Multi-model Projection of Global Warming Impact on Tropical Cyclone ...

Journal of the Meteorological Society of Japan, Vol. 87, No. 3, pp. 525--538, 2009.

525

DOI:10.2151/jmsj.87.525

Multi-model Projection of Global Warming Impact on Tropical Cyclone Genesis Frequency over the Western North Pacific

Satoru YOKOI Center for Climate System Research, The University of Tokyo, Kashiwa, Japan

and Yukari N. TAKAYABU Center for Climate System Research, The University of Tokyo, Chiba, Japan Research Institute for Global Change, Japan Agency for Marine-Earth Science and Technology, Yokosuka, Japan (Manuscript received 30 September 2008, in final form 19 January 2009)

Abstract This study examines the global warming impact on tropical cyclone (TC) genesis frequency over the western North Pacific basin (0 –40 N, 100 E–180 ; WNP) projected by five atmosphere–ocean coupled general circulation models that participate in the World Climate Research Programme’s Coupled Model Intercomparison Project phase 3 (CMIP3), and exhibit high performances in simulating horizontal distribution of annual-mean frequency under the current climate condition. TC-like disturbances are detected and counted in simulations for the 20th-century climate experiment and global warming experiments. It is revealed that all of the five models project an increasing trend of the frequency in the eastern part of the analysis domain, especially over the central North Pacific (5 –20 N, 150 E–180 ; CNP), and a decreasing trend in the western part, with a maximum decrease over the South China Sea (10 –25 N, 110 –120 E; SCS). The former increasing trend can be interpreted by analogy with interannual variability related to El Ninõ and Southern Oscillation (ENSO). This is because projected changes of sea surface temperature and large-scale circulation field exhibit an El Ninõ-like pattern, and on the other hand, more TCs are observed in the CNP during the El Ninõ phases. Relative vorticity in the lower troposphere and vertical wind shear would become more favorable for TC genesis, as in El Ninõ situation. The authors conclude that these two dynamic factors are major contributors to the projected increase of the frequency in the CNP. Over the SCS, projected environmental conditions are diagnosed as more favorable for TC genesis than the current ones, in spite of the decrease projection of the frequency. The authors discuss that the projected decrease may be associated with a projected weakening of the activity of tropical depression-type disturbance that can later be developed into TC.

1. Introduction The western North Pacific basin (WNP) is the most active region of tropical cyclone (TC) formaCorresponding author: Satoru Yokoi, Center for Climate System Research, University of Tokyo, 5-1-5 Kashiwanoha, Kashiwa, Chiba 277-8568, Japan. E-mail: [email protected] 6 2009, Meteorological Society of Japan

tion, where about a third of TCs that form worldwide are generated. Since the TCs are one of the atmospheric phenomena that cause serious damage to the society in and around the WNP, the question of how TC statistics, for example, frequency and intensity, would change due to the projected global warming is of social importance as well as of scientific interest. Based on analyses of observation datasets, Emanuel (2005) and Webster et al. (2005)

526

Journal of the Meteorological Society of Japan Table 1.

Model name

Vol. 87, No. 3

List of CGCMs used in this study.

Abbreviation

Originating groups, country

Resolution

CGCM3.1(T63)

CCCMA

T63*

CSIRO–Mk3.0

CSIRO3.0

CSIRO–Mk3.5

CSIRO3.5

INGV–SXG ECHAM5/MPI–OM

INGV MPI

Canadian Centre for Climate Modeling and Analysis, Canada Commonwealth Scientific and Industrial Research Organisation Atmospheric Research, Australia Commonwealth Scientific and Industrial Research Organisation Atmospheric Research, Australia Instituto Nazionale di Geofisica e Vulcanologia, Italy Max Planck Institute for Meteorology, Germany

T63 T63 T106 T63

* T63 with ‘‘linear’’ reduced Gaussian grid equivalent to T42 quadratic grid.

showed that TCs had become more intense in this basin. On the other hand, some studies (Chan and Liu 2004; Chan 2006) emphasized that it was di‰cult to discuss anthropogenically induced long-term trends, since they were masked by strong natural variability with interannual to interdecadal time scales. As for the frequency or number of TCs, it is further di‰cult to discuss the trends because of discontinuities inherent in the dataset (Trenberth et al. 2007). Numerical simulation is a powerful tool to project global warming impacts on the TC statistics. One of the major approaches is called direct approach, which detects TC-like vortex disturbances represented by models and takes their statistics. Using this approach, several of recent studies projected a decreasing trend in global-mean TC frequency (Sugi et al. 2002; McDonald et al. 2005; Oouchi et al. 2006; Bengtsson et al. 2007; Gualdi et al. 2008). On the other hand, there seems no clear consensus on regional impacts of global warming on the TC frequency. Over the WNP, some studies (e.g., Sugi et al. 2002; Oouchi et al. 2006; Bengtsson et al. 2007) projected a decreasing trend of TCs, while others (e.g., Stowasser et al. 2007; Emanuel et al. 2008) projected an increasing trend. McDonald et al. (2005) showed that the TC genesis would decrease in the western part of the WNP, while it would increase in the eastern part. Considering large discrepancies in projected frequency change over the WNP among di¤erent climate models, it seems important to examine the global warming impact on a multi-model basis to find what aspects of projected changes are consistent among the models. In order to take the multimodel approach, the present study analyzes outputs from global warming experiments by atmosphere– ocean coupled general circulation models (CGCMs) that participate in the World Climate

Research Programme’s (WCRP’s) Coupled Model Intercomparison Project phase 3 (CMIP3) (Meehl et al. 2007a). Yokoi et al. (2009) evaluated performances of eight of the CMIP3 models in simulating annual-mean TC frequency. They showed that five models, which are listed in Table 1, realistically simulated horizontal distribution of the frequency over the WNP, nevertheless horizontal resolutions of these models are only T63 (@200 km) or T106 (@120 km). In this study, we analyze outputs of these five models to perform the multi-model based projection. The target domain is (0 –40 N, 100 E–180 ), which is displayed in Fig. 1. We will then discuss possible reasons for the projected trends in the TC frequency. One of the e¤ective measures for the discussion is so-called indirect approach, which diagnoses environmental condi-

Fig. 1. Domain where TC detection is performed. Rectangles indicate locations of the South China Sea (10 –25 N, 110 – 120 E; SCS), westernmost part of the western North Pacific (10 –30 N, 120 –135 E; wWNP), and the central North Pacific (5 –20 N, 150 E–180 ; CNP), where projected global warming impacts are discussed in detail.

June 2009

S. YOKOI and Y. N. TAKAYABU

tions that are considered to influence the TC genesis. Such conditions include lower-tropospheric circulation field, vertical wind shear, atmospheric temperature and humidity profile, and sea surface temperature (SST). Camargo et al. (2007) demonstrated that interannual variability (IAV) of TC genesis over the WNP related with El Ninõ and Southern Oscillation (ENSO) can be interpreted from the viewpoint of that of the environmental conditions. Their results imply availability of this approach to discuss the variability of TC genesis. As for the global warming issue, Vecchi and Soden (2007a) projected that vertical shear over the North Atlantic and East Pacific would strengthen and lead to decrease of TC genesis. Over the WNP, Caron and Jones (2008) showed that the conditions would become more favorable for TC genesis. In addition to the environmental conditions, we will also focus on activity of the tropical depression (TD)-type disturbances that can later be developed into TCs (e.g., Heta 1990; Liebmann et al. 1994). This is because it seems di‰cult to form a TC without such disturbances, even if the environmental conditions are satisfied. Most of the CMIP3 models projected that SST would rise all over the ocean, and the horizontal pattern of the warming rate would become an El Ninõ-like one (Meehl et al. 2007b). Since the ENSO is considered to play a dominant role in modulating the TC activity over the WNP (Chan 1985, 2007; Lander 1994; Wang and Chan 2002; Chia and Ropelewski 2002 and others), it is anticipated that horizontal distribution of the TC genesis frequency may also be projected to become an El Ninõ-like. In fact, McDonald et al. (2005) attributed the east–west contrast in the frequency trend projected by their own model to such an SST warming pattern. It should be noted, however, that their model projected SST increase trends over the

527

central and eastern North Pacific that are at a higher rate than other models (e.g., their Fig. 10a), and thus their model possibly overemphasized the east–west contrast. Anyway, we will examine whether such an El Ninõ-like change is projected in the five models studied here, and what aspects of projected changes can and what aspects cannot be interpreted by analogy with ENSO. The rest of the paper is organized as follows. Section 2 explains model simulations and observation datasets analyzed in this paper. Results of the direct approach are shown in Section 3. Then we discuss mechanisms of projected TC frequency trends in terms of changes in environmental conditions and activity of the TD-type disturbance in Sections 4 and 5, respectively. Finally, Section 6 presents summary and discussion. 2. CGCM simulations and observation datasets Among more than ten series of experiments carried out by CMIP3 (Meehl et al. 2007a), we analyze simulations of six experiments: 20th-century climate simulation (20C3M), 21st-century climate change experiments with Special Report on Emission Scenarios (SRES; Nakicenovic et al. 2001) A1B, A2, and B1, and climate change commitment experiments with SRES A1B and B1. Daily-mean three-dimensional atmospheric outputs are archived at the Program for Climate Model Diagnosis and Intercomparison (PCMDI) database, and data lengths are 40 years for 20C3M (corresponding to the years 1961–2000), 40 years for the 21st– century experiment with SRES A2 (2046–2065 and 2081–2100), and 60 years for the 21st–century and commitment experiments with SRES A1B and B1 (2046–2065, 2081–2100, and 2181–2200). Table 2 summarizes the periods and experiments of each CGCM available for analysis. While all of the five CGCMs have the 20C3M output archived at

Table 2. Periods and experiments used for analysis in this paper. Output periods for 20C3M, 21C-m, 21C-e, and 22C-e are 1961–2000, 2046–2065, 2081–2100, and 2181–2200, respectively. Projected warming rate of annual- and global-mean surface temperature averaged over all periods of the SRES experiments from 20C3M is also shown. SRESA1B Model CCCMA CSIRO3.0 CSIRO3.5 INGV MPI

SRESA2

SRESB1

20C3M

21C-m

21C-e

22C-e

21C-m

21C-e

21C-m

21C-e

22C-e

Warming rate [K]

X X X X X

X X X X X

X X X X X

X X — — —

— X X X —

— X X X X

X X X — X

X X X — X

X X X — —

3.2 1.9 2.9 2.9 2.9

528

Journal of the Meteorological Society of Japan

PCMDI, some periods of the global warming experiments are not available. Table 2 also shows a projected warming rate of global- and annualmean surface temperature averaged over the available periods of global warming experiments of each model compared with that simulated in respective 20C3M. In order to mitigate influences of the inter-model di¤erences in climate sensitivity, we divide projected trends with this warming rate prior to making composite. We detect TC-like vortex disturbances from daily-mean atmospheric outputs with the following five conditions which are the same as in Yokoi et al. (2009). (1) The candidate for the disturbance center is a grid point at which the 850-hPa daily-mean relative vorticity attains a local maximum with value greater than a chosen threshold. (2) The daily-mean temperature at the 300-hPa level at the center is higher than the environmental temperature by at least a chosen threshold. (3) Conditions 1 and 2 are satisfied for at least two consecutive time steps of the dataset, i.e., two days. (4) The genesis position, which is defined as the first position where both conditions 1 and 2 are satisfied, is over the ocean. (5) At the genesis instant, the horizontal wind speed at the 850-hPa level around the center is larger than that at the 300-hPa level. The two threshold values are determined for each model independently, so that horizontal distribution of annual-mean TC genesis frequency simulated in 20C3M represents the observed distribution as adequately as possible. The reader is referred to Yokoi et al. (2009) for further information about the TC definition. In order to check performances of the models and to understand the projected global warming impacts by comparing with the IAV in the current climate, we use the 40-yr reanalysis data produced by the European Centre for Medium-Range Weather Forecasts (ERA-40; Uppala et al. 2005), optimum interpolated SST data prepared by the National Oceanic and Atmospheric Administration (Reynolds et al. 2002), and TC best-track data issued by the Japan Meteorological Agency. The best-track data include position and maximum 10-min-mean surface wind speeds of TCs found in (0 –60 N, 100 E–180 ) domain. The position of

Vol. 87, No. 3

TC genesis is defined as the first position where the maximum wind speed exceeded 34 knots (17.5 m s1 ) for each TC. The period used for analysis is 20 years from 1982 to 2001, when all of the three datasets are available. 3. Genesis frequency of TC-like vortex disturbances In this section, we perform the direct approach to project TC genesis frequency under the global warming condition. Before doing this, it is necessary to check performances of the models in simulating the frequency in the current climate. Figures 2a and 2b show observed annual-mean frequency and corresponding multi-model ensemble in 20C3M, respectively. There are a number of similarities between the observation and simulation; both exhibit the maximum found to the east of the Philippines, the secondary maximum over the South China Sea, and lower frequencies to the east of 150 E. Quantitatively speaking, however, the models slightly underestimate the maximum values, and overestimate the frequency to the east of 150 E. Average lifetime of simulated TCs is 4.8 days, which is somewhat shorter but in good accordance with observed lifetime of 5.3 days. Yokoi et al. (2009) showed that the horizontal distribution of the frequency was reasonably reproduced in each of the five models as well. In addition, they reported that these models also realistically simulated seasonal contrast of the frequency, although detailed seasonal march during the TC season (May– October) was not well reproduced. Multi-model ensemble of projected global warming impact on the genesis frequency is shown in Fig. 2c. Its horizontal distribution exhibits clear east–west contrast. There are two regions where all of the five models project the frequency trends with the same sign, shown by gray tones. The frequency would increase in the central North Pacific (5 – 20 N, 150 E–180 ; CNP) and decrease in the South China Sea (10 –25 N, 110 –120 E; SCS) and the westernmost part of the WNP (10 –30 N, 120 – 135 E; wWNP). These regions are indicated in Fig. 1. In both regions, average rates of trend are about 4 [% K1 ]. Such east–west contrast leads to the eastward extension of the frequent TC genesis region. The fact that all models project the same sign of trend suggests robustness of this projected pattern. On the other hand, there is no consensus about the global warming impact on basin-wide and global-mean frequency among the five models.

June 2009


529

Fig. 3. Monthly TC genesis frequency over the CNP (5 –20 N, 150 E–180 ) in units of number per 15 30 latitude–longitude box per 10 years. Thin solid, thick solid, and thick broken lines represent observation, multi-model ensemble of 20C3M, and projection if the mean surface temperature would rise by 2.8 K, respectively.

Fig. 2. Climatological annual TC genesis frequency (a) observed and (b) simulated in 20C3M (five model ensemble), and (c) projected global warming impact on the annual genesis frequency per 1 K warming of the mean surface temperature. Unit is number per 5 5 latitude–longitude box per 10 years. Shadings in (a) and (b) show frequency more than 6 units, and those in (c) indicate that all of the five models project the same sign of trend.

CSIRO3.5, INGV, and MPI project decreasing trends in the TC genesis over the WNP basin and globally as well, while the other two models project increasing trends over both of the domains. Note that Gualdi et al. (2008) and Bengtsson et al. (2007) already reported projections of global-mean frequency with INGV and MPI models, respectively, which were consistent with the present results. Figure 3 displays the annual cycle of the genesis frequency over the CNP. Most of TC genesis is observed during the July–November period (thin solid line). The models fairly reproduce the phase of the annual cycle, while the amplitude is overestimated which is consistent with Figs. 2a,b. The thick broken line shows the projected annual cycle under the condition of the mean surface temperature warming by 2.8 K which is an average value in the end of the 21st century projected by SRES A1B experiments of the CMIP3 models (Meehl et al. 2007b). Overall increasing trends are projected during the July–November period, corresponding to the TC active season. On the other hand, when we concentrate on the SCS and the wWNP, the projected deceasing trends are found in most of the months during May–October, which is also the TC active season there (figure not shown). 4. Environmental conditions In order to discuss reasons for the projected trends in the TC genesis frequency, we examine

530


Vol. 87, No. 3

Fig. 4. Multi-model ensemble of environmental conditions during July–November simulated in 20C3M shown by contour as well as observed climatology by gray tone with a legend at the bottom of each panel. Shown are (a) relative vorticity at the 850-hPa level in units of 107 s1 , (b) vertical shear of horizontal wind between the 850- and 200-hPa levels in units of m s1 (650 hPa)1 , (c) PI in units of m s1 , (d) relative humidity at the 700-hPa level in units of %, and (e) the GP.

global warming impacts on environmental conditions that are considered to influence the TC genesis. Those favorable for the genesis include cyclonic circulation in the lower troposphere, weak vertical wind shear, high SST, low atmospheric static stability, and high relative humidity in the lower and middle troposphere (Gray 1975). Bister and Emanuel (1998, 2002) combined influences of SST and the static stability to define Potential Intensity (PI) that represents the upper limit of maximum cyclonic wind speed of TC in units of m s1 , determined from given SST and atmospheric thermodynamic profile. Higher PI means more favorable condition for TC genesis (Emanuel and Nolan 2004). We will examine relative vorticity at the 850-hPa level ðz850 Þ, vertical wind shear between 850- and 200hPa levels ðVshear Þ, PI, and relative humidity at the 700-hPa level ðH700 Þ. From these four variables, Emanuel and Nolan (2004) defined a TC Genesis Potential (GP) as GP ¼ j10 5 ð f þ z850 Þj 3=2 ð1 þ 0:1 Vshear Þ2 3 PI H700 3 ; 70 50

ð1Þ

where f is the Coriolis parameter. Higher GP values indicate more favorable condition for TC genesis.

Hereafter, we refer to relative vorticity and wind shear as dynamic conditions, while the others as thermodynamic conditions. Figure 4 compares the multi-model ensemble of the environmental conditions during July– November simulated by 20C3M (contour) with observed climatology (gray tone). Over the WNP, most of the TC genesis is associated with monsoon trough environment (Ritchie and Holland 1999), which is characterized by a zonally-elongating band of positive z850 . The main part of the trough is over the SCS and wWNP, while the CNP is covered with the eastern tail of the trough with only weak positive z850 (Fig. 4a). The models broadly reproduce the strength and zonal structure of the trough, apart from northward displacement of the simulated one which was already pointed out by Yokoi et al. (2009). The models also reproduce the gross horizontal pattern of vertical wind shear (Fig. 4b). Relatively weak shear is found in the monsoon trough region. Vertical shear is below 9 m s1 (650 hPa)1 at most of the CNP except for its northeastern sector. Gray (1975) reported that it was di‰cult for TCs to develop when wind shear between 950- and 200-hPa levels was greater than 10 m s1 (750 hPa)1 , suggesting that the vertical shear in the northeastern sector of the CNP is un-

June 2009


531

Fig. 5. Multi-model ensemble of global warming impacts projected by the five models. Di¤erences between global warming conditions with 1 K warming of the mean surface temperature, and the simulated current climate conditions are shown. Plotted variables are (a–e) the same as in Fig. 4, and (f ) zonal wind at the 850-hPa level in units of m s1 . Light and heavy shadings indicate that at least four of the five models project positive and negative trends, respectively.

favorable for TC genesis. As for thermodynamic conditions, observed PI (Fig. 4c) and H700 (Fig. 4d) over the SCS, wWNP, and CNP are higher than tropical (25 S–25 N) mean of about 50 m s1 and 50%, respectively. The models realistically simulate patterns of PI and H700 , while the latter exhibits slight northward shift bias which is associated with that of the monsoon trough. As a result of these conditions, the pattern of GP (Fig. 4e) is fairly reproduced by the models apart from a slight northward displacement. Note that we do not consider the absolute values of PI and GP, since di¤erent convection schemes produce slightly di¤erent temperature profiles in convective regions, resulting in sometimes large di¤erences in overall magnitude of PI and thus GP (Emanuel et al. 2008). Multi-model ensemble projection of environmental condition changes due to global warming is shown in Fig. 5. Increase of z850 is found in the area of the monsoon trough with the largest increment over the CNP (Fig. 5a), which indicates intensification and eastward extension of the trough. Vertical wind shear would increase to the west of the weak shear area and decrease to its east (Fig. 5b). These changes indicate eastward shift of the weak shear area. The PI (Fig. 5c) would increase all over the region south of 25 N, with the maximum increment in the equatorial area east of the date line. H700

(Fig. 5d) would also increase almost all over the basin and its increment is maximized in the area to the west of the date line near the Equator. As a result of these changes, GP would increase all over the WNP between the Equator and 30 N (Fig. 5e), with the maximum increment at 20 N, 140 E. These projected changes can be interpreted in terms of SST change during July–November shown in Fig. 6a. Higher rate of the projected SST warming in the equatorial zone implies that the situation would become more El Ninõ-like one, as pointed out by Meehl et al. (2007b), since the observed SST pattern during the El Ninõ phase exhibits a maximum positive anomaly in the equatorial zone as shown in Fig. 6b. The CNP region is located at the northwestern side of the maximum in both projected and El Ninõ-related patterns. On the other hand, there are several di¤erences between these two patterns. The projection exhibits a maximum rate in the so-called Nino4 region (160 E–150 W, 5 S–5 N), while the El Ninõ-related anomaly has a maximum to its east. Another di¤erence can be found along 20 N where SST would rise at a higher rate than to its north and south, while there is no counter part found in the observed El Ninõ situation. The maximum rate of SST warming in the Nino4 region is associated with the maximum in-

532


Vol. 87, No. 3

Fig. 7. Observed May–November TC genesis frequency simultaneously regressed on the Nino4 SST in units of number per 5 5 longitude–latitude box per 10 years. Shadings indicate that correlation coe‰cient is statistically significant at the 99% confidence level.

Fig. 6. (a) Multi-model ensemble of projected global warming impact on July– November SST in units of K per 1 K warming of the mean surface temperature. Light and heavy shadings represent warming rate higher than 0.6 and 0.7 K, respectively. (b) Observed July–November SST simultaneously regressed on the Nino4 SST. Light and heavy shadings indicate that correlation coe‰cient is negative and positive, respectively, and statistically significant at the 99% confidence levels.

crease of precipitation and amplification of lowertropospheric horizontal convergence (figure not shown). These changes would cause a positive trend in lower-tropospheric zonal wind to the west of Nino4 region (Fig. 5f ), corresponding to the weakening of the trade easterly. This trend in turn is directly linked with the increase of z850 to its north and thus the eastward extension of the monsoon trough. The SST maximum is also accompanied by the maximum increase of PI and H700 . The correspondence between SST and PI is reasonable, since the latter measures how much the local SST is higher than the overlying atmospheric thermodynamic profile that depends largely on tropicalmean SST (Vecchi and Soden 2007c). On the other hand, the amplified convective activity and precipitation would supply more amount of water vapor into the free troposphere, resulting in the increase of H700 .

Over the CNP, z850 , PI and H700 would become more favorable for TC genesis all over the region. The vertical shear would also become more favorable in the northeastern sector of the region, where stronger shear is observed and simulated under the current climate condition. In order to discuss which of these changes plays a more important role in the projected increasing trend of TC genesis than the others, we compare the global warming impacts with the observed IAV. Many previous studies revealed that higher frequency of TC genesis over the CNP was observed during El Ninõ phase (Chan 2000; Saunders et al. 2000; Chia and Ropelewski 2002; McDonald et al. 2005; Camargo et al. 2007). Figure 7 shows May–November genesis frequency simultaneously regressed on the Nino4 SST. As in the previous studies, positive anomalies with statistical significance can be found in the CNP, accompanied by significant negative anomalies to the east of Taiwan Island. Figure 8 shows observed environmental conditions during July–November simultaneously regressed on the number of TC genesis over the CNP. The regressed dynamic conditions, namely z850 (Fig. 8a) and vertical shear (Fig. 8b), exhibit quite similar horizontal patterns to the projected global warming impacts (Figs. 5a,b). The monsoon trough intensifies and extends eastward in years with more TCs, with westerly wind anomalies at 850 hPa south of the CNP (Fig. 8f ). As for the vertical shear, negative anomaly is found in the north-

June 2009


533

Fig. 8. Observed July–November environmental conditions simultaneously regressed on the TC genesis number over the CNP. Variable presented in each panel is the same as in Fig. 5. Light and dark shadings indicate that correlation coe‰cient is positive and negative, respectively, and statistically significant at the 99% confidence level.

eastern sector of the CNP with a positive anomaly to its southwest. In contrast, the regressed thermodynamic conditions (PI and H700 ) exhibit somewhat a di¤erent horizontal pattern from the projected global warming impacts. While the regressed PI (Fig. 8c) possesses a positive maximum in the equatorial region, negative anomalies prevail in the SCS, wWNP, and a substantial part of the CNP. As for H700 (Fig. 8d), positive anomalies with statistical significance is restricted to the equatorial areas east of 150 E. As a result of these changes, to the south of 20 N, the regressed GP shows clear east–west contrast with positive anomalies east of 150 E and negative ones to its west. Note that the regressed maps can also be obtained with the use of ENSO indices such as Nino4 SST as a reference variable, because of the significant correlation between the genesis frequency over the CNP and ENSO. The regressed GP pattern is also quite similar to that associated with ENSO (Camargo et al. 2007). The correlation coe‰cients between IAVs of the TC number and the environmental conditions over the CNP during July–November are summarized in Table 3. Both of the dynamic components correlate significantly with the TC number, suggesting that if these components become more favorable for TC genesis, the TC number will actually tend to increase. In contrast, the thermodynamic components

Table 3. Simultaneous correlation coe‰cients among observed IAVs of TC number and the environmental conditions over the CNP during July–November. Asterisks indicate that the correlation coe‰cient is statistically significant at the 99% confidence level.

z850 vertical shear PI H700

vertical shear

PI

H700

TC number

0.29

0.57* 0.27

þ0.16 0.27

þ0.61* 0.62*

0.18

0.27 þ0.31

do not show significant correlation with TC number. Besides, the PI and TC number correlates negatively, which is inconsistent with a widely-accepted fact that higher PI, and thus higher SST, are conductive to TC genesis. There is a possibility that this inconsistency is due to the significant negative correlation of PI against z850 shown in Table 3. To discuss this possibility, we plot observed July–November z850 and PI over the CNP along with TC numbers in Fig. 9. Even if we focus our discussion on the years with z850 between þ2 and þ3 [106 s1 ], we can not find positive correlation between the PI and TC numbers, the latter of which are indicated as markers in the figure. For example, in 1991 and 1992, the TC number is more than the 20-yr average (3.25)

534


Fig. 9. Scatter plot of observed July– November z850 and PI over the CNP of each year. The numbers plotted as markers represent TC number over the CNP in July–November of the respective years. The marks of the years 1987, 1990, 1991, and 1992 are labeled.

Vol. 87, No. 3

genesis, and vertical shear would not change significantly. The resultant GP trend also exhibits that the environment would become more favorable condition. These changes are inconsistent with the projected decrease of TC genesis. Such disagreement between results of the direct and indirect approaches was also pointed out by Caron and Jones (2008). On the other hand, the observed IAV of the TC numbers during the TC season over the SCS and wWNP (May–November) is not correlated significantly with any of the four environmental conditions studied here, suggesting that these conditions already meet the necessary condition for TC genesis almost every year in the current climate. Therefore, the projected changes that make environmental conditions slightly more favorable for the genesis probably have little impacts on the actual TC genesis frequency. So we try to find another candidate that contributes to the decrease projection of TC genesis. In addition, the observed IAV of the TC number over the SCS and wWNP does not correlate significantly with ENSO indices (Fig. 7), and therefore it seems di‰cult to discuss the global warming impacts by analogy with ENSO phenomena in these regions. 5. Activity of TD-type disturbance

while the PI is less than the average (74.6 m s1 ). On the other hand, in 1987 and 1990, the TC number is less than the average but the PI is more than the average. Actually, PI is above 70 m s1 every year, which is considerably higher than the tropical average of about 50 m s1 . In addition, July–November SST is above 28.4 C every year (figure not shown), which is also considerably higher than SST threshold for TC genesis of 26 C in the current climate (Gray 1975). It can be said that PI and SST satisfy necessary conditions for TC genesis every year, irrespective of ENSO phases in the current climate. We can therefore conclude that the projected slight increases of PI and SST due to global warming do not play a role in TC increase. Similar discussion can also be applied to the relative humidity; H700 averaged over the CNP is higher than the tropical mean of 50%. In summary, the increase projection of TC genesis over the CNP can be interpreted by analogy with the El Ninõ phenomenon, and the changes in the dynamic environmental conditions play a significant role in the increase projection. Over the SCS and wWNP, Fig. 5 shows that z850 , PI, and H700 would become more favorable for TC

In order to discuss a possible cause of the decrease projection of TC genesis over the SCS, we examine projected global warming impact on the activity of pre-existing disturbances, which is considered to be necessary for TC genesis in addition to establishment of the favorable environmental conditions. Heta (1990) reported that most of the TCs generated over the WNP in 1980 originated from vortex disturbances, or TD-type disturbances (Takayabu and Nitta 1993), which had propagated westward or northwestward along the monsoon trough. Liebmann et al. (1994) revealed that, in intra-seasonal time scales, larger number of TC genesis was associated with the larger number of the TD-type disturbances. It is therefore possible that the projected trend in the TC genesis frequency is associated with that of the activity of the TDtype disturbance. Here, the activity is measured by variance of 3–5-day component of meridional wind at the 850-hPa level, following Takayabu and Nitta (1993). Figure 10a shows the observed activity during the May–November period normalized by that averaged over the domain shown in this figure. The horizontal distribution is characterized by a

June 2009


Fig. 10. Climatological activity of the TDtype disturbance during May–November (a) observed and (b) simulated in 20C3M (multi-model ensemble). The activity is normalized by that averaged over the plotted domain. (c) Projected global warming impact on the activity per 1 K warming of the mean surface temperature. Light and dark shadings indicate that all of the five models project positive and negative trends, respectively. (d) The activity averaged over three years with more TCs generated over the SCS minus that over seven years with less TCs. Shadings indicate the di¤erence is larger than þ0.2.

band of high activity orienting in a southeast to northwest direction from the equatorial region near the date line to the northern part of the SCS.

535

Individual disturbances propagate northwestward along this band (Lau and Lau 1990; Takayabu and Nitta 1993). Figure 10b shows a multi-model ensemble of the activity simulated in 20C3M. The models fairly reproduce the band of high activity, while it is located slightly to the northeast of the observed one in regions east of 120 E, which is probably consistent with the northward shift bias of the monsoon trough. Over the SCS, the meridional distribution is realistically reproduced. Projected trend in the activity of the TD-type disturbance due to global warming is shown in Fig. 10c. The activity would decrease over most of the domain shown. The decrease rate is large in regions southwest of the band of high activity. On the other hand, the activity would increase slightly to the east of the band. Such horizontal distribution indicates weakening and eastward shift of the band. The latter change may be consistent with the projected eastward extension of the monsoon trough. The substantial decrease over the northern Philippines and to its east is projected by all of the five models. The projected decrease trend of TC genesis in the SCS is probably associated with the decrease of the TD-type disturbances to its southeast. The relationship between the TC genesis frequency and the activity of the TD-type disturbance can also be found in the observed IAV in the current climate. To explain this relationship, we compare the activities of TD-type disturbance observed in years with more TCs generated over the SCS and that in years with less TCs. In the three years of 1983, 1990, and 1995, the number of TCs generated over the SCS during May–November was larger than the climatological average (4.3) by at least one standard deviation (1.3). On the other hand, the number was less than the average by at least one standard deviation in the seven years of 1982, 1987, 1988, 1989, 1991, 1993, and 2001. Figure 10d shows di¤erence between the activity averaged over the three years with more TCs and that over the seven years with less TCs. Not only over the SCS but also over the Philippines and the wWNP, the variance in the years with more TCs is larger than that in the years with less TCs. In fact, in all of the three years with more TCs, the variance averaged over the (5–15 N, 120–130 E) area is larger than climatology. As for the seven years with less TCs, on the other hand, the variance in four years (1982, 1987, 1991, and 2001) is smaller than climatology, while in the other three years it is

536


larger. It can therefore be said that activeness of the TD-type disturbance to the southeast of the SCS is a necessary condition, but not a su‰cient condition, for the large number of TC genesis over the SCS. If IAV characteristics would not change significantly, projected weakening of the climatological activity of the TD-type disturbance may decrease the number of years with higher activity and the larger number of TC genesis, leading to the decreasing trend of the climatological TC genesis frequency. This confirms the contribution of the change in the activity of the TD-type disturbance to the decrease projection of TC genesis over the SCS. Note that the overall weakening of the activity of the TD-type disturbance may be consistent with discussion in Sugi et al. (2002). They demonstrated that intensity of tropical circulation measured by strength of instantaneous upward and downward mass fluxes integrated all over the tropics would decrease due to CO2 doubling, which can be interpreted as decrease of total activity of tropical disturbances. The weakening of tropical circulation is projected not only by Sugi et al. (2002) but also by most of the CMIP3 models (Vecchi and Soden 2007b). 6. Summary and discussion This study examined global warming impacts on the TC genesis frequency over the WNP basin projected by five CGCMs that participate in the CMIP3. Since projected trends in regional genesis frequency are di¤erent among models and studies (e.g., Meehl et al. 2007b), we took the multi-model approach to try to find projected trends that are consistent among the models. We analyzed dailymean outputs of 20C3M (40-year length) and global warming scenario experiments with the emission scenarios SRES A1B, A2, and B1 (80- to 160-year length in total) that have been archived at PCMDI database. All of the five models show that the TC genesis would increase over the CNP and decrease over the SCS and wWNP. The fact that all of the five models project the same sign of the frequency trends suggests robustness of the projection. The former increase projection is attributable to the changes in environmental conditions that are considered to a¤ect the TC genesis. The monsoon trough, where most of the TC genesis over the WNP is observed, would intensify and extend eastward, and the vertical wind shear would weaken where it is strong in the current climate. These projected changes closely resemble the El Ninõ condi-

Vol. 87, No. 3

tions, which is in good agreement with the fact that the pattern of projected SST warming is El Ninõ-like. While both dynamic and thermodynamic conditions would become more favorable for TC genesis over the CNP due to global warming, the change of the former condition is considered to be more important than that of the latter. This is because the observed IAV of TC genesis frequency over the CNP in the current climate is significantly correlated only with the dynamic conditions. On the other hand, the thermodynamic conditions are satisfied for TC genesis every year in the current climate, and thus their projected slight changes do not seem to contribute to the increase trends of TC genesis frequency. As for the SCS and wWNP, the environmental conditions would also become more favorable for TC to develop, in spite of the decrease trend of TC genesis frequency. We attribute this trend to the weakening of the activity of the TD-type disturbance, which is one of the disturbances that can later be developed into TCs. The active area of the TD-type disturbance in the current climate extends from the equatorial region near the date line to the northern SCS, which is projected to weaken and to shift eastward. As a result, the activity over and to the southeast of the SCS would decrease significantly. Note that the decrease projection of TC genesis over the SCS and wWNP is not El Ninõlike, since the observed IAV of TC genesis in these regions does not correlate significantly with ENSO indices. McDonald et al. (2005) showed a similar east– west contrast in projected genesis frequency change with the aid of the direct approach. Since they analyzed only 15-year-length simulations of presentday climate and global warming experiments performed with their own atmospheric general circulation model, our study has advantages in the robustness of the projection. In addition, figures in Bengtsson et al. (2007) showed that the genesis frequency would decrease substantially over the SCS and wWNP while it would not change significantly over the CNP. The zonal contrast of their result is consistent with ours. On the other hand, Stowasser et al. (2007) projected substantial increase of TC genesis over the SCS with the use of their regional climate model. They attributed the increase projection to the increase in relative humidity and weakening of vertical shear, with the latter of which our multi-model ensemble result disagrees. Since they analyzed an extreme global warming experiment

June 2009


with an environment of sixfold CO2 concentration, it may not be appropriate to directly compare our results with theirs. Nevertheless, it seems necessary to discuss such di¤erences among studies in detail in order to further confirm the degree of certainty of our multi-model projection. This study suggested that TCs would be generated over the CNP more frequently in the warmer climate than in the current climate, which leads to slight eastward shift of the frequent TC genesis area. This shift may lead to the change of favorable track and mean intensity of TCs over the WNP. Wu and Wang (2004) suggested that the number of TCs that approach Japanese Islands may increase if the mean location of TC genesis shifted eastward by 10 . The TCs that develop over the CNP tend to have longer lifetimes over the tropical warm ocean than that generated over the wWNP, and thus have a higher chance to become more vigorous (Chan 2007). Therefore, our results imply that the mean intensity of TCs over the WNP would increase at a higher rate than that over the other TC basin such as the tropical North Atlantic and Indian Ocean. These issues remain for our future study. Acknowledgments The authors acknowledge the developers of the CGCMs used in this study and WCRP’s CMIP3 project. Support of the PCMDI dataset is provided by the O‰ce of Science, U.S. Department of Energy. The ‘‘Data Integration and Analysis System’’ Fund for National Key Technology from the Ministry of Education, Culture, Sports, Science and Technology, Japan supported the authors in obtaining and processing the dataset. The authors are also grateful to the European Centre for MediumRange Weather Forecasts, National Oceanic and Atmospheric Administration, and Japan Meteorological Agency for their provision of reanalysis and observation datasets. This study is financially supported by the Global Environmental Research Fund (S-5-2) of the Ministry of the Environment, Japan. References Bengtsson, L., K. I. Hodges, M. Esch, N. Keenlyside, L. Kornblueh, J. -J. Luo, and T. Yamagata, 2007: How may tropical cyclones change in a warmer climate? Tellus, 59A, 539–561. Bister, M., and K. A. Emanuel, 1998: Dissipative heating and hurricane intensity. Meteorol. Atmos. Phys., 65, 233–240.

537

Bister, M., and K. A. Emanuel, 2002: Low frequency variability of tropical cyclone potential intensity 1. Interannual to interdecadal variability. J. Geophys. Res., 107, doi:10.1029/2001JD000776. Camargo, S. J., K. A. Emanuel, and A. H. Sobel, 2007: Use of a genesis potential index to diagnose ENSO e¤ects on tropical cyclone genesis. J. Climate, 20, 4819–2834. Caron, L. -P., and C. G. Jones, 2008: Analysing present, past and future tropical cyclone activity as inferred from an ensemble of coupled global climate models. Tellus, 60A, 80–96. Chan, J. C. L., 1985: Tropical cyclone activity in the northwest Pacific in relation to the El Ninõ/ Southern Oscillation phenomenon. Mon. Wea. Rev., 113, 599–606. Chan, J. C. L., 2000: Tropical cyclone activity over the Western North Pacific associated with El Ninõ and La Ninã events. J. Climate, 13, 2960–2972. Chan, J. C. L., and K. S. Liu, 2004: Global warming and Western North Pacific typhoon activity from an observational perspective. J. Climate, 17, 4590–4602. Chan, J. C. L., 2006: Comment on ‘‘Changes in tropical cyclone number, duration, and intensity in a warmer environment’’. Science, 311, 1713. Chan, J. C. L., 2007: Interannual variation of intense typhoon activity. Tellus, 59A, 455–460. Chia, H. H., and C. F. Ropelewski, 2002: The interannual variability in the genesis location of tropical cyclones in the Northwest Pacific. J. Climate, 15, 2934–2944. Emanuel, K. A., and D. S. Nolan, 2004: Tropical cyclone activity and global climate. Preprints, 26th Conf. on Hurricanes and Tropical Meteorology, Miami, FL, Amer. Meteor. Soc., 240–241. Emanuel, K., 2005: Increasing destructiveness of tropical cyclone over the past 30 years. Nature, 436, 686– 688. Emanuel, K., R. Sundararajan, and J. Williams, 2008: Hurricanes and global warming: Results from downscaling IPCC AR4 simulations. Bull. Amer. Meteor. Soc., 89, 347–367. Gualdi, S., E. Scoccimarro, and A. Navarra, 2008: Changes in tropical cyclone activity due to global warming: Results from a high-resolution coupled general circulation model. J. Climate, in press. Gray, W. M., 1975: Tropical Cyclone Genesis, Dept. of Atmospheric Science Paper, 234, Colorado State University, Fort Collins, CO, 121 pp. Heta, Y., 1990: An analysis of tropical wind fields in relation to typhoon formation over the Western Pacific. J. Meteor. Soc. Japan, 68, 65–77. Lander, M., 1994: An exploratory analysis of the relationship between tropical storm formation in the Western North Pacific and ENSO. Mon. Wea. Rev., 122, 636–651.

538


Lau, K. -H., and N. -C. Lau, 1990: Observed structure and propagation characteristics of tropical summertime synoptic scale disturbances. Mon. Wea. Rev., 118, 1888–1913. Liebmann, B., H. H. Hendon, and J. D. Glick, 1994: The relationship between tropical cyclones of the Western Pacific and Indian Ocean and the MaddenJulian oscillation. J. Meteor. Soc. Japan, 72, 401– 412. McDonald, R. E., D. G. Bleaken, D. R. Cresswell, V. D. Pope, and C. A. Senior, 2005: Tropical storms: representation and diagnosis in climate models and the impacts of climate change. Clim. Dyn., 25, 19–36. Meehl, G. A., C. Covey, T. Delworth, M. Latif, B. McAvaney, J. F. B. Mitchell, R. J. Stou¤er, and K. E. Taylor, 2007a: The WCRP CMIP3 multimodel dataset: A new era in climate change research. Bull. Amer. Meteor. Soc., 88, 1383–1394. Meehl, G. A., and co-authors, 2007b: Global Climate Predictions. Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change Solomon, S., and co-editors (Eds.), Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA. Nakicenovic, N., and co-authors, 2001: IPCC Special Report on Emissions Scenarios. Cambridge University Press, 599 pp. Oouchi, K., J. Yoshimura, H. Yoshimura, R. Mizuta, S. Kusunoki, and A. Noda, 2006: Tropical cyclone climatology in a global-warming climate as simulated in a 20 km-mesh global atmospheric model: Frequency and wind intensity analyses. J. Meteor. Soc. Japan, 84, 259–276. Reynolds, R. W., N. A. Rayner, T. M. Smith, D. C. Stokes, and W. Wang, 2002: An improved in situ and satellite SST analysis for Climate, J. Climate, 15, 1609–1625. Ritchie, E. A., and G. J. Holland, 1999: Large-scale patterns associated with tropical cyclogenesis in the Western Pacific. Mon. Wea. Rev., 127, 2027–2043. Saunders, M. A., R. E. Chandler, C. J. Merchant, and F. P. Roberts, 2000: Atlantic hurricanes and NW Pacific typhoons: ENSO spatial impacts on occurrence and landfall. Geophys. Res. Lett., 27, 1147– 1150.

Vol. 87, No. 3

Stowasser, M., Y. Wang, and K. Hamilton, 2007: Tropical cyclone changes in the western North Pacific in a global warming scenario. J. Climate, 20, 2378– 2396. Sugi, M., A. Noda, and N. Sato, 2002: Influence of the global warming on tropical cyclone climatology: An experiment with the JMA global model. J. Meteor. Soc. Japan, 80, 249–272. Takayabu, Y. N., and T. Nitta, 1993: 3–5 day-period disturbances coupled with convection over the tropical Pacific Ocean. J. Meteor. Soc. Japan, 71, 221– 246. Trenberth, K. E., and co-authors, 2007: Observation: Surface and atmospheric climate change. Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change Solomon, S., and co-editors (Eds.), Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA. Uppala, S. M., and co-authors, 2005: The ERA-40 reanalysis. Q. J. R. Meteorol. Soc., 131, 2961–3012. Vecchi, G. A., and B. J. Soden, 2007a: Increased tropical Atlantic wind shear in model projections of global warming. Geophys. Res. Lett., 34, doi:10.1029/ 2006GL028905. Vecchi, G. A., and B. J. Soden, 2007b: Global warming and the weakening of the tropical circulation. J. Climate, 20, 4316–4340. Vecchi, G. A., and B. J. Soden, 2007c: E¤ect of remote sea surface temperature change on tropical cyclone potential intensity. Nature, 450, 1066–1070. Wang, B., and J. C. L. Chan, 2002: How strong ENSO events a¤ect tropical storm activity over the western North Pacific. J. Climate, 15, 1643–1658. Webster, P. J., G. J. Holland, J. A. Curry, and H. -R. Chang, 2005: Changes in tropical cyclone number, duration, and intensity in a warmer environment. Science, 309, 1844–1846. Wu, L., and B. Wang, 2004: Assessing impacts of global warming on tropical cyclone tracks. J. Climate, 17, 1686–1698. Yokoi, S., Y. N. Takayabu, and J. C. L. Chan, 2009: Tropical cyclone genesis frequency over the western North Pacific simulated in medium-range coupled general circulation models. Clim. Dyn., doi:10.1007/s00382-009-0593-9.