Abstract

Tropical cyclone–like vortices (TCLVs) in the South Pacific Ocean are studied using the CSIRO Division of Atmospheric Research Limited Area Model (DARLAM), nested in a transient carbon dioxide simulation of the CSIRO Mark2 global coupled GCM. This GCM is able to simulate El Niño–Southern Oscillation (ENSO)–like interannual variations, although the amplitude of these is considerably smaller than observed. A comparison is made between observed geographical variations of cyclone formation caused by ENSO and similar variations simulated by DARLAM. An analysis of the simulated interannual variability of TCLV formation suggests that under La Niña conditions TCLVs tend to occur closer to the coast of Australia, whereas under El Niño conditions TCLVs tend to occur farther eastward, in agreement with observations. Under enhanced greenhouse conditions, this geographical variation continues. In addition, the total number of TCLVs in the South Pacific region decreases in a warmer world. As in previous simulations using DARLAM, there is a southward movement in TCLV occurrence under enhanced greenhouse conditions, although this has not been simulated to date by other climate models. The GCM simulation of sea surface temperatures also exhibits coherent decadal variability that has some similarities to the observed ENSO-like decadal variability. This variability forces decadal variations in TCLV formation that, like ENSO-forced variations, have geographically distinct centers of action.

1. Introduction

The characteristics of tropical cyclones in an enhanced greenhouse climate have been the subject of recent research (Haarsma et al. 1993; Henderson-Sellers et al. 1998). There is a growing consensus that tropical cyclones could be somewhat more intense in a warmer world (Tonkin et al. 1997; Knutson et al. 1998; Knutson and Tuleya 1999; Walsh and Ryan 2000). One recent climate model simulation suggests possible poleward movement of typical tropical cyclone occurrence (Walsh and Katzfey 2000), while other simulations suggest little such movement (Knutson and Tuleya 1999). Conflicting results have been obtained for changes in tropical cyclone numbers, with some simulations suggesting increased numbers in some regions (Tsutsui et al. 1999), others decreased (Bengtsson et al. 1996; Yoshimura et al. 1999), and some little or no change (Walsh and Katzfey 2000). Thus considerable work remains to be performed on this topic.

Other studies have examined the ability of climate models to simulate the observed variability of the formation of model-generated tropical cyclone–like vortices (TCLVs) using climate models forced with observed sea surface temperatures (SSTs; Vitart et al. 1997; Vitart et al. 1999). One aspect of this process that has yet to be studied is the response of the TCLVs simulated in such models to the interannual variability generated by a coupled ocean–atmosphere model.

The ability of general circulation models (GCMs) to generate El Niño–Southern Oscillation (ENSO)–like variations is improving rapidly (e.g., Wilson 2001, manuscript submitted to J. Climate). As a result, confidence is increasing that these models will also be able to make predictions regarding the effect of climate change on ENSO. ENSO has a substantial effect on the formation and tracks of observed tropical cyclones in a number of regions of the globe, including the North Atlantic (Goldenberg and Shapiro 1996) and the southwest Pacific (Hastings 1990; Evans and Allan 1992; Basher and Zheng 1995). Any substantial effect of global warming on ENSO would therefore likely have an impact on the regional climatology of tropical cyclone formation.

A coupled GCM can also potentially simulate aspects of the observed variability on longer, decadal timescales if the simulation is lengthy enough. Decadal variability has been documented in the observed record and simulated in GCMs (e.g., Knutson and Manabe 1998). Associations between decadal timescale variations and tropical cyclone formation have also been documented (Gray et al. 1997; Walsh and Kleeman 1997; Goldenberg and Landsea 1999). Nevertheless, decadal variations of TCLVs simulated by a climate model have yet to be examined.

These variations are examined in the current study, along with the response of TCLV formation and movement to model-generated ENSO-like oscillations, as well as to the secular transient greenhouse forcing. Section 2 describes the modelling system and the methodology for detection of TCLVs. Section 3 gives the results of the simulations, while section 4 summarizes the conclusions of the study.

2. Model and methodology

a. Model and domain

The numerical study is carried out using the three-dimensional Commonwealth Scientific and Industrial Research Organisation (CSIRO) Division of Atmospheric Research Limited Area Model (DARLAM) of McGregor and Katzfey (1998). DARLAM is a two-time-level, semi-implicit, hydrostatic primitive equation model on an Arakawa staggered C-grid, employing a Lambert conformal projection. A combination of semi-Lagrangian and bicubic spatial interpolation is used for the horizontal advection. In the vertical, the turbulent mixing is parameterized in terms of stability-dependent K theory; the shallow cumulus convection schemes of Geleyn (1987) is also used. Here, a model domain of horizontal grid size 108 × 57 is used with a constant horizontal resolution of 125 km (Fig. 1), with nine levels in the vertical. This resolution is low compared to the structure of real tropical cyclones and so the model is unlikely to simulate all aspects of tropical cyclone structure. However, previous studies have shown that the vortices generated at this resolution are able to give a reasonable representation of the climatology of tropical cyclone formation (Walsh and Watterson 1997). To capture the formation and subsequent movement of TCLVs caused by the different phases of ENSO, the model domain extends well to the east of dateline into the central Pacific, and therefore spans approximately latitudes 55°S–10°N and longitudes 70°E–120°W. The simulation is forced by output from the CSIRO Mark2 atmosphere–ocean coupled global model (Gordon and O'Farrell 1997) in which the equivalent CO2 concentrations are gradually increased. The equivalent CO2 concentrations are computed as in Kattenberg et al. (1996):

 
formula

where Co is the initial concentration (330 ppm) and ΔQ is a specified radiative forcing change, here taken following the IPCC/IS92a emission scenario of Kattenberg et al. (1996) [see Hirst (1999) for details].

Fig. 1.

DARLAM domain. Crosses are DARLAM grid points

Fig. 1.

DARLAM domain. Crosses are DARLAM grid points

The current study analyses simulated results for two time periods: the current climate (represented by the first 30 yr from year 1961 to year 1990) and 3 × CO2 conditions (the 30-yr period from year 2061 to year 2090).

The current study focuses on the portion of the domain that is most sensitive to ENSO variations, the southwest Pacific region. Therefore, there is no discussion of the formation and tracks of TCLVs in the part of the domain west of the Australian continent.

b. Detection and tracking methodology

The method for the detection and tracking of the TCLVs is the same as that documented in Walsh and Watterson (1997) and Walsh and Katzfey (2000). Briefly described, this method relies upon detection criteria that are based upon observed tropical cyclone characteristics. These criteria must be satisfied for a low pressure area generated by the model to be identified as a TCLV, and are summarized as follows:

  1. minimum vorticity must be 10−5 s−1;

  2. there must be a closed pressure minimum within a radius of 250 km from a point satisfying criterion 1, a distance chosen empirically to give a good geographical association between vorticity maxima and pressure minima; this minimum pressure is taken as the center of the storm;

  3. the total tropospheric temperature anomaly, calculated by summing temperature anomalies at 700, 500, and 300 hPa around the center of the storm (anomalies from the mean environmental temperature at each level in a band on either side of the storm), must be greater than zero;

  4. mean wind speed in a region 500 km × 500 km square around the center of the storm at 850 hPa must be higher than at 300 hPa;

  5. the temperature anomaly at 300 hPa must be greater than at 850 hPa at the center of the storm;

  6. the outer core wind strength (OCS) (Weatherford and Gray 1988), defined as the mean tangential wind speed between a radius of 1° and 2.5° of latitude from the center of the storm, must be greater than 5 m s−1.

A modification was made to the method of Walsh and Watterson (1997) in that after a TCLV has satisfied all of the above criteria for at least 24 h, criteria 3, 4, and 5 were relaxed. The subsequent evolution of the storm was followed until criterion 6 alone was no longer satisfied, in other words until the storm winds had weakened. Observed tropical cyclones are tracked in a similar manner, and whether the storm is warm or cold cored is not usually used as a definition of whether the storm track has ended (J. Kepert 1998, personal communication). This is partly because there is no objective definition for transition of a tropical cyclone to an extratropical system (Malmquist 1999). Comparison between storm tracks with and without the relaxation of criteria 3–5 showed that the effect of the relaxation was to extend the poleward track of existing storms rather than to create new detections. This gave a more realistic simulation of poleward tracks.

3. Results

a. Model simulation of ENSO-related variations in the current climate

1) Mean sea level pressure, rainfall, and winds

Interannual climatic variability of the tropical Pacific is dominated by the ENSO phenomenon (e.g., Allan et al. 1996); its extremes are referred to as El Niño and La Niña states. A commonly used indicator of ENSO phases is the Southern Oscillation index [(SOI, the normalized differences of mean sea level pressure (MSLP), Tahiti minus Darwin; Troup (1965)]. A positive SOI above a specified threshold is defined as La Niña conditions and negative SOI below a specified threshold defined as El Niño conditions. A crucial aspect of the model's performance is its simulation of the ENSO phenomenon.

Unless otherwise stated, all subsequent analysis of the interannual variability carried out in the current study is for the cyclone season in the Southern Hemisphere, November to April (Revell and Goulter 1986).

The correlation map between the 17-yr (1979–95) observed rainfall anomaly (Xie and Arkin 1997) and observed SOI is shown in Fig. 2a. The results for the 30-yr simulated current climate in DARLAM (1961–90) are shown in Fig. 2b. Both figures show positive correlations in the western southwest Pacific and over northern Australia, and negative correlations in the central equatorial Pacific. Similar correlation patterns are seen between MSLP and the SOI (not shown). The simulated intensity of correlations is weaker than that of the observations. In general, this is probably because the forcing GCM is incapable of producing a sharply defined thermocline in the eastern Pacific, consequently reducing the amplitude of the simulated ENSO (Wilson 2000).

Fig. 2.

Correlation map of rainfall anomaly and SOI under current climate conditions, for (a) observations and (b) DARLAM. Dark shaded areas are regions of significant positive correlation, while light shaded areas are regions of significant negative correlation

Fig. 2.

Correlation map of rainfall anomaly and SOI under current climate conditions, for (a) observations and (b) DARLAM. Dark shaded areas are regions of significant positive correlation, while light shaded areas are regions of significant negative correlation

In order to construct composites of typical simulated El Niño and La Niña conditions, the following definition is used. La Niña conditions are defined as model SOI > 2, whereas El Niño is defined as model SOI < −2. Additionally, it was found that the sign of the anomaly of east–west equatorial SST gradient needed to be opposed to the sign of the SOI for rainfall patterns to be clearly defined and unambiguous. The SST gradient anomaly is represented by the normalized difference of area-averaged SST between a region bounded by 5°S–5°N, 140°–170°E and the Niño-4 region (5°S–5°N, 160°E–150°W). These regions overlap slightly but give a good indication of the east–west gradient across the portion of the Pacific Ocean that is included in the DARLAM domain. The magnitude of this SST gradient anomaly was chosen to be greater than 0.2°C.

The interrelationships among variables are illustrated in Fig. 3, which shows the ENSO indices for the 30-yr current climate. The duration of the simulated ENSO phases is between 1–5 yr. Based upon a canonical definition of ENSO, we expect the SOI to be anticorrelated with the SST gradient and Niño-3 (5°S–5°N, 90°–150°W) indices. This is the case when there is a strong SOI signal (|SOI| > 2), except for the year 1990, when the SOI, Niño-3 and the SST gradient anomaly are all positive. This year was therefore excluded from the composites. For the 30-yr period representing current climate, the model-defined El Niño and La Niña states comprise six model years each. The anticorrelation between the SOI and the SST gradient is not as pronounced under 3 × CO2 conditions (not shown here), so fewer years were selected for the enhanced greenhouse composites: four for each phase of ENSO.

Fig. 3.

Simulated interannual variability of SOI, Niño-3 SST anomaly (°C), and east–west Pacific SST gradient anomaly (°C) for 30-yr simulation under the current climate

Fig. 3.

Simulated interannual variability of SOI, Niño-3 SST anomaly (°C), and east–west Pacific SST gradient anomaly (°C) for 30-yr simulation under the current climate

The main observed features of El Niño (e.g., Allan 1996; Harrison and Larkin 1998) are reproduced by the model, as shown schematically in Fig. 4 for the 6-yr El Niño composite. There are westerly wind anomalies along the equator, as well as high anomalous mean sea level pressure (light-shaded region) over part of Australia and the western part of the South Pacific, south to New Zealand. There is a low pressure anomaly (dark-shaded region) on the eastern model boundary at approximately 25°S and 130°W. Similar features are seen in observed conditions (Fig. 1 in Harrison and Larkin 1998). The main difference is that the magnitude of the simulated SST anomalies is only about one-third of that observed (e.g., Allan 1996).

Fig. 4.

Plot of composite simulated El Niño showing three characteristic features: eastern and central equatorial Pacific warm SST anomalies (contoured); equatorial westerly wind anomalies (vectors); and MSLP anomalies from the 6-yr composite of current climate conditions (shaded; light is positive anomaly > 0.5 hPa, and dark is negative < −0.8 hPa). Contour interval for SST is 0.2°C

Fig. 4.

Plot of composite simulated El Niño showing three characteristic features: eastern and central equatorial Pacific warm SST anomalies (contoured); equatorial westerly wind anomalies (vectors); and MSLP anomalies from the 6-yr composite of current climate conditions (shaded; light is positive anomaly > 0.5 hPa, and dark is negative < −0.8 hPa). Contour interval for SST is 0.2°C

Rainfall anomaly patterns from the 6-yr composites give qualitative agreement with observations (Allan 1996) as shown in Fig. 5. Under El Niño conditions (Fig. 5a), it is dry over the northern part of Australia, New Guinea, and Indonesia. In the La Niña state (Fig. 5b), the rainfall maximum shifts westward across the Pacific toward Australia and Indonesia. In the eastern and northern parts of Australia, regions that have observed substantial correlations with the SOI (e.g., McBride and Nicholls 1983), there is a simulated increase in rainfall in La Niña conditions as compared with El Niño, also in agreement with observations. Simulated wind anomaly patterns under El Niño and La Niña conditions are qualitatively similar to those observed (not shown).

Fig. 5.

DARLAM simulation of rainfall anomaly (mm day−1) of 6-yr composite under current climate conditions for (a) El Niño and (b) La Niña

Fig. 5.

DARLAM simulation of rainfall anomaly (mm day−1) of 6-yr composite under current climate conditions for (a) El Niño and (b) La Niña

In summary, the regional climate model DARLAM, as forced by the GCM, gives a simulation that reproduces in many respects the important aspects of the observed interannual variability in this region of the globe. The simulated anomaly patterns associated with the extremes of ENSO are similar to those from observations. The main deficiency in the DARLAM simulations is that the amplitude of the simulated ENSO variations is smaller than in reality because of the smaller simulated SST variations in the forcing GCM.

2) TCLV formation and occurrence

Observed tropical cyclone numbers and simulated formation and occurrence of TCLVs are summarized in Tables 1 and 2. These tables show that in this simulation, given the detection criteria summarized in section 2b, too many TCLVs are formed in comparison with reality. This is probably caused by two main factors. There is too much simulated precipitation in the current simulation, which produces too many TCLV precursor systems (not shown). In addition, the simulated vertical wind shear is too weak in comparison with observations (not shown). Vertical wind shear is here defined as

 
formula

where u is the zonal wind, υ is the meridional wind, and the subscripts refer to pressure levels in hectoPascals. High values of Sz tend to inhibit tropical cyclone formation (e.g., Gray 1979; McBride 1995; Goldenberg and Shapiro 1996), whereas in this model run, values of wind shear are generally too low as compared with the National Centers for Environmental Prediction reanalysis fields (Kalnay et al. 1996) over most of the formation region. This highlights the importance of a good simulation of the mean climate of the region in order to obtain a good simulation of derived variables such as TCLV formation (e.g., Vitart et al. 1999).

Table 1.

Observed formation of tropical cyclones as compared with simulated numbers in the current climate, Nov–Apr, 145°E–140°W. Observed data taken from the Joint Typhoon Warning Center (JTWC) Best Track Data, 1967–96. Percentages indicate the fraction of total cyclone numbers occurring in each band

Observed formation of tropical cyclones as compared with simulated numbers in the current climate, Nov–Apr, 145°E–140°W. Observed data taken from the Joint Typhoon Warning Center (JTWC) Best Track Data, 1967–96. Percentages indicate the fraction of total cyclone numbers occurring in each band
Observed formation of tropical cyclones as compared with simulated numbers in the current climate, Nov–Apr, 145°E–140°W. Observed data taken from the Joint Typhoon Warning Center (JTWC) Best Track Data, 1967–96. Percentages indicate the fraction of total cyclone numbers occurring in each band

Also given in Tables 1 and 2 are the percentages of storms by latitude band, which enables a comparison between observations and the simulation of their latitudinal distribution. The latitudinal distribution of simulated formation (Table 1) and occurrence (Table 2) is reasonable compared with that observed, although formation at higher latitudes is less likely in the model than observed.

Table 2.

The same as Table 1 but for occurence, in cyclone days

The same as Table 1 but for occurence, in cyclone days
The same as Table 1 but for occurence, in cyclone days

Observed and simulated formation are shown in Fig. 6, which illustrates the overestimation of formation, particularly in the eastern regions of the model domain. Some formation is also simulated in the Northern Hemisphere, but this is not shaded since the proximity to the boundary of the model makes the numbers quite small, no more than one formation per 5° grid square.

Fig. 6.

Comparison between (a) observed tropical cyclone formation for the period 1967–96 and (b) simulated TCLV formation for the 30-yr current climate period. Observations from the Joint Typhoon Warning Center (JTWC) Best Track Data Set

Fig. 6.

Comparison between (a) observed tropical cyclone formation for the period 1967–96 and (b) simulated TCLV formation for the 30-yr current climate period. Observations from the Joint Typhoon Warning Center (JTWC) Best Track Data Set

The relationship between the ENSO variations and TCLV occurrence is shown in Fig. 7, for both observations (Fig. 7a) and simulations (Fig. 7b). For the observations, Southern Hemisphere occurrence totals in 2.5° longitude bands are shown for months when the Troup index of the SOI was greater than 10 (La Niña) or less than −10 (El Niño). For the model results, because DARLAM is not implemented on a regular latitude–longitude grid (Fig. 1), meridional totals are calculated for each model x coordinate rather than in a longitude band. For better comparison with the observations in Fig. 7a, meridional totals for the model results shown in Fig. 7b are summed together two x grid points at a time in a band 250 km wide. This gives 54 x-coordinate data points instead of 108 as in the model grid. Under the grid system shown in Fig. 7b, the Indian Ocean extends from x-coordinate 1 to 12, Australia is between gridpoints 12 and 26 and the Pacific Ocean extends from gridpoints 27 onward.

Fig. 7.

Occurrence in cyclone days during El Niño and La Niña conditions of (a) observed tropical cyclones, total south of the equator in 2.5° long bands, and (b) simulated TCLVs in current climate, totals as described in the text. In (b), Australia is located approximately between gridpoints 12 and 26 as indicated. From gridpoint 27 onward is the South Pacific Ocean

Fig. 7.

Occurrence in cyclone days during El Niño and La Niña conditions of (a) observed tropical cyclones, total south of the equator in 2.5° long bands, and (b) simulated TCLVs in current climate, totals as described in the text. In (b), Australia is located approximately between gridpoints 12 and 26 as indicated. From gridpoint 27 onward is the South Pacific Ocean

Figure 7 shows similarities between longitudinal variations in observed and simulated formation with the phase of ENSO. Under both observed and simulated El Niño conditions, cyclone occurrence is higher in the eastern South Pacific, whereas during La Niña, higher occurrence of cyclones is found close to the Australian coast. These shifts are associated with the movement of warm SST anomalies across the Pacific. The observed geographical shift has been previously documented by several authors (e.g., Revell and Goulter 1986; Hastings 1990; Evans and Allan 1992; Basher and Zheng 1995).

TCLV tracks are shown in Fig. 8 for simulated El Niño and La Niña conditions. Here, to more clearly demonstrate the difference in tracks, we have restricted the El Niño and La Niña plots to those years where the magnitude of the SST gradient anomaly across the equatorial Pacific was at least 1.5°C. In this figure only 3 yr of data are included for El Niño conditions, and 2 yr for La Niña. Under El Niño conditions (Fig. 8a), TCLVs form and travel farther eastward, whereas under La Niña conditions (Fig. 8b), the storms form much closer to the northeast (NE) coast of Australia.

Fig. 8.

DARLAM simulation of TCLV tracks under current climate conditions for (a) El Niño and (b) La Niña. Initial positions of TCLVs are indicated by crosses, subsequent positions by dots

Fig. 8.

DARLAM simulation of TCLV tracks under current climate conditions for (a) El Niño and (b) La Niña. Initial positions of TCLVs are indicated by crosses, subsequent positions by dots

b. Variations under enhanced greenhouse conditions

1) Precipitation

Here, simulated 3 × CO2 conditions are analyzed, corresponding to years 2061–90. The variation of rainfall anomaly patterns between El Niño and La Niña states simulated in the current climate (Fig. 5) is still seen in enhanced greenhouse conditions (not shown); however, the magnitude of variations under 3 × CO2 conditions seems higher when compared with that simulated for the current climate: for instance, positive rainfall anomalies off the NE coast of Australia are 2 times as large in 3 × CO2 conditions.

2) TCLVs and ENSO

The effect of ENSO on the spatial distributions of TCLVs is similar under greenhouse conditions to that in the current climate, as shown in Fig. 9. In common with the simulation for the current climate (Fig. 7), higher occurrence of TCLVs is seen close to the Australian east coast under La Niña conditions than under El Niño conditions. There are some minor differences from the longitudinal distribution simulated in the current climate: In Fig. 7, more storms are simulated under La Niña conditions between gridpoints 26 and 38; under 3 × CO2 conditions (Fig. 9) this is reduced to the region between gridpoints 26 and 33. Many GCM simulations have predicted that oscillations associated with ENSO will continue under enhanced greenhouse conditions (e.g., Meehl and Washington 1996; Knutson et al. 1997; Timmerman et al. 1998). If this is so, the results presented here suggest that the geographical variation in tropical cyclone formation and occurrence associated with ENSO in the current climate will also continue in a warmer world. The dependence of cyclone tracks on ENSO in enhanced greenhouse conditions (not shown) is similar to those simulated for the current climate (see Fig. 8). Again, as in the current climate simulations, the movement of TCLVs is influenced by the SST anomalies through their effect on preferred regions of TCLV formation.

Fig. 9.

Distribution of simulated TCLVs under the enhanced greenhouse climate. In the east–west direction, Australia is between gridpoints 12 and 26. From gridpoint 27 onward is the Pacific Ocean

Fig. 9.

Distribution of simulated TCLVs under the enhanced greenhouse climate. In the east–west direction, Australia is between gridpoints 12 and 26. From gridpoint 27 onward is the Pacific Ocean

3) Total number of TCLVs

The number of TCLVs decreases under greenhouse conditions compared with current climate conditions (Fig. 10). During the period of simulation, there is a distinct and statistically significant downward trend in total TCLV numbers east of Australia, as evaluated by Kendall's tau test (e.g., Press et al. 1992). It is difficult to diagnose unambiguously the reason for this result. This is because there is no diagnostic tropical cyclogenesis parameter that is considered to be reliable under enhanced greenhouse conditions, despite continued work on this issue (Gray 1979; Royer et al. 1998). Nevertheless, we have investigated changes in a number of factors that are known to contribute to tropical cyclone formation in the current climate.

  • Moist convection

    Tropical cyclogenesis is intimately associated with the formation of deep convection. Therefore, substantial changes in the amount of deep convection should be related to changes in tropical cyclogenesis, all other things being equal. However, moist convection largely increases rather than decreases over the regions east of Australia under enhanced greenhouse conditions (not shown). Thus, although this variable may have an effect on cyclone formation, is not likely to be a cause of the decrease in cyclogenesis in this simulation.

  • Vertical wind shear

    The difference of this quantity between current climate and enhanced greenhouse conditions is shown in Fig. 11, for the months November–April. Some increases in shear are seen off the coast of eastern Australia and over most of the regions of TCLV formation, except over the extreme northern part of the formation region (compare Fig. 6). This increase in shear may be a contributing factor toward the simulated decrease in TCLV formation, as Shapiro (1987) shows that changes in mean wind shear of 2 m s−1 can strongly affect tropical cyclone formation.

  • Midtropospheric relative humidity

    High values of this quantity have been shown to support cyclone formation (e.g., McBride 1995). Nevertheless, in this simulation the midtropospheric relative humidity is actually higher under enhanced greenhouse conditions (not shown), so this does not explain the decrease in TCLV formation.

  • Low-level vorticity

    Changes in vorticity could be a contributing factor, with more cyclonic vorticity perhaps leading to more storm formation. On average, vorticity is more anticyclonic (positive in the Southern Hemisphere) under enhanced greenhouse conditions in the formation regions east of Australia (not shown), while an area of more cyclonic (negative) vorticity is seen just east of the Solomon Islands. This pattern could tend to suppress cyclogenesis most in the regions closer to Australia and to a lesser extent in the regions further east. This is in agreement with the behavior shown in Fig. 12, where TCLV formation decreases more just east of Australia and less farther east.

  • Sea surface temperatures

    SSTs are higher in the enhanced greenhouse simulation, with temperature rising more in the eastern Pacific than in the west, a more “El Niño–like” pattern. In the observations, there is little relationship between ENSO variations and total numbers of tropical cyclones in the South Pacific region as a whole, although there is a strong relationship between ENSO and numbers of cyclones close to the Australian coast (Hastings 1990). Thus the decline in total numbers over the South Pacific region is not likely to have been caused purely by a change toward a more El Niño–like SST pattern.

Fig. 11.

Difference in vertical wind shear, 3 × CO2 minus current climate for Nov–Apr climatological conditions. Contour interval is 0.5 m s−1

Fig. 11.

Difference in vertical wind shear, 3 × CO2 minus current climate for Nov–Apr climatological conditions. Contour interval is 0.5 m s−1

Fig. 12.

10-yr low-pass filter of (a) observed annual tropical cyclone numbers and (b) simulated TCLV numbers, and associated best-fit trend lines. Solid lines indicate 145°–170°E, while dashed lines indicate 170°E–130°W

Fig. 12.

10-yr low-pass filter of (a) observed annual tropical cyclone numbers and (b) simulated TCLV numbers, and associated best-fit trend lines. Solid lines indicate 145°–170°E, while dashed lines indicate 170°E–130°W

Fig. 10.

Occurrence (JFM) of observed tropical cyclones and TCLVs by latitude band, east of 145°E and west of 140°W. Observations are from the JTWC Best Track Data, 1967–96

Fig. 10.

Occurrence (JFM) of observed tropical cyclones and TCLVs by latitude band, east of 145°E and west of 140°W. Observations are from the JTWC Best Track Data, 1967–96

4) Simulation of TCLVs by the forcing GCM

The role played by the GCM in forcing the simulated trends in DARLAM was evaluated, as well the ability of the GCM to simulate tropical cyclogenesis. If the same TCLV detection criteria used in the analysis of the DARLAM results are applied to the GCM output, the GCM simulates only about one-tenth of the observed tropical cyclone formation rate (not shown). Despite the overestimation by DARLAM of observed tropical cyclone formation, it is certainly simulates formation more realistically than does the GCM.

Under 3 × CO2 conditions, the number of vortices generated by the GCM is about 0.5 times that simulated by it for the current climate. This trend is similar to that shown for DARLAM, but more dramatic. The similarity of this trend in the two models suggests that the GCM is forcing a substantial portion of the large-scale circulation changes causing the simulated decrease in TCLVs in DARLAM. To examine this question further, changes under enhanced greenhouse conditions simulated by the GCM were examined for the fields discussed above. In the GCM, the simulated changes in vorticity and vertical wind shear are quite similar to those simulated by DARLAM, which suggests that these fields are largely determined by the large-scale forcing supplied by the GCM (not shown). Relative humidity increases in a warmer world in the GCM as it does in DARLAM, although the pattern of relative humidity increase is somewhat different in the GCM. Precipitation changes, though, are different in the GCM from those of DARLAM. Between the east coast of Australia and 170°E, the GCM simulates precipitation decreases, while east of 170°E increases are simulated. DARLAM simulates mostly precipitation increases in the entire tropical South Pacific, however. This may be a cause of the greater percentage decline in TCLV formation in the GCM than in DARLAM, as the increase in moist convection in DARLAM under enhanced greenhouse conditions partially counteracts the more hostile vertical wind shear environment, thereby causing a smaller decline in TCLV numbers than in the GCM.

As mentioned in the introduction, there are conflicting results at present on the issue of whether tropical cyclone numbers will change in a warmer world. In the south Pacific region, recent climate model results have simulated little change in numbers (Tsutsui et al. 1999; Walsh and Katzfey 2000) or slight decreases (Yoshimura et al. 1999). The model results presented here show a decrease in tropical cyclone numbers in this region. This was ascribed partly to increases in the vertical wind shear under enhanced greenhouse conditions. However, other climate models produce different changes in shear. For example, the Australian Bureau of Meteorology Research Centre GCM (Power et al. 1998) simulates much smaller increases in vertical wind shear under enhanced greenhouse conditions (not shown). Because of the differences in model responses and the lack of a reliable diagnostic cyclogenesis parameter, the actual response of total tropical cyclone numbers in this region to the enhanced greenhouse effect remains substantially uncertain at this time.

5) Poleward extent of TCLVs

It has been suggested that the general increase in SSTs predicted as a result of climate change may cause tropical cyclones to persist farther poleward than in the current climate (Henderson-Sellers et al. 1998). The recent study of Walsh and Katzfey (2000) found some support for this hypothesis, although they also concluded that considerably more work needed to be performed to make this conclusion more robust. The simulations of Walsh and Katzfey were performed using the same regional climate model (DARLAM) but forced by the output of a “slab” ocean model coupled to an atmospheric GCM under 2 × CO2 conditions (Watterson et al. 1997).

The poleward extent of TCLVs under greenhouse conditions can also be examined quantitatively as shown in Fig. 10, which compares TCLV occurrence in 1× and 3 × CO2 conditions, by latitude band. As mentioned earlier, too many cyclone days are simulated in the model as compared with observations. Despite the fact that the number of TCLVs is less in the enhanced greenhouse simulation, there is a small increase in the occurrence of TCLVs from 30° to 35°S and farther south. The average latitude of occurrence for the current climate simulation is 17.7°S; under 3 × CO2 conditions it is 18.6°S. This change is not large but is statistically significant as assessed by the Mann–Whitney U test (e.g., von Storch and Zwiers 1999). This result is also consistent with that of Walsh and Katzfey (2000) and indicates the continued need for research on this issue. However, other simulations show no such poleward extension of tracks (e.g., Knutson and Tuleya 1999).

c. Decadal variations of TCLV formation

The issue of decadal variability, both simulated and observed, is one that is drawing increasing attention (e.g., Latif 1998; Knutson and Manabe 1998). The non-ENSO variability in observed SSTs has been recently described by Enfield and Mestas-Nuñez (1999). In the Pacific, the leading non-ENSO EOF mode (their Fig. 3) is characterized by negative anomalies in the central and western north Pacific centered near 35°N, with positive anomalies along the western shores of north and south America as well over the ocean along the equator, and to north and south of it. This pattern closely resembles that previously referred to as the Pacific decadal oscillation (PDO; Mantua et al. 1997). Many suggestions have been made regarding possible mechanisms for the generation of the PDO, but the issue remains unresolved.

In the context of climate change, decadal variability is important because variations on these timescales can act to mimic or obscure climate change signals, particularly in the early part of the expected increase in the effects of global warming. Additionally, decadal variability is important because of the possibility that it modulates and alters the relationships between El Niño and a number of other phenomena (Power et al. 1999), including possibly the formation of tropical cyclones (Grant and Walsh 2000).

In the current experiment, the forcing GCM simulates decadal variability that resembles observational estimates in the regions over which DARLAM is implemented (Walland et al. 2000). The simulated decadal variability accounts for some 20%–40% of the variability of precipitation on timescales greater than one year. The length of the DARLAM simulation performed here (130 yr) is long enough for decadal oscillations in tropical cyclone occurrence to be extracted. Unfortunately, the same is not true in general of the observed tropical cyclone record in this region. Truly reliable data over the oceans of this region date only from the late 1960s, after the introduction of routine satellite reconnaissance. Over land, the data may conceivably be more reliable for a longer period of time, but the completeness and correctness of these data have yet to be evaluated for this purpose.

The observed and simulated decadal variation of tropical cyclogenesis in the model is shown in Fig. 12, where a 10-yr low-pass filter has been applied to the annual data. The solid line shows formation from 145° to 170°E, while the dotted line shows formation from 170°E to 130°W. A longitude of 170°E was chosen as the dividing line for the purpose of this analysis because observed tropical cyclone formation differs on either side of this line in terms of its correlation with the SOI (Basher and Zheng 1995). There are substantial observed decadal variations in formation during the limited record available for analysis (Fig. 12a). Close to the eastern Australian coast (solid line), numbers reached a peak in the mid-1970s and have trended downward since, while storms east of 170°E (dashed line) have slightly increased in number. In comparing the observations to the simulated decadal variability (Fig. 12b), numbers simulated between 145° and 170°E are roughly comparable to observed, but simulated numbers are considerably greater than observed east of 170°E. The two time series shown in Fig. 12b are substantially anticorrelated (r = −0.40 for 1966–2084), similar to the results shown above for the interannual variability. Additionally, the decrease in tropical cyclone formation is greater in the western Pacific than in the eastern, which is consistent with the trend to a more El Niño–like mean state seen in many transient climate model simulations (e.g., Meehl and Washington 1996; Knutson et al. 1997; Timmerman et al. 1998).

Correlations between detrended, 10-yr low-pass filtered simulated TCLV formation and similarly detrended, filtered Pacific SST variations are shown in Fig. 13. For formation in the Southern Hemisphere from 145° to 170°E (Fig. 13a), there are reasonably strong correlations in the model in a V-shaped region running from the southeast Pacific through the region east of Australia and up into the northern Pacific. This correlation pattern is similar to the PDO-related pattern of observed SST decadal variability shown in Enfield and Mestas-Nuñez (1999), although the simulated center of action in the North Pacific is displaced slightly southwest of its observed position. This suggests that a mode of oscillation on decadal timescales is having a substantial effect on DARLAM's formation of TCLVs in this region.

Fig. 13.

Correlations between decadal variations of TCLV formation and SSTs, for TCLVs in the Southern Hemisphere between (a) 145°–170°E and (b) 170°E–130°W. Shaded areas are correlations greater than 0.4

Fig. 13.

Correlations between decadal variations of TCLV formation and SSTs, for TCLVs in the Southern Hemisphere between (a) 145°–170°E and (b) 170°E–130°W. Shaded areas are correlations greater than 0.4

For the area between 170°E and 130°W (Fig. 13b), the relationship with the decadal mode of variation is less clear. The correlations with the El Niño source regions of the central Pacific have the opposite sign to those in Fig. 13a, but are less strong. Substantial correlations now exist between TCLV formation in this region and SST east of New Zealand.

The causes of the association between SST variations and TCLV formation on decadal timescales are clearly analogous to the similar relationship between ENSO-forced SST anomalies and observed tropical cyclone formation. At both timescales, positive SST anomalies in the southwest Pacific are associated with more TCLV formation in this region (compare Figs. 13a and Fig. 7). Observations also show similar behavior at both interannual and decadal timescales, with higher SSTs in this region associated with more tropical cyclone formation (Grant and Walsh 2000). At decadal timescales, the precise mechanism for this behavior is not yet clearly established, although vertical wind shear is typically higher during low SST episodes.

In summary, there is substantial decadal variability in simulated TCLV formation. This is associated with a model-generated mode of Pacific SST decadal variability that has some resemblance to observed variability at this timescale.

4. Conclusions

A regional climate model, DARLAM, has been nested within the CSIRO Mark2 coupled ocean–atmosphere GCM and a transient enhanced greenhouse simulation. The results are analyzed in terms of the ability of the model to simulate tropical cyclone formation in the current climate, the associations between tropical cyclones occurrence and ENSO, and any changes under enhanced greenhouse conditions. The following conclusions were made:

  • Total numbers of simulated tropical cyclone–like vortices (TCLVs) are overestimated in the current climate, probably as a result of excessive simulated precipitation and insufficient simulated vertical wind shear. Under enhanced greenhouse conditions, numbers decrease, but given the differing predictions of other climate models regarding tropical cyclone numbers in a warmer world and changes in vertical wind shear, only low confidence can be placed in this result.

  • The model seems to produce well the spatial distribution of TCLV formation and occurrence in the current climate. A higher incidence of TCLVs is seen near the Australian east coast under La Niña, while under El Niño conditions formation and occurrence shifts more toward the central Pacific. In enhanced greenhouse conditions, a similar spatial variation is simulated.

  • Since climate models predict that ENSO variations of some kind will continue in a warmer world, the above geographical relationship between ENSO and TCLV occurrence may also continue.

  • Tracks of cyclones under enhanced greenhouse conditions are slightly more poleward than simulated in the current climate. This is consistent with results from a previous version of this model, but other models do not give the same result.

  • The model simulation of TCLVs exhibits considerable coherent decadal variability that is in some respects similar to previously identified observed patterns of large-scale variations on this timescale. An improved version of this model may therefore potentially be used to investigate aspects of the relationship between tropical cyclone formation and dynamical variables on these timescales.

Acknowledgments

The first author would like to thank Mr. Barrie Hunt for his encouragement to perform this research. Acknowledgments are made also to Dr. Robert Allan for his permission to use the wind data, to Dr. Ian Smith for supplying Niño-3 SST data, and to Dr. Stephen Wilson for useful discussions. Thanks go to Rob Colman of the Australian Bureau of Meteorology Research Centre for supplying output from their GCM. Last, thanks are due to my supervisors Dr. John McGregor and Dr. Jack Katzfey for giving me time to perform this research. The comments of two reviewers improved the manuscript.

REFERENCES

REFERENCES
Allan
,
R. J.
,
J. A.
Lindesay
, and
D. E.
Parker
,
1996
:
El Niño–Southern Oscillation and Climatic Variability.
CSIRO Publishing, 416 pp
.
Basher
,
R. E.
, and
X.
Zheng
,
1995
:
Tropical cyclones in the southwest Pacific: Spatial patterns and relationships to Southern Oscillation and sea surface temperature.
J. Climate
,
8
,
1249
1260
.
Bengtsson
,
L.
,
M.
Botzet
, and
M.
Esch
,
1996
:
Will greenhouse gas induced warming over the next 50 years lead to higher frequency and greater intensity of hurricanes?
Tellus
,
48A
,
57
73
.
Enfield
,
D. B.
, and
A. M.
Mestas-Nuñez
,
1999
:
Multiscale variabilities in global sea surface temperatures and their relationships with tropospheric climate patterns.
J. Climate
,
12
,
2719
2733
.
Evans
,
J. L.
, and
R. J.
Allan
,
1992
:
El Niño/Southern Oscillation modification to the structure of the monsoon and tropical cyclone activity in the Australasian region.
Int. J. Climatol
,
12
,
611
623
.
Geleyn
,
J. F.
,
1987
:
Use of a modified Richardson number for parameterizing the effect of shallow convection.
Short- and Medium-Range Numerical Weather Prediction, T. Matsuno, Ed., Special Volume, J. Meteor. Soc. Japan, 141–149
.
Goldenberg
,
S. B.
, and
L. J.
Shapiro
,
1996
:
Physical mechanisms for the association of El Niño and West African rainfall with Atlantic major hurricane activity.
J. Climate
,
9
,
1169
1187
.
——, and
Landsea
,
C. W.
,
1999
:
Relationships between decadal-scale fluctuations in vertical shear from NCEP/NCAR reanalyses data and Atlantic basin tropical cyclone activity.
Preprints, 23rd Conf. on Hurricanes and Tropical Meteorology, Dallas, TX, Amer. Meteor. Soc., 1089–1091
.
Gordon
,
H. B.
, and
S. P.
O'Farrell
,
1997
:
Transient climate change in the CSIRO coupled model with dynamic sea ice.
Mon. Wea. Rev
,
125
,
875
907
.
Grant
,
A.
, and
K. J. E.
Walsh
,
2000
:
Interdecadal variability in Queensland tropical cyclone occurrence.
Preprints, 24th Conf. on Hurricanes and Tropical Meteorology, Ft. Lauderdale, FL, Amer. Meteor. Soc., 470–471
.
Gray
,
W. M.
,
1979
:
Hurricanes: Their formation, struture and likely role in the tropical circulation.
Meteorology over the Tropical Oceans, D. B. Shaw., Ed., Roy. Meteor. Soc., 155–218
.
——,
Sheaffer
,
J. D.
, and
C. W.
Landsea
,
1997
:
Climate trends associated with multidecadal variability of Atlantic hurricane activity.
Hurricanes: Climate and Socioeconomic Impacts, H. F. Diaz and R. S. Pulwarty, Eds., Springer, 15–54
.
Haarsma
,
R. J.
,
J. F. B.
Mitchell
, and
C. A.
Senior
,
1993
:
Tropical disturbances in a GCM.
Climate Dyn
,
8
,
247
257
.
Harrison
,
D. E.
, and
N. K.
Larkin
,
1998
:
El Niño–Southern Oscillation sea surface temperature and wind anomalies, 1946–1993.
Rev. Geophys
,
36
,
353
399
.
Hastings
,
P. A.
,
1990
:
Southern Oscillation influences on tropical cyclone activity in the Australian/south-west Pacific region.
Int. J. Climatol
,
10
,
291
298
.
Henderson-Sellers
,
A.
, and
and Coauthors
,
1998
:
Tropical cyclones and global climate change: A post-IPCC assessment.
Bull. Amer. Meteor. Soc
,
79
,
19
38
.
Hirst
,
A. C.
,
1999
:
The Southern Ocean response to global warming in the CSIRO coupled ocean–atmosphere model.
Environ. Model. Software
,
14
,
227
241
.
Kalnay
,
E.
, and
Coauthors
,
1996
:
The NCEP/NCAR 40-Year Reanalysis Project.
Bull. Amer. Meteor. Soc
,
77
,
437
471
.
Kattenberg
,
A.
, and
Coauthors
,
1996
:
Climate models: Projections of future climate.
Climate Change 1995: The Science of Climate Change, J. T. Houghton et al., Eds., Cambridge University Press, 285–357
.
Knutson
,
T. R.
, and
S.
Manabe
,
1998
:
Model assessment of decadal variability and trends in the tropical Pacific Ocean.
J. Climate
,
11
,
2273
2296
.
——, and
Tuleya
,
R. E.
,
1999
:
Increased hurricane intensities with CO2-induced warming as simulated using the GFDL hurricane prediction system.
Climate Dyn
,
15
,
503
519
.
——,
Manabe
,
S.
, and
D.
Gu
,
1997
:
Simulated ENSO in a global coupled ocean–atmosphere model: Multidecadal amplitude modulation and CO2 sensitivity.
J. Climate
,
10
,
138
161
.
——,
Tuleya
,
R. E.
, and
Y.
Kurihara
,
1998
:
Simulated increase of hurricane intensities in a CO2-warmed world.
Science
,
279
,
1018
1020
.
Latif
,
M.
,
1998
:
Dynamics of interdecadal variability in coupled ocean–atmosphere models.
J. Climate
,
11
,
602
624
.
Malmquist
,
D.
,
1999
:
Meteorologists and insurers explore extratropical transition of tropical cyclones.
Eos, Trans. Amer. Geophys. Union
,
80
,
79
80
.
Mantua
,
N.
,
S.
Hare
,
Y.
Zhang
,
J.
Wallace
, and
R.
Francis
,
1997
:
A Pacific interdecadal climate oscillation with impacts on salmon production.
Bull. Amer. Meteor. Soc
,
78
,
1069
1079
.
McBride
,
J. L.
,
1995
:
Tropical cyclone formation.
Global perspectives on tropical cyclones, WMO Tech. Doc. WMO/TD-No. 693, 63–105
.
——, and
Nicholls
,
N.
,
1983
:
Seasonal relationships between Australian rainfall and the Southern Oscillation.
Mon. Wea. Rev
,
111
,
1998
2004
.
McGregor
,
J. L.
, and
J. J.
Katzfey
,
1998
:
Climate modelling for Australian region using DARLAM.
Proc. Int. Workshop on General Monsoon System, Beijing, China, START Regional Center for Temperate East Asia, 13–18
.
Meehl
,
G. A.
, and
W. M.
Washington
,
1996
:
El Niño-like climate change in a model with increased atmospheric CO2 concentrations.
Nature
,
382
,
56
60
.
Power
,
S. B.
,
F.
Tseitkin
,
R.
Colman
, and
A.
Sulaiman
,
1998
:
A coupled GCM for seasonal prediction and climate change research.
BMRC Res. Rep. No. 66, 52 pp. [Available from Bureau of Meteorology, GPO Box 1289K, Melbourne, Victoria 3001, Australia.]
.
——
,
T.
Casey
,
C.
Folland
,
A.
Colman
, and
V.
Mehta
,
1999
:
Interdecadal modulation of the impact of ENSO on Australia.
Climate Dyn
,
15
,
319
324
.
Press
,
W. H.
,
S. A.
Teukolsky
,
W. T.
Vetterling
, and
B. P.
Flannery
,
1992
:
Numerical Recipes. 2d ed.
Cambridge University Press, 963 pp
.
Reason
,
C. J. C.
,
R. J.
Allan
,
J. A.
Lindesay
, and
T. J.
Ansell
,
1999
:
ENSO and climatic signals across the Indian Ocean Basin in the global context: Part 1, Interannual composite patterns.
Int. J. Climatol
,
20
,
1285
1327
.
Revell
,
C. G.
, and
S. W.
Goulter
,
1986
:
South Pacific tropical cyclones and the Southern Oscillation.
Int. J. Climatol
,
6
,
1138
1145
.
Royer
,
J-F.
,
F.
Chauvin
,
B.
Timbal
,
P.
Araspin
, and
D.
Grimal
,
1998
:
A GCM study of the impact of greenhouse gas increase on the frequency of occurrence of tropical cyclones.
Climatic Change
,
38
,
307
343
.
Shapiro
,
L. J.
,
1987
:
Month-to-month variability of Atlantic tropical circulation and its relationship to tropical cyclone formation.
Mon. Wea. Rev
,
115
,
2598
2614
.
Timmerman
,
A.
,
J.
Oberhuber
,
A.
Bacher
,
M.
Esch
,
M.
Latif
, and
E.
Roeckner
,
1998
:
ENSO response to global warming.
Max-Planck-Institut für Meteorologie Rep. 251, 13 pp
.
Tonkin
,
H.
,
C.
Landsea
,
G. J.
Holland
, and
S.
Li
,
1997
:
Tropical cyclones and climate change: A preliminary assessment.
Assessing Climate Change: Results from the Model Evaluation Consortium for Climate Assessment. W. Howe and A. Henderson-Sellers, Eds., Gordon and Breach, 327–360
.
Troup
,
A. J.
,
1965
:
The Southern Oscillation.
Quart. J. Roy. Meteorol. Soc
,
91
,
490
506
.
Tsutsui
,
J.
,
A.
Kasahara
, and
H.
Hirakuchi
,
1999
:
The impacts of global warming on tropical cyclones—A numerical experiment with the T42 version of NCAR CCM2.
Preprints, 23rd Conf. on Hurricanes and Tropical Meteorology, Dallas, TX, Amer. Meteor. Soc., 1077–1080
.
Vitart
,
F.
,
J. L.
Anderson
, and
W. F.
Stern
,
1997
:
Simulation of interannual variability of tropical storm frequency in an ensemble of GCM integrations.
J. Climate
,
10
,
745
760
.
——, ——, and ——,
.
1999
:
Impact of large-scale circulation on tropical storm frequency, intensity and location simulated by an ensemble of GCM integrations.
J. Climate
,
12
,
3237
3254
.
von Storch
,
H.
, and
F. W.
Zwiers
,
1999
:
Statistical Analysis in Climate Research.
Cambridge University Press, 484 pp
.
Walland
,
D. J.
,
S. B.
Power
, and
A. C.
Hirst
,
2000
:
Decadal climate variability simulated in a coupled general circulation model.
Climate Dyn
,
16
,
201
211
.
Walsh
,
K. J. E.
, and
R.
Kleeman
,
1997
:
Predicting decadal variations in Atlantic tropical cyclone numbers and Australian rainfall.
Geophys. Res. Lett
,
24
,
3249
3252
.
——
, and
I. G.
Watterson
,
1997
:
Tropical cyclone-like vortices in a limited area model: Comparison with observed climatology.
J. Climate
,
10
,
2240
2259
.
——, and
Katzfey
,
J. J.
,
2000
:
The impact of climate change on the poleward movement of tropical cyclone–like vortices in a regional climate model.
J. Climate
,
13
,
1116
1132
.
——, and
Ryan
,
B. F.
,
2000
:
Tropical cyclone intensity increase near Australia as a result of climate change.
J. Climate
,
13
,
3029
3036
.
Watterson
,
I. G.
,
S. P.
O'Farrell
, and
M. R.
Dix
,
1997
:
Energy and water transport in climates simulated by a general circulation model that includes dynamic sea ice.
J. Geophys. Res
,
102
,
11 027
11 037
.
Weatherford
,
C. L.
, and
W. M.
Gray
,
1988
:
Typhoon structure as revealed by aircraft reconnaisance. Part I: Data analysis and climatology.
Mon. Wea. Rev
,
116
,
1032
1043
.
Xie
,
P.
, and
P. A.
Arkin
,
1997
:
Global precipitation: A 17-year monthly analysis based on gauge observation, satellite estimates, and numerical model outputs.
Bull. Amer. Meteor. Soc
,
78
,
2539
2558
.
Yoshimura
,
J.
,
M.
Sugi
, and
A.
Noda
,
1999
:
Influence of greenhouse warming on tropical cyclone frequency simulated by a high-resolution AGCM.
Preprints, 23rd Conf. on Hurricanes and Tropical Meteorology, Dallas, TX, Amer. Meteor. Soc., 1081–1084
.

Footnotes

Corresponding author address: Kim Nguyen, CSIRO Atmospheric Research, PMB1, Aspendale, Victoria 3195, Australia.Email: kcn@dar.csiro.au