency change from control experi-ment to 2 � CO2 experiment resembles thepattern of SST anomaly. Thus, the patterns oftropical cyclone frequency change, precipitationchange and SST anomaly resemble each other.This suggests a chain of links from tropical SSTanomaly to convective activity (or precipitation),from convective activity to tropical circulation,and from tropical circulation to tropical cyclo-genesis.

To understand the regional changes in theconvective activity in the tropics, the changesin the distribution of the mean fields of quan-tities associated with tropical convection areexamined. Figure 9 shows the differences(2 � CO2 � control) in the annual mean verticalp-velocity at 500 hPa, surface specific humidity,dry static stability and convective available po-tential energy (CAPE). In Fig. 9(a), anomalousupward motions are seen over the Atlanticocean and the region extending south-eastwardfrom North Indian Ocean, to the western southPacific Ocean. Over these regions, anomalousconvection is positive in Fig. 8(a). On the otherhand, anomalous downward motions are seenover the most past of the Pacific Ocean andSouth Indian Ocean, where anomalous convec-tion is negative. Figure 9(b) shows that the


surface specific humidity over the ocean is 5–15% larger in the 2 � CO2 experiment than inthe control experiment. Although the differencein the surface specific humidity is positiveeverywhere, a large difference of more than10% is seen over the regions where the anom-alous convection is positive in Fig. 8 (a).

The change in dry static stability is shown inFig. 9 (c). Here the dry static stability is definedas the difference in potential temperature at200 hPa and 1000 hPa. The figure shows thatthe dry static stability is 6–8% larger in the2 � CO2 experiment than the control experi-ment over the most part of the tropical belt. Itshould be noted that the increase in the drystatic stability is closely related to the increasein the moist adiabatic lapse rate in the warmerand wetter atmosphere. In contrast to therather uniform increase in the dry static stabil-ity, the change in the convective instability asmeasured by CAPE shows distinct regionaldistribution (Fig. 9(d)). It increases over the re-gions where the anomalous convection is posi-tive in Fig 8a. It should be noted that the CAPEdecreases over the most part of the PacificOcean, and South Indian Ocean, although the

surface moisture, and therefore, the moist staticenergy of the surface air increases everywherein the tropics. The reason for the decrease inCAPE over these regions is that the reductionin CAPE due to the increased dry static stabil-ity is larger than the increase in CAPE due tothe increase of moist static energy of surfaceair. To further confirm this idea, the verticalprofiles of potential temperature and moiststatic energy at the grid point in the westernPacific, where the change in CAPE due to theglobal warming is close to zero, are shown inFig. 10. The potential temperature difference isabout one degree at 1000 hPa and it increaseswith height up to 200 hPa, where the differenceis about four degrees. On the other hand, thedifference in the saturated moist static energyis nearly the same at all levels between thesurface and 200 hPa. The magnitude of the dif-ference is almost the same as the difference inthe moist static energy at the surface. Thus, thedifference in the CAPE at this grid point isclose to zero. At the other grid point where theCPAE increases (decreases) due to the globalwarming, the effect of moist static energy in-crease at the surface is larger (less) than the

Fig. 8. (a) Annual mean precipitation intensity difference (mm/day) between the control experimentand the 2 � CO2 experiment. (b) Annual mean SST difference (�C) between the control experimentand the 2 � CO2 experiment. One degree is subtracted from the SST difference.


effect of dry static energy increases in the upperlevels.

In order to understand the regional change inthe tropical cyclone frequency in response tothe CO2 increase, yearly genesis parameter(YGP) proposed by Gray (1975) is examined.Gray identified six factors which contribute to

tropical cyclogenesis. They are three dynamicalfactors: Coriolis parameter, low level relativevorticity and vertical shear of horizontal wind;and three thermodynamical factors: thermalenergy of ocean (or SST), moist instability oflower troposphere and relative humidity. Graycombined these six factors into a single param-

Fig. 9. Differences in the annual mean quantities related to tropical circulation between the controlexperiment and the 2 � CO2 experiment. (a) vertical p-velocity at 500 hPa (hPa/h), (b) surfacespecific humidity (difference in %) , (c) dry static stability (difference in %), (d) CAPE (J kg�1).


eter called YGP, which accounts for the clima-tological geographical distribution of tropicalcyclogenesis fairy well.

To find out which factor most contributes tothe change in tropical cyclone frequency, YGPis computed for the control experiment, andthe 2 � CO2 experiment. The difference in theYGP of the two experiments is shown in Fig.11 together with the contributions to the YGPchange from the five factors constituting theYGP (no contribution from Coriolis factorswhich does not change on doubling CO2). Sincethe YGP is a sum of four seasonal genesis pa-rameters (SGP), the contribution from eachfactor to the SGP change is calculated first bycalculating the SGP change associated with thechange in the respective factor. Although thesecond order terms are neglected in this calcu-lation of SGP change, the sum of contributionsfrom the five factors (Fig. 11(b)–(f )) calculatedin this way is nearly the same as the totalchange in the YGP (Fig. 11(a)).

From Fig. 11(a), it is seen that the change inthe YGP is positive over most regions, exceptfor the western central Pacific. The total num-ber of tropical cyclones estimated from YGP is83.5 per year for the control experiment, and130.4 per year for the 2 � CO2 experiment. Thechange in YGP indicates a 56% increase in the

number of tropical cyclone due to the CO2 in-crease, in spite of the 34% decrease in the num-ber of simulated tropical cyclones. The YGPclearly overestimates the number of tropicalcyclones in the 2 � CO2 experiment. This over-estimation is mainly due to the ocean thermalenergy factor (SST factor) as shown in Fig.11(d).

It is interesting to note that the changes inthe two dynamical factors, low level relativevorticity (Fig. 11(b)) and vertical shear of hori-zontal wind (c), contribute to the reduction ofthe YGP, particularly over the Pacific. This in-dicates that the dynamical factors play impor-tant roles in the significant reduction in the si-mulated tropical cyclone frequency, particularlyover the Pacific. On the other hand, the threethermodynamical factors, SST (d), moist insta-bility (e) and relative humidity (f ), mostly con-tribute to the increase of the YGP. Positive SSTanomalies are considered to be favorable fortropical cyclogenesis as indicated by the ther-modynamical factors in the YGP. However, rel-atively little SST anomaly compared with thesurrounding regions, even though it is positive,would lead to a circulation that is unfavorablefor tropical cyclogenesis. The results of the ex-periment indicate that the change in tropicalcyclogenesis is strongly controlled by dynamical

Fig. 10. Annual mean vertical profile of (a) potential temperature (K), and (b) moist static energy(103 J kg�1) at 10�N, 160�E. Solid and dashed lines indicate the control experiment and 2 � CO2experiments. Thick lines in (b) show the saturated moist static energy.


factors associated with the change in SST dis-tribution, rather than the thermodynamicalfactors associated with the change in absolutevalue of local SST.

In summary, the changes in tropical circula-tion associated with the SST anomaly distribu-tion play an important role in the regionalchange of tropical cyclone frequency due to theglobal warming. In the previous section wehave noted that the major portion of inter-an-nual variation of the simulated tropical cyclonefrequency is not forced by SST variation. How-ever, this does not exclude a possibility of aninfluence of SST anomaly on regional tropicalcyclogenesis. It should be noted that the SSTanomaly used in the 2 � CO2 experiment is al-most the same pattern throughout the 10 yearperiod of the experiment, and therefore, verypersistent. Such persistent SST anomaly has

never been observed in the past. In this regard,it is interesting to note that the number oftyphoons in 1998 summer is only 4, only onethird of the normal year. The SST anomalypattern in 1998 summer was persistent, andsimilar to the pattern of that of the 2 � CO2experiment. Therefore, a significant reductionin typhoon frequency indicated by the presentstudy may not be unrealistic, if the SST anom-aly is reasonable.

4.2 Global changeIn the previous subsection, it has been shown

that the regional change in the tropical cyclonefrequency due to the global warming is closelyrelated to the distribution of SST increase.Over the regions with relatively large SST in-crease, convective activity and tropical cyclonefrequency also increase. On the other hand, the

Fig. 11. (a) Difference in the YGP between the control experiment and 2 � CO2 experiment. (b)–(f )Contributions to the difference in YGP from the changes in constituting factors: (b) relative vor-ticity, (c) vertical shear of horizontal wind.


regions with relatively small SST increase,convective activity and tropical cyclone fre-quency decrease, even though the SST does in-crease. The result of the experiment shows asignificant reduction in the global tropical cy-clone frequency due to the global warming.Dose this correspond to a overall weakening ofconvective activity?

To examine the change in overall convectiveactivity, zonal mean precipitation and evapora-tion for the control experiment and the 2 � CO2experiment are shown in Fig. 12. Both theglobal average precipitation and evaporationincrease about 2% due to the global warming.Both of them show larger increase in the extra-

tropics. The precipitation and evaporation inthe tropics increase 1% and 1.4%, respectively(Table 2). In the Southern Hemisphere tropics,the zonally averaged precipitation of the2 � CO2 experiment is less than that of thecontrol experiment. On the other hand, precip-itable water in the tropical atmosphere in-creases as much as 14% due to the globalwarming (Table 2). Why does the tropical pre-cipitation increase only 1% in spite of the 14%increase of precipitable water? If the tropicalcirculation does not change on doubling CO2,the precipitation should increase with increas-ing atmospheric moisture. That the precipi-tation does not much increase in spite of a

Fig. 11. (d) ocean thermal energy (SST), (e) moist instability, (f ) relative humidity.


significant increase of atmospheric moisturesuggests a weakening of tropical circulation ondoubling CO2.

To examine the change in the tropical circu-lation, mass flux at 500 hPa is computed (Table3). The upward (downward) flux in the Table 3is the vertical p-velocity averaged over theupward (downward) motion region within thetropics (30�N–30�S) multiplied by the frac-

tional area. The instantaneous mass flux iscomputed using snapshots with 24 hour in-terval. Since upward motion and downwardmotion take place at the same grid point atdifferent time, the magnitude of mass fluxescalculated from time averaged vertical velocitywould be smaller than that calculated from theinstantaneous vertical velocity. Indeed, themass fluxes calculated from the monthly meanvertical velocity or annual mean vertical veloc-ity in Table 3 are smaller than the instanta-neous mass fluxes. By a similar reason, the in-tensity of spatially averaged circulation such as(zonally averaged) Hadley circulation or (mer-idionally averaged) Walker circulation, whichare often used as a measure of tropical circula-tion intensity, would underestimate the inten-sity of tropical circulation by cancellation ofupward motion and downward motion. Here wetake the instantaneous mass flux as an index oftropical circulation intensity. As the tropicalregion (30�N–30�S) considered here is not aclosed region, magnitudes of the upward massflux and downward mass flux are not exactlythe same. As the upward motion area of the

Fig. 12. (a) Zonally averaged annual mean precipitation (mm/day) in the control experiment and2 � CO2 experiment, and their difference. (b) Zonally averaged annual mean evaporation (mm/day)in the control experiment and 2 � CO2 experiment, and their difference. Solid and dotteded curvesin the top panels indicate the control experiment and 2 � CO2 experiment, respectively. The dif-fernce is shown in the bottom panels.

Table 2. Variables related to tropical cir-culation


mean Hadley circulation is mostly confined inthe tropical area (30�N–30�S), we consider theinstantaneous upward mass flux in Table 3 asan index of the tropical circulation intensity(Table 2). This mass flux is about 6% less forthe 2 � CO2 experiment than the contral ex-periment. Thus, it has been shown that thetropical circulation weakens in response to theglobal warming.

The wakening of tropical circulation due toglobal warming may be understood by energybalance of the tropical atmosphere. The ap-proximate energy equation in the tropics maybe expressed as,

oqy

qpA

y

T

Q

Cp; ð1Þ

(Holton 1979; Kuntson and Manabe 1995).Here we consider a schematic tropical circula-tion as shown in Fig. 13. Then, the energy bal-ance in the upward motion region and down-ward motion region at 500 hPa may be writtenas,

Muqy

qp

��

��A

y

T

1

CpðQc � QRÞAu; ð2Þ

Mdqy

qp

��

��A

y

T

1

CpQRAd; ð3Þ

where MuðMdÞ and AuðAdÞ are the mass flux,

and fractional area of the upward (downward)motion area, Qc and QR are the condensationheating rate and radiative cooling rate, respec-tively. The change in condensation heating ratedue to the global warming may be proportionalto the change in precipitation rate. Therefore,from Table 2, the Qc would increase about 1%due to the global warning. We note that the

Table 3. Tropical mass fluxes

Fig. 13. A schematic diagram showingthe energy balance of tropical circula-tion.


radiative cooling rate also increases about 1%.On the other hand, mass flux is about 6% lessin the 2 � CO2 experiment than the control ex-periment, while the atmospheric stability (de-fined as the potential temperature difference at200 hPa and 1000 hPa) is about 6% larger inthe 2 � CO2 experiment. We note that the drystatic stability is nearly uniform in the tropics,and we assumed for simplicity that the changein dry static stability over the upward motionregion is the same as that over the downwardmotion region. Thus, we see that the increasein dry static stability and the decrease in massflux nearly cancel each other and balance withthe small increase in the heating terms.

In summary, there is a balance as illustratedin Fig. 14. The tropical atmosphere becomesmore stable (dry static stability increases) inresponse to the global warming, while the con-densation heating and radiative cooling, whichare the driving force of tropical circulation, in-crease only a little. Hence, the tropical circula-tion weakens due to the global warming. Onthe other hand, as the atmospheric moisturein the tropics significantly increases with theglobal warming, condensation heating can in-crease even with weaker circulation. In otherwords, in the warmer tropical atmosphere withmore moisture and larger dry static stability,the energy balance is achieved by a weakeningof the atmospheric circulation.

A weakening of tropical circulation due to theglobal warming has been reported in severalGCM studies. Bengtsson et al. (1996) pointedout the weakening of Hadley cell and tropicalhydrological cycle on doubling CO2 in their

experiment. Knutson and Manabe (1995) re-ported that the intensity of the upward motionover the tropical western Pacific slightly de-creases when the atmospheric CO2 concentra-tion is quadrapled, although the precipitationover the same region is significantly enhanced.Kitoh et al. (1996) has shown that the intensityof the south-westerly monsoon flow becomesweaker on doubling CO2, although the mon-soon precipitation increases.

Next, we examine the influence of the globalwarming upon individual tropical disturbances.Figure 15 shows the frequency distribution oftropical disturbances as a function of precipita-tion intensity, vertical p-velocity at 500 hPaand vorticity at 850 hPa. Here, we have se-lected the grid points in the tropics (30�N–30�S) where the 850 hPa vorticity takes localmaximum value, and we regard such grid pointsas the centers of tropical disturbances. For thevalues of precipitation and vertical velocity of adisturbance, we take the maximum value with-in the two grid distances (@ 500 km) from thecenter of the disturbance. The frequency dis-tributions with respect to precipitation inten-sity are almost the same for control and 2 � CO2experiments. On the other hand, in the fre-quency distribution as a function of vertical ve-locity and vorticity, we can see a tendency thatthe frequency of tropical disturbances withstrong (weak) vertical velocity or vorticity de-creases (increases) on doubling CO2.

To further confirm this tendency, two dimen-sional frequency distributions of disturbancesas a function of precipitation, vertical p-velocityat 500 hPa and vorcitity at 850 hPa are shownin Fig. 16. The top panels show the frequencydistribution with respect to precipitation andvertical velocity for the control and 2 � CO2experiments, and the difference between thetwo experiments. The pattern of distributionindicates a positive correlation between precip-itation, and vertical velocity of disturbances. Itshould be noted that both the axes are scaled inlogarithm. Therefore, for a certain value of pre-cipitation (vertical velocity), the values of ver-tical velocity (precipitation) extend over aconsiderably wide range. In the left panel, thedistributions for control and 2 � CO2 experi-ments mostly overlap. The difference betweenthe two experiments on the right panel indicatesthat for the same value of precipitation the

Fig. 14. Changes in quantities related totropical circulation on doubling CO2.


vertical velocity of disturbances tends to beweaker in the 2 � CO2 experiment than in thecontrol experiment.

On the other hand, as shown in the middlepanels, for the frequency distribution with re-spect to vertical velocity and vorticity, the dif-ference between the two experiments extendsalong the 45 degree sloped line. Both the verti-cal velocity, and vorticity tend to be weaker inthe 2 � CO2 experiment than the control ex-periment. Thus the relationship between vor-ticity and vertical velocity is rather invarianton doubling CO2. As expected from the aboveresults, for the same amount of precipitation,the vorticity of disturbances tends to be weakerin the 2 � CO2 experiment than in the controlexperiment, as shown in the bottom panels.These results may be closely related to the sta-bilization of tropical atmosphere on doublingCO2. The equation (1) indicates that for thesame amount of heating (precipitation), verti-cal velocity should be weaker in the stabilizedatmosphere. As in the discussion for the energybalance of the tropical circulation, the conden-sation heating associated with a tropical dis-turbance would not necessarily decrease withweakening of the vertical velocity because of theatmospheric moisture increase on doubling CO2.

4.3 Inter-model differenceAs mentioned in the introduction, there have

been several studies using GCMs on the influ-ence of global warming on tropical cyclone cli-matology. Some of them have shown an increasein global total number of tropical cyclones dueto the global warming, while the others haveshown a decrease. Regional change in the trop-ical cyclone frequency is also different frommodel to model. The reason for such differencesamong the models will be discussed in thissubsection.

According to the discussion in the previoussubsection, there are two opposing factors forthe change in global tropical cyclone frequency:reduction due to the stabilization of the atmo-sphere (increase in dry static stability), and in-crease due to the enhanced precipitation (con-densation heating). In the experiment of thepresent study, the increase in the tropical pre-cipitation is not significant and the increase indry static stability of the atmosphere is domi-nant, leading to a significant reduction in the

Fig. 15. Normalized frequency distribu-tion of tropical disturbances (%) as afunction of (a) precipitation (mm/day),(b) verticl p-velocity (hPa/h) at 500 hPa,and (c) vorticity (sec�1) at 850 hPa.Solid and dashed curves indicate thecontrol experiment and 2 � CO2 experi-ment, respectively.


Fig. 16. Two dimensional normalized frequency distribution of tropical disturbances (%) as a func-tion of (a) precipitation and vertical p-velocity at 500 hPa, (b) verticl p-velocity at 500 hPa andvorticity at 850 hPa, and (c) precipitation and vorticity at 850 hPa. Solid and dashed curves indi-cate the control experiment and 2 � CO2 experiment, respectively. (d)–(f ): differences between thecontrol experiment and the 2 � CO2 experiment.


global total number of tropical cyclones in the2 � CO2 experiment. However, the amount ofincrease in tropical precipitation on doublingCO2 is very much different from model to model.It takes values between �1% and 5% accordingto our calculation using the data from eightmodels collected at IPCC Data DistributionCenter. In the present study, the increase intropical precipitation is 1% and relatively smallcompared with other models. It may be possiblethat the global frequency of tropical cycloneshow less decrease or even increase in the modelwith more tropical precipitation increase ondoubling CO2. It is noted that the amount ofincrease in tropical precipitation is well bal-anced with the amount of increase in the radi-ative cooling in the present study (Table 2). Itis suggested that the change in the tropicalprecipitation due to the global warming isclosely related to the change in the radiativecooling, and therefore the change in clouds. Inthis regard, it is interesting to note that the re-sults of experiments by Broccoli and Manabe(1990) differ by different treatment of clouds inthe same model. They found an increase of trop-ical cyclone frequency in the experiment withprescribed climatological clouds, while a de-crease in the experiment with variable clouds.

As for the difference among the models in theregional change of tropical cyclone frequency,the difference in the distribution of SST anom-aly used in the experiments seems to be a ma-jor factor. It is known that there are largeintermodel differences in the SST anomaly dis-tribution among the results of global warmingexperiments with coupled atmosphere-oceanGCMs (Noda et al. 1999). In section 4.1, it hasbeen noted that the regions where the SST in-crease due to the global warming is relativelylarge (small) coincide with the regions wherethe convective activity (precipitation) and trop-ical cyclone frequency are increased (decreased).

From above discussion it is indicated thata reliable estimate of the distribution of SSTchange and a quantitative estimate of tropicalprecipitation change are essential for a reliableprediction of possible change in tropical cyclonefrequency and distribution.

5. Summary

In the present study, the influence of theglobal warming on tropical cyclones has been

examined using a high resolution AGCM. Twoten-year integrations were performed with theJMA global model at T106 horizontal resolu-tion. For the control experiment, the observedSST for the period 1979–1988 is prescribed,while for the 2 � CO2 experiment, SST anom-aly due to the global warming estimated froma coupled model transient CO2 experiment(Tokioka et al. 1995), is added to the SST usedin the control experiment.

The results of experiments show that a sig-nificant decrease in the frequency of tropicalcyclones is possible in response to the green-house gas-induced global warming. Most sig-nificant decrease is indicated over the NorthPacific. On the other hand, a considerable in-crease in tropical cyclone frequency is indicatedfor the North Atlantic. As for the maximumintensity of tropical cyclones, no significantchange has been noted.

It has been found that the regional change intropical cyclone frequency is closely related tothe distribution of the SST anomaly and thechange in convective activity associated withit. Over the regions where the SST anomalyis relatively small compared with surroundingregions, even though it is positive, convectiveactivity and tropical cyclone frequency tendto decrease. This indicates that the climateof convective activity, and tropical cyclone fre-quency distribution, depends on dynamicalfactors associated with the SST distribution,rather than thermodynamical factors asso-ciated with the absolute value of local SST.

As for the decrease in the global total numberof tropical cyclones on doubling CO2, a weak-ening of tropical circulation associated with thestabilization of the atmosphere (the increase indry static stability) seems to be responsible forit. It is found that the rate of increase in thetropical precipitation due to the global warm-ing is much less than the rate of increase in theatmospheric moisture. With this little increasein precipitation (convective heating), a consid-erable increase in the dry static stability of theatmosphere leads to a weakening of the tropicalcirculation.

There are considerable differences among theglobal or regional climate changes in tropicalcyclone frequency indicated by different GCMs.A large difference in the amount of tropicalprecipitation increase seems to be responsible


for the difference in the global climate change,while differences in the SST anomaly used inthe respective model seem to be responsible forthe difference in the regional climate change. Inother words, for a reliable prediction of futureclimate change in tropical cyclone frequency, areliable estimate of SST change and tropicalprecipitation change is necessary.

The model used for the present study is ahigh-resolution model as a present-day AGCM.However, it is clear that the resolution of themodel is still too coarse for studying a possibleclimate change of tropical cyclones, particularlystudying the change in intensity of mature stagetropical cyclones. In the near future, it wouldbe possible to do experiments with much higherresolution global models. Then, we will be ableto simulate tropical cyclones much more realis-tically, and increase reliability of the experi-ment. To simulate tropical cyclones realistically,however, not only increasing the resolution ofmodels, but also introducing more sophisticatedphysics, particularly convection process, is nec-essary. For example, Yamasaki (1977) pointedout an important role of drag force and evapo-ration of rain drops in the early stage of devel-opment of a tropical cyclone. Such processes arenot included in the prameterization of cumulusconvection in the present-day GCMs, but maybe required in the future to simulate develop-ment of tropical cyclones more realistically.

The present study indicates that, for a reli-able prediction of future climate change in trop-ical cyclone climatology, not only a realisticsimulation of tropical cyclones with high-resolution models, but also a reliable estimateof SST change with a low resolution coupledGCMs is needed. A reliable prediction of thechange in large scale tropical circulation andprecipitation is also important.

Acknowledgment

This study was first conducted as a part of‘‘Study of Disaster Prediction in Global Hydro-logical Processes’’ at the National Research In-stitute for Earth Science and Disaster Preven-tion (NIED). Then, it was continued as ‘‘EarthSystem Modeling Project’’ funded by the Scienceand Technology Agency, and completed in ajoint project of Meteorological Research Insti-tute and Frontier Research System for GlobalChange. The computation of the experiments

was carried out on CRY-YMP2 at NIED. Theauthors thank Professors T. Matsuno and S.Manabe for their comments and encourage-ment throughout this study.

References

Bengtsson, L., M. Botzet and M. Esch, 1995: Hurri-cane-type vortices in a general circulationmodel. Tellus, 47A, 175–196.

———, M. Botzet and M. Esch, 1996: Will green-house gas-induced warming over the next 50years lead to higher frequency and greater in-tensity of hurricanes? Tellus, 48A, 57–73.

Broccoli, A.J. and S. Manabe, 1990: Can existing cli-mate models be used to study anthropogenicchanges in tropical cyclone climate? Geophys.Res. Let., 17, 1917–1920.

Emanuel, K.A., 1987: The dependence of hurricaneintensity on climate, Nature, 326, 483–485.

Evans, J.L., 1992: Comment on ‘‘Can existing cli-mate models be used to study anthropogenicchanges in tropical cyclone climate’’. Geophys.Res. Let., 19, 1523–1524.

Gray, W.M., 1975: Tropical cyclone genesis. Dept. ofAtmos. Sci. Paper No. 234, Colorado StateUniv., Ft. Collins, CO.

Haarsma, R.J., J.F.B. Mitchell and C.A. Senior,1992: Tropical Disturbances in a GCM, Clim.Dyn., 8, 247–257.

Iwasaki, T., S. Yamada and K. Tada, 1989: A pa-rameterization scheme of orographic gravitywave drag with the different vertical partition-ings, Part I: Impact on medium range fore-casts. J. Meteor. Soc. Japan, 67, 11–27.

Knutson, T. and S. Manabe, 1995: Time mean re-sponse over the tropical Pacific to increasedCO2 in a coupled ocean-atmosphere model. J.Climate, 8, 2181–2199.

———, R.E. Tuleya and Y. Kurihara, 1998: Simu-lated increase of hurricane intensities in aCO2-warmed climate. Science, 279, 1018–1020.

Kuo, H.L., 1974: Further studies of the influenceof cumulus convection on large scale flow. J.Atmos. Sci., 31, 1232–1240.

Lacis, A.A. and J.E. Hansen, 1974: A parameter-ization for the absorption of solar radiation inthe earth’s atmosphere. J. Atmos. Sci., 31,118–133.

Lighthill, J., G. Holland, W. Gray, C. Landsea, G.Criag, J. Evans, Y. Kurihara and C. Guard,1994: Global climate change and tropical cy-clones. Bull. Amer. Meteor. Soc., 75, 2147–2157.

Louis, J., M. Tiedtke and J.-F. Geleyn, 1982: A pa-rameterization shame of orographic gravity


wave drag with the different vertical partition-ings, Part I: Impact on medium rage forecasts.J. Meteor. Soc. Japan, 67, 11–27.

May W. and U. Andersen, 1999: A time-slice experi-ment with the ECHAM4 AGCM: Changes inthe occurrence of storms. Forth InternationalConference on Modelling of Global ClimateChange and Variability. 13–17 September1999, Max Planck Institute for Meteorology,Hamburg.

McDonald, R., 1999: Changes in tropical cyclones asgreenhouse gases are increased. DETR Report,March 1999. Hadley Centre for Climate Pre-diction and Research.

Noda, A., K. Yoshimatsu, S. Yukimoto, K. Yama-guchi, and S. Yamaki, 1999: Relationship be-tween natural variability and CO2-inducedwarming pattern: MRI AOGCM Experi-ment. Preprnts, 10th Symp. on Global ChangeStudies, Dallas, Texas, Amer. Metor. Soc., 359–362.

Ryan, B.F., I.G. Watterson and J. Evans, 1992:Tropical cyclone frequencies inferred fromGray’s Yearly Genesis Parameter: Validationof GCM tropical climate. Geophys. Res. Let., 19,1831–1834.

Sato, N., P.J. Sellers, D.A. Randall, E.K. Schneider,J. Shukla, J.L. Kinter III, Y.-T. Hou and E.Albertazzi, 1989: Effects of implementing theSimple Biosphere Model in a general circula-tion model. J. Atmos. Soi., 46, 2757–2787.

Sellers, P.J., Y. Mintz, Y.C. Sud and A. Dalcher.1986: A simple biosphere model (SiB) for usewithin general circulation models. J. Atmos.Sci., 43, 505–531.

Sugi, M. and N. Sato, 1994: Possibility of using theJMA global model for studying tropical cycloneclimatology. Proceedings of International Meet-ing on Numerical Prediction of Tropical Cyclo-nes. Tokyo, Japan, 17–21 January 1994, 211–213.

———, R. Kawamura and N. Sato, 1995a: The cli-mate simulated by the JMA global model, Part1: Global feature, Report of National ResearchInstitute for Earth Science and Disaster Pre-vention, 54, 155–180.

———, R.D. Nair and N. Sato, 1995b: The climatesimulated by the JMA global model, Part 2:Tropical precipitation, Report of National Re-search Institute for Earth Science and DisasterPrevention, 54, 181–197.

——— and N. Sato, 1995c: A ten year integration ofthe T106 JMA global model. AMIP ScientificConference, Monterey, California, 15–19 May1995.

———, A. Noda and N. Sato, 1997: Influence ofglobal warming on tropical cyclone climatology:An experiment with the JMA global model.WMO/TD-No. 792.

Tokioka, T., A. Noda, A. Kitoh, Y. Nikaidou, S. Na-kagawa, T. Motoi, S. Yakimoto and K. Takata,1995: A transient CO2 experiment with theMRI-CGCM—Quick report. J. Meteor. Soc.Japan, 73, 817–826.

Tsusui, J. and A. Kasahara, 1995: Simulated tropicalcyclones using the National Center for Atmo-spheric Research Community Climate Model(CCM2). J. Geophys. Res., 101, D10, 15,013–15,032.

Walsh, K., B.F. Ryan, 2000: Tropical cyclone inten-sity increase near Australia as a result of cli-mate change. J. Climate, 13, 3029–3036.

Ymasaki, M., 1977: The role of surface friction intropical cyclones. J. Meteor. Soc. Japan, 55,559–572.

Yoshimura, J., M. Sugi and A. Noda, 1999: Influenceof greenhouse warming on tropical cyclone fre-quency simulated by a high resolution AGCM.Preprints, 10th Symp. on Global ChangeStudies, Dallas, Texas, Amer. Meteor. Soc.,555–558.


inﬂuence of the global warming on tropical …...merous studies have been conducted. emanuel...

Documents