A mechanism of interdecadal variability of tropical cyclone activity over the western North Pacific
- First Online:
- Cite this article as:
- Matsuura, T., Yumoto, M. & Iizuka, S. Climate Dynamics (2003) 21: 105. doi:10.1007/s00382-003-0327-3
- 374 Views
Tropical cyclone (TC) activity in the western North Pacific (WNP) has changed interdecadally with an approximately 20-year period between 1951 and 1999. The cause and mechanism of interdecadal variability of TC frequency in the WNP is investigated using NCEP/NCAR reanalysis and the result obtained from a high-resolution coupled general circulation model (CGCM). The interdecadal variability of TC activity in the WNP correlates with long-term variations in sea surface temperatures (SSTs) in the tropical central Pacific and with those of westerly wind anomalies associated with the monsoon trough that appears over the tropical WNP during the typhoon season of July to October. The westerly wind anomalies at near 10°N show positive feedback with the SST anomalies in the central Pacific. Therefore, the interdecadal variability of TC frequency is related to long-term variations in atmosphere–ocean coupling phenomena in the tropical North Pacific. A 50-year long-run simulation using the high-resolution CGCM showed the robustness of interdecadal variability of TC frequency.
Tropical cyclones (TCs) are the most devastating of all natural disasters in terms of loss of human life, property damage, and other economic consequences. Approximately 80 TCs with maximum sustained surface wind exceeding 17.3 ms–1 are annually generated over the globe (Gray 1979). About one-third of all TCs over the globe occur in the western North Pacific (WNP). Their average annual birthrate is 27.2 during the 30-year period of 1971–2000. Hereafter, we use the term "tropical cyclone (TC)" to mean any of the following: a tropical storm (maximum sustained surface wind speeds between 17.3 ms–1 and 23 ms–1), a severe tropical storm (23–32 ms–1), or a typhoon (>33 ms–1) in the WNP.
TC activity changes over the long term and appears to be related to climatic variations (Raper 1993; Lighthill et al. 1994). One of the most interesting questions is how TC activity will change when global warming occurs due to anthropogenic effects (Broccoli and Manabe 1990; Bengtsson et al. 1996; Sugi et al. 2002). Quite apart from any modification of the climatology of TCs caused by anthropogenic effects, however, there are several pieces of evidence indicating that variations in TC activity have occurred on decadal and longer time scales (Chan and Shi 1996; Chu and Clark 1999; Elsner et al. 2001; Goldenberg et al. 2001; Chu 2002). We now define interdecadal variability as a temporal variation lasting in the order of 10 to 50 years. This variability is important in its own right since an interdecadal time scale reveals the most prominent variability in TC frequency during the 50-year period from 1951 to 1999 (Yumoto and Matsuura 2001). Moreover, when evaluating the relationship between TC activity and global warming, it is important to consider the interdecadal variability of the TC activity, since it is possible to analyze reliable data on observed TCs for a relatively short period of about 50 years.
We were able to capture the interdecadal variability of TC frequency as one of the dominant variations from the analysis of observations during the 49-year period of 1951–1999. To investigate the interdecadal variability of TC frequency, we show the relationship between TC activity and the changes in the dynamic and thermodynamic conditions of both the atmosphere and the ocean. A high-resolution coupled ocean–atmosphere general circulation model (CGCM), developed at the National Research Institute for Earth Science and Disaster Prevention (NIED), was able to reproduce realistic-appearing TCs (Matsuura et al. 1999). We also used this CGCM to investigate the relationship between the interdecadal variability of climate and TC activity in the WNP.
The goal of this study is to reveal the cause and mechanism of interdecadal variability of TC frequency in the WNP using both observational data and simulation data obtained from the CGCM. The data and methodology are described in Sect. 2. In Sect. 3, the interdecadal variability of TC frequency is defined and discussed in the context of the observations and the CGCM simulation. Section 4 shows the relationship between the genesis locations of TCs and the oceanic and atmospheric conditions during the July to October typhoon season. The cause and mechanism of interdecadal variability of TC genesis is shown from the relationship among SST anomaly, zonal wind-stress anomaly, and extension of the monsoon trough in Sect. 5. We summarize the interdecadal feature of TC activity with a schematic representation in Sect. 6.
2.1 Data analysis
We used the data on TC activity in the WNP from the monthly Geophysical Review, issued by the Japan Meteorological Agency (JMA) from 1991 to 1999, and Tropical Cyclone Tracks in the WNP 1951–1990 issued by the Regional Specialized Meteorological Center (RSMC) Tokyo Typhoon Center (1992). It was not until the creation of the Joint Typhoon Warning Center in 1960 that a warning on all TCs was issued systematically. Moreover, satellite monitoring of weather events has become routine since 1965. Therefore, we mainly analyze the reliable data and discuss the relationship between TC frequency and the atmospheric and ocean conditions after 1960.
The smoothing method used in this study to analyze the long-term variability of TC frequency is the 4(3RSR) 2H twice method (Tukey 1977), which involves running medians of the annual number of TCs. This method is explained in detail in Yumoto et al. (submitted 2002). This method is able to capture a broad and smoothed curve for the TC frequency. However the method is inappropriate for addressing the issue of when a change-point takes place in the TC time series (Chu 2002). Although a large swing was noted in the late 1960s as the interannual variability of TC frequency (see Fig. 3), we focus only on the interdecadal variability of TC frequency. Therefore, we consider that the method used here is appropriate for investigating the interdecadal variability of TC frequency. As discussed later, the interdecadal variability shows the maximum peak in the wavelet power spectrum for TC frequency (Yumoto et al. 2002).
The monthly mean global sea surface temperature (SST) data used in this study are Reynolds SSTs (Reynolds and Smith 1995). We also used the National Center for Environmental Prediction (NCEP)/National Center for Atmospheric Research (NCAR) Reanalysis data set on a 2.5° × 2.5° grid (Kalnay et al. 1996).
2.2 Coupled general circulation model description and climatology
2.2.1 Atmospheric model
The atmospheric component of our CGCM is the global spectral model (GSM 8911), which was used for weather forecasting at JMA (JMA 1993). In the horizontal direction, all variables are truncated triangularly at total wave number 106 (T106). There are 21 levels in the vertical from the surface to about 10 hPa. The radiation in this model changes daily and seasonally: longwave radiation (Katayama 1972; Sugi et al. 1990) and shortwave radiation (Lacis and Hansen 1974). The physics package includes a convection scheme (Kuo 1974; Tiedtke 1985) and a cloud fraction using Saito and Baba's (1988) method. The Level 2 turbulence closure scheme by Mellor and Yamada (1974) is used as the planetary boundary layer processes. The surface eddy fluxes of momentum, sensible heat, and latent heat are computed using bulk formulas for stability dependence by Louis et al. (1982). The effect of the gravity waves induced by subgrid scale topography is parametrized according to the scheme developed by Iwasaki et al. (1989). The Simple Biosphere scheme (SiB) (Sellers et al. 1986; Sato et al. 1989) is used for land surface processes.
2.2.2 Ocean model
The ocean model used in our CGCM is based on Modular Ocean Model 2-2 developed at the Geophysical Fluid Dynamics Laboratory (Pacanowski 1996). The model domain covers the global oceans except for 70°N in the Arctic Ocean. The horizontal resolution is 1.125° in longitude and 0.5625° in latitude. There are 37 levels in the vertical direction, with the upper 400 m divided into 25 levels. We employed a biharmonic type of mixing for the horizontal subgrid scale mixing parametrization: both the horizontal eddy viscosity coefficient and the horizontal eddy diffusion coefficient are –5 × 1020 cm4 s–1. A Richardson number-dependent diffusivity by Pacanowski and Philander (1981) is used in the vertical mixing scheme. Penetrative solar radiation is parametrized using a formula given by Paulson and Simpson (1977). The external mode is solved using an implicit free-surface method developed by Dukowicz and Smith (1994); and a convection scheme introduced by Marotzke (1991) is used to eliminate unstable conditions immediately. Since no sea-ice model is included in the model, the temperature and salinity near the southern and northern boundaries are restored to the climatological values (Levitus 1982) and the observed climatological sea-ice distribution is always used in our atmospheric general circulation model (AGCM). River runoff into the ocean is not considered.
We began the CGCM run using daily atmospheric conditions recorded on January 1, 1989, which were obtained from the integration of the atmospheric model from April 1, 1988 using the Meteorological Office Sea Surface Temperature of the United Kingdom Meteorological Office as the boundary condition. The oceanic model was spun up for 10 years from the static state using the annual mean temperature and salinity provided by Levitus (1982) as the initial conditions. Under these conditions, the atmospheric and oceanic models were coupled through daily mean SST and atmospheric fluxes. No correction is applied to the heat, momentum, and water fluxes between the atmosphere and the ocean. We integrated the data for 59 years and analyzed the simulated results for the 50-year period from year 10 to year 59.
2.2.4 Model climate during the typhoon season
We verified the model climatology for the typhoon season (July–October) obtained from the simulation. To compare the simulation with the observations we used the Reynolds SST (Reynolds and Smith 1995), precipitation data (Xie and Arkin 1996), and the reanalysis data set provided by NCEP/NCAR as the other atmospheric data (Kalnay et al. 1996). Climatology of the observational data, except for precipitation, is averaged from 1951 to 1999.
In the eastern equatorial Pacific, on the other hand, a cold tongue extends westward along the equator from the coast of Peru, also similar to that in the observations. The SST structure shows a meridionally asymmetric pattern. Compared to the observations, however, the model cold tongue has a deficiency: it extends far into the western equatorial Pacific. Moreover, the SST minimum peaks of the model cold tongue split into two areas: one area along the equator and another area off the South American coast. The model SSTs off the South American coast are higher than the observed SSTs and their zonal gradients are weaker. The bias towards higher SSTs in the model increases during the boreal winter–spring (not shown). This difference between the model and observed SSTs also seems to be found in other CGCMs (Mechoso et al. 1995). This is because the downward shortwave radiation is too intense off the western side of continents: the effect of low-level marine stratus is not modeled in the CGCMs. The model SST values are 1–2 °C higher than observed in the tropical oceans (the Indian Ocean and the eastern Pacific).
We now address the atmospheric circulation in the typhoon season (Fig. 1b). The observed atmospheric circulation shows that easterly winds extend over the tropical Pacific. In the Indian Ocean, the trade winds extend from near 30°S to the equator. Once these pass over the equator, the wind direction turns to the east and they change into westerly winds. The westerly winds extend to the Philippines and then meet the Pacific trade winds accompanied by intense precipitation (see Fig. 1d).
The model lower atmospheric circulation in the typhoon season corresponds to the observed circulation (Figs. 1b, 2b). In particular, trade winds show similar pattern between the model simulation and observations over the tropical Pacific and south Indian Ocean. However, the wind direction over the Bay of Bengal through to over the Philippines is different in the observations, which show westerly winds, and the simulation, which shows southwesterly winds.
In the boreal summer, a low-pressure area is formed in the neighborhood of Tibet and high-pressure areas lie over the North Pacific (Figs. 1c, 2c). In the Southern Hemisphere, a high-pressure belt appears close to 30°S all the year round, whereas the seasonal change is less than in the Northern hemisphere. In the typhoon season, the model Pacific high develops farther westward than the observations and the monsoon trough does not penetrate to the WNP.
Although the model precipitation belt in mid-latitude shifts farther poleward than that obtained from the observations, in addition to its westerly location, the model global precipitation roughly corresponds to the observed distribution of precipitation. The model distribution of precipitation in the tropics is very different from that of the observations: the model South Pacific convergence zone (SPCZ) is present along the south of the equator all the year round, and the precipitation belt in the tropical Pacific is symmetrical around the equator (double inter-tropical convergence zone, ITCZ). In the observations, this double ITCZ appears only in early spring. The double ITCZ is seen in the AGCM (T106) stand-alone simulation (Kar et al. 1996), too. When the horizontal resolution is T42, however, this double ITCZ is not seen (Sugi et al. 1995a, b).
In the western tropical Pacific through to the Indian Ocean, significant differences are found in the simulated precipitation fields during the southwest monsoon season (Figs. 1d, 2d). There is less precipitation over near the Philippines and over the Bay of Bengal and more precipitation over the eastern Tibetan Plateau than the actual observed precipitation (see Figs. 1d, 2d). This bias also appears in the AGCM stand-alone simulation and in the low resolution T42 AGCM (Sugi et al. 1995a, b).
3 Long-term variability of TC frequency
We also obtained the interdecadal variability of TC frequency analyzing the result obtained from a 50-year simulation of the high-resolution CGCM (Matsuura et al. 1999; Yumoto et al. 2002) (Fig. 3b). We define a model TC as a typhoon-like vortex meeting the following criteria: it is located in the WNP between 120°E and 180° longitude except for the South China Sea, has a maximum wind speed at 850 hPa exceeding 17.0 ms–1 near a point of minimum sea level pressure, has a relative vorticity at 850 hPa that exceeds 1.2 × 10–4 s–1, and has a minimum sea level pressure less than 1008 hPa. Using these definitions, 741 model TCs appeared over the WNP except the South China Sea in the 50-year integration. Thirty-three of these TCs occurred on the east side of 180° longitude, moved toward the west, and entered the WNP region.
The CGCM simulation shows that there are wavelet spectrum peaks of 5-year interannual variability and of 20-year interdecadal variability for typhoon frequency (Fig. 4a in Yumoto et al. 2002). This 20-year period matches the results obtained from the observations, as shown in Fig. 3a, although the interannual variability is more intense than that in the observations. HFPs for model TC genesis are in two periods during years 25–36 (HFP years 25–36) and years 53–59 (HFP years 53–59) and LFPs during years 18–24 (LFP years 18–24) and years 37–52 (LFP years 37–52).
4 Relationship between TC frequency and oceanic and atmospheric conditions
It is noted that favorable conditions for the atmosphere and ocean are needed for TC development (Gray 1979). First, an ocean temperature of at least 26 °C is needed down to a depth of about 60 m. The energy source of TCs is latent heat, which is released to the atmosphere in the form of water vapor. In general, evaporation is active with an increase in SSTs and more water vapor finds its way into the atmospheric boundary layer. Second, the atmosphere must be capable of permitting deep convection to occur: the stratification of the atmosphere should be kept conditionally unstable so that cumulus convection occurs. Third, the dynamic balance of the TC is approximated by the gradient relation; and the Coriolis force plays an important role. Therefore, TCs form mainly northward of 5°N, not over the equator. Fourth, there is a requirement for the prior existence at low altitude of a substantial level of cyclonic vorticity. Fifth, for TC generation, the vertical shear of the horizontal wind should be low to allow incipient warm air to accumulate. The presence of the monsoon trough (a region of low pressure associated with a monsoon), with its relatively low vertical wind shear and high relative vorticity, is extremely favorable for tropical cyclogenesis (Gray 1968; Ramage 1974). It is therefore likely that the TC genesis is influenced by large-scale atmospheric and oceanic conditions.
Figure 5b shows the differences in SSTs between the HFP (years 25–36) and LFP (years 18–24) of the CGCM. The model SSTs of the HFP in the tropical central and western North Pacific are higher than those of the adjoining LFP, as in the observations. The statistically significant areas occupy narrower regions in the simulation than that in the observations (see Fig. 5a, b). They also shift westward to the western Pacific in the simulation. The feature of SST anomalies in Fig. 5a, b is similar to the well-known interdecadal pattern of SST anomalies that show a wedge-shaped pattern: the large positive SST anomalies extend to the tropical central and eastern Pacific.
Focusing on the high-variable area of TC frequency (10°N–20°N, 120°E–180°) (see Fig. 5 in Yumoto and Matsuura 2001), we find that the differences in relative vorticity at 850 hPa (Fig. 5c, d) and divergence at 200 hPa (Fig. 5e, f) show anomalous positive vorticity and anomalous positive divergence for both the observations and the simulation. From this result, we can see anomalous cyclonic vorticity in the lower atmosphere and divergence in the upper atmosphere over the oceanic area where TCs actively occur. In addition, the precipitation rate over the WNP between 10°N and 20°N was greater in the HFP than in the LFP for both the observations and the simulation (Fig. 5g, h). The atmospheric and oceanic conditions shown in Fig. 5 for the HFP in the TC genesis area are more favorable for TC genesis than that for the LFP.
A marked difference of TC frequency between the HFP and LFP appears over an ocean area of 10°N–20°N, 130°E–170°E (Yumoto and Matsuura 2001; Yumoto et al. 2002). Over this ocean area are present monsoon shear lines and confluence regions between the easterly and westerly flow at the eastern extremity of the monsoon trough or gyre (see Fig. 6). This is different from the observations in that the intrusion of a monsoon trough in the CGCM simulation does not appear in the tropical WNP. However, strong horizontal shear extends from 120°E toward the eastern tropical Pacific between the zonal band of 10°N–20°N (Fig. 4b, d).
5 Correlation of interdecadal TC genesis with the decadal/interdecadal climatology in the tropical North Pacific
Decadal/interdecadal climate variations in the tropical Pacific are of interest and have been actively investigated (Latif 1998; Kleeman et al. 1999; Knutson and Manabe 1998; Yukimoto et al. 2000). We discuss now whether the interdecadal TC frequency is related to the decadal/interdecadal climate variations in the tropical North Pacific.
Correlation coefficients between the number of tropical cyclones (TC) and selected three areas' SST anomalies for pre typhoon season (March to June), typhoon season (July to October), and post-typhoon season (November to February) for observations
Pre-typhoon season (MAMJ)
Typhoon season (JASO)
Post-typhoon season (NDJF)
Correlation coefficients between the number of TCs and the SST anomalies of three ocean areas for the CGCM simulation
Pre-typhoon season (MAMJ)
Typhoon season (JASO)
Post-typhoon season (NDJF)
For the simulation, the zonal wind anomalies in the tropical WNP also correlated with the SST anomalies in ocean area II (Fig. 9b). Although in the simulation the monsoon trough does not intrude into the tropical WNP, the intensive westerly anomalies match the intensive horizontal shear regions (see Figs. 4d and 7b). The horizontal shear line extends farther eastward in the HFPs than in the LFPs.
We have filtered out a prominent abrupt change at 1976/1977: the tendency of negative anomalies appears in the whole basin of the tropical Pacific during the first half before 1976/1977 and the tendency of positive anomalies appears during the latter half. This tendency is prominent on the side of the tropical eastern Pacific (not shown).
We can separate two ocean areas between 5°N and 15°N in the North Pacific: 120°E–170°E and 170°E–120°W in the longitudinal direction to enable discussions of variations in SST anomalies, zonal wind stress anomalies, and net surface heat-flux anomalies. Although the number of TCs varies mostly in the ocean area 10°N–20°N, 130°E–170°E as long-term variations (Yumoto and Matsuura 2001), we have averaged the physical quantities during 5°N–15°N where the zonal wind anomalies at 850 hPa vary prominently (see Fig. 7). SST anomalies propagated westward from 1955 to 1980 and after 1980 they showed a standing oscillation pattern divided at 160°E. The TC frequency correlates with SST anomalies more in the central North Pacific at 180°–120°W than in the WNP at 120°E–170°E. The zonal wind stress anomalies are positive during the HFPs and negative during the LFPs and propagate eastward between 120°E and 170°E.
Peaks of zonal wind-stress anomalies appear over near 150°W; in particular, a positive peak in the long-time trend of TC frequency in 1967 corresponds to a positive peak of zonal wind-stress anomalies, and a negative peak of TC frequency at 1975 corresponds to a negative peak of zonal wind stress anomalies (see Fig. 10b). However, a positive peak in the long-term trend of TC frequency in 1990 is ahead of a positive peak of zonal wind-stress anomalies. Although peaks of zonal wind stress anomalies appear over the central North Pacific, the long-term variations of TC frequency correlate closely with them in the WNP (120°E–170°E).
The maximum and minimum peaks of zonal wind-stress anomalies in the central North Pacific correspond closely to those of SST anomalies there. The positive and negative net surface heat-flux anomalies during 120°E–170°E correspond to those of zonal wind-stress anomalies: during the positive wind stress anomalies, trade winds weaken and net surface heat-flux from the ocean to the atmosphere decreases.
We have confirmed the interdecadal variability of TC frequency whose period is about 20 years, using observational data and data obtained from a CGCM simulation. It was also demonstrated that the interdecadal variability of TC activity correlates with that of zonal wind stresses in the tropical western Pacific and that of SSTs in the tropical western and central Pacific (Tables 1 and 2).
SST anomalies in the North Pacific from 5°N to 15°N show different patterns before 1980 and after 1980. Before the 1980s, the SST anomalies propagated westward from near 160°W. After the 1980s, however, the SST anomalies showed a standing oscillation pattern divided at 160°E. When the SST anomalies propagate westward, positive wind-evaporation-SST (WES) feedback occurs (Xie and Philander 1994). In brief, the trade winds weaken because of westerly wind anomalies. The release of latent heat-flux from the ocean decreases, and the SST anomalies increase. The WES feedback propagates westward accompanied by an ocean–atmosphere coupled instability phenomenon (Xie 1996).
The standing oscillation of SST anomalies after 1980 may be due to the ENSO-like delayed-oscillator phenomenon (Yukimoto et al. 2000). Concerning the standing pattern of SST anomalies, a change of thermocline depth propagates westward as downwelling and upwelling Rossby waves (not shown). In the tropical WNP (west of 160°E) the positive SST anomalies correspond to the positive heat content anomalies. In this case, positive feedback occurs among wind stress curl, thermocline depth, and SST anomalies. In short, once negative wind-stress curl anomalies occur near 170°E, downwelling Rossby waves are excited and propagate westward. At the same time, the SST anomalies west of 160°E increase.
A positive peak in TC frequency in 1967 corresponds to a positive peak in the WES feedback mechanism, and a negative peak in TC frequency in 1975 corresponds to a negative peak in the WES feedback mechanism. During a positive peak in TC frequency in the first half of 1990, the westerly wind anomalies become positive in the tropical WNP, and the release of latent heat from the ocean decreased. However, during this time the SST change was decided not only by surface heat flux but also by ocean dynamics.
The intensity of westerly wind anomalies is accompanied by far eastward extension of a monsoon trough during the typhoon season (see Figs. 6 and 7). The monsoon shear line and the confluence region between easterly and westerly flow are major areas where TCs are generated (Ritchie and Holland 1999). Actually, the most prominent ocean area for interdecadal variability of TC genesis is between 10°N–20°N, 130°E–170°E (Yumoto and Matsuura 2001; Yumoto et al. 2002). In the HFPs the monsoon trough extends farther eastward (Fig. 6) and the monsoon shear line becomes stronger than in the LFPs.
It was also shown that the TC genesis in the central North Pacific is related to the extension of the monsoon trough (Clark and Chu 2002). In the El Niño hurricane season, the TC frequency in the central North Pacific is three times greater than that in the La Niña hurricane season. Clark and Chu (2002) showed that during the El Niño summer or autumn, the monsoon trough intrudes to near 180°. On the other hand, during the La Niña autumn, the monsoon trough intrudes to near 145°E. Briefly, the monsoon trough extends farther eastward in the El Niño hurricane season than in the La Niña hurricane season.
Cyclonic circulation appears over the oceanic area at 850 hPa between east of the Philippines and south of Japan, and is associated with active mesoscale convection (see Fig. 7). The reverse relationship appears in the LFPs. This result corresponds to the relative vorticity differences at 850 hPa between the HFP (1986–94) and the LFP (1973–85) for the observations and between the HFP (years 25–36) and LFP (years 18–24) of the simulation (see Fig. 5c, d): the cyclonic (anticyclonic) circulation area in Fig. 12 has a tendency towards positive (negative) relative vorticity as shown in Fig. 5c, d. Cyclonic circulation may be due to the accumulation of active convection including active TC genesis there and the anticyclonic circulation may be generated from the Rossby response of active convection. At the present stage, we are not able to conclude that anomalous cyclonic circulation plays an important role in active TC genesis or that active TC genesis derives anomalous cyclonic circulation.
Our CGCM was able to reproduce the interdecadal variability of TC genesis frequency over a 50 year-long run. The interannual variability of TC frequency in the simulation was more pronounced than that in the observations from 1951 to 1999 (Yumoto et al. 2002). The genesis rate of TCs in the simulation was lower over the ocean area between 20°N–30°N and 130°E–170°E than that in the observations; and the genesis locations of TCs in the simulation were dispersed zonally, unlike those in the observations (see Fig. 4). This is because the climate of CGCM SSTs in the tropical North Pacific shows that the zonal asymmetry is weaker than in the observations: the development of the warm pool is weak (see Figs. 1a and 2a). The SST bias in the CGCM influences the distributions of strong horizontal shear and weak vertical shear of atmospheric winds, which play an important role in TC genesis (see Fig. 4).
We were able to obtain similar results between the NCEP/NCAR reanalysis and the CGCM simulation for the relationship between the interdecadal variability of TC frequency and global oceanic and atmospheric conditions (Fig. 5). This indicates that there are favorable conditions for the atmospheric and oceanic fields for long-term TC genesis in the CGCM simulation. However, the fine structures of SST, horizontal vorticity at 850 hPa, vertical divergence at 200 hPa, and precipitation rate differences appeared in the simulation, not in the observations (Fig. 5) because of the T106 resolution in the CGCM and T42 resolution in the NCEP/NCAR reanalysis.
The chief deficiency of the CGCM climate in the tropical WNP is that no monsoon trough develops during summer–autumn. The monsoon trough is related to the genesis of TCs (Ritchie and Holland 1999); however, the interdecadal variability of TC genesis still appeared in our CGCM simulation. The lack of a monsoon trough in the CGCM simulation may also result in fewer TCs than in the observations.
In future, by embedding sophisticated convection schemes (for example, Arakawa and Shubert 1974), we may be able to improve the reproduction of the monsoon trough (Iizuka et al. personal communication). To predict long-term variation in TC activity, we need a higher resolution CGCM. Although our model is still incomplete, we believe that the CGCM simulation and analysis through comparing observations opens the way to studies on the relationship between long-term TC activity and interdecadal climate change.
The interdecadal variability of hurricane activity in the eastern Pacific may be related to the abrupt change at 1976/1977 in the tropical North Pacific (Chu and Clark 1999). However, the interdecadal variability of typhoon activity depends on the decadal variability of monsoon westerlies in the western tropical Pacific, not on the abrupt change seen in 1976/1977. We conclude that the interdecadal variability of typhoon frequency is associated with the decadal/interdecadal variability of atmosphere–ocean coupling phenomena in the tropical North Pacific.
The authors would like to thank Dr. R. Kawamura, Dr. P-S. Chu, and an anonymous reviewer for helpful suggestions. This research was conducted as a part of the project study "Study on extreme weather events and water-related disasters due to climate change" in the National Research Institute for Earth Science and Disaster Prevention of Japan. We are greatful to the Japan Meteorological Agency and Geophysical Fluid Dynamics Laboratory/NOAA for providing the code of GCMs used in the present study.