Climate Dynamics (2004) 93: 165–176 DOI 10.1007/s00382-004-0427-8
C. Kobayashi Æ M. Sugi
Impact of horizontal resolution on the simulation of the Asian summer monsoon and tropical cyclones in the JMA global model
Received: 25 July 2003 / Accepted: 8 March 2004 / Published online: 29 May 2004 Ó Springer-Verlag 2004
Abstract To investigate the impact of increasing horizontal resolution on a simulated model climate, we conducted an experiment using the Japan Meteorological Agency (JMA) operational global atmosphere model (JMA-GSM0103). The models with four different horizontal resolutions ranging from T42 to T213 have been integrated over three years with prescribed climate sea surface temperature in the experiment. The distributions of 3-year averaged seasonal mean fields are basically similar among the models with different resolution, although there are some monotonic and systematic differences with increasing resolution. However, the climatology of synoptic scale phenomena is well represented in higher resolution models. The position and amount of precipitation in Baiu front (or ‘‘Mei-yu’’) at higher resolution models agree well with observations. The start time of northward propagation of heavy precipitation over the Bay of Bengal, which is associated with Indian monsoon development, is also well simulated in higher resolution models. The number of tropical cyclones increases monotonically with resolutions. The simulated tropical cyclones become more realistic with increasing resolution.
1 Introduction Recent development of parallel super computer technology has made possible to run a very high-resolution atmospheric general circulation models (AGCM) for climate simulation. Is the simulated climate significantly C. Kobayashi (&) Æ M. Sugi Climate Research Department, Meteorological Research Institute, Japan Meteorological Agency, 1-1 Nagamine Tsukuba-shi, Ibaraki-ken, Japan 305-0052 E-mail:
[email protected] Present address: C. Kobayashi Climate Prediction Division, Japan Meteorological Agency, 1-3-4 ote-machi chiyoda-ku, Tokyo, Japan 100-8122
improved by using a very high-resolution model? To investigate the influence of horizontal resolution on simulated climate produced by AGCM, a large number of studies have been conducted (e.g. Tibaldi et al. 1990, Boyle 1993; Williamson et al. 1995). Many of them concluded that considerable differences among the simulations occur from T21 to T42 and T42 is adequate for simulations of monthly to seasonal time scale global climate. They found large differences only in the detailed structure of various variables that are related to the representation of orography. However, Sperber et al. (1994) showed that T42 fails to correctly simulate monsoons when using the European Centre for Medium-Range Weather Forecasts (ECMWF) model. They also showed that T106, which is the highest resolution of their experiment, represents the best simulation of monsoon flow. However, they integrated over only one annual cycle at each of four horizontal resolutions and they did not use an ensemble method. Therefore, there is a possibility that their result is affected by including the internal variability of a particular year. On the other hand, the representation of tropical cyclones is expected to be better in higher resolution models. The structure of individual storms is represented more realistically in T106 (e.g. Bengtsson et al. 1995). An estimation of the number of tropical cyclones under global warming conditions has been investigated using T106 models (e.g. Bengtsson et al. 1996; Sugi et al. 2002), but there has been no clear description of the dependency of simulated tropical cyclone frequency on horizontal resolution. The main purpose of this study is to evaluate the effect of horizontal resolution on the simulated climate of Asian monsoon and tropical cyclones. The design of experiment is described in next section. After describing some large-scale features in Sect. 3, an evaluation of the simulation of Indian and East Asian monsoon are presented in Sect. 4. Sect. 5 describes the result for tropical cyclones and a general discussion follows in Sect. 6.
166
Kobayashi and Sugi: Impact of horizontal resolution on the simulation of the Asian summer monsoon
Fig. 1 Seasonal mean zonal averaged temperature differences from ERA15 climate a–e for DJF f–j JJA a, f: ERA15 climate, b, g: T42, c, h: T63, d, i: T106 and e, j: T213, respectively. Contours; 5 K for climate and 0.5 K for differences. Negative errors dashed. The areas where the difference is 95% significant are shaded
Table 1 Annual mean global averaged precipitation (unit: mm/ day). Additional experiment denoted T42TS1 GPCP T42 Prescription rate 2.64 (mm/day) Rate against T42
T63
2.60 2.69
T106
T213
T42TS1
2.77
2.82
2.70
103% 106% 108% 104%
2 Design of the experiment In this study, we utilize the Japan Meteorological Agency (JMA) Global Spectral Model (JMA-GSM0103) which
is used for operational forecasts from March 2001. Detailed descriptions are reported in JMA (2002). This model was integrated through three annual cycles at each of four horizontal resolutions, T42, T63, T106 and T213. These models have the same vertical resolution. The number of vertical layers is set to 40. The top level of the model is located at 0.4 hPa. All the models have common physical processes. The model is tuned for operational 1-week prediction with T106 resolution first, then only the gravity wave drag parameter is modified for T213. Because parameters of the lower resolution models are not modified, the results of low resolution models may have some disadvantage.
Kobayashi and Sugi: Impact of horizontal resolution on the simulation of the Asian summer monsoon
167
Fig. 2 Seasonal mean zonal averaged precipitation for a–d DJF and e–h JJA. a, e: T42, b, f: T63, c, g: T106 and d, h: T213, respectively. GPCP climate is represented by heavy solid line in each panel. Model’s climate is represented by a thin solid line
To use possible longest time steps, the time step length is calculated every hour using CFL condition during the integration. The maximum time step length is given on each horizontal resolution model. For example, T42’s time step is limited to 2000 s, which is five times longer than that of T213. To assess the impact of time steps on the simulation, an additional integration was performed at T42 resolution using shorter time steps than the maximum time steps of standard experiments. The initial time of the integration is 12 UTC on October 2nd in 2000 of the JMA Global analysis. These simulations use the monthly mean climatological SSTs of Reynolds and Smith (1994). The SST is updated daily using linear interpolation. Although three annual cycles is a short time to discuss the climate of a model, experience suggests it is sufficient length to see the climate of monthly mean or longer mean field. Most of the figures in this study show a seasonal mean field over three years as model climate.
For validations of model climate, the 15-year mean fields of the ERA15 (Gibson et al. 1997) have mainly been used. The 21-year mean precipitation data of GPCP (Global Precipitation Climatology Project, Huffman et al. 1997) version 2 has been used for validation of monthly or seasonal mean precipitation. The pentad precipitation data of Xie and Arkin (1997) has also been used for validation. For validation of tropical cyclones, the best track data of the US Navy has been used.
3 Seasonal mean field 3.1 Temperature The differences of seasonal mean zonally averaged temperatures from ERA15 climate are shown in Fig. 1. The areas where the difference is 95% significant are shaded.
168
Kobayashi and Sugi: Impact of horizontal resolution on the simulation of the Asian summer monsoon
Fig. 3 DJF mean meridional mass stream functions. a: ERA15 climate, b: T42, c: T63, d: T106 and e: T213, respectively. Contours; 2*1010 kg/s
The tropospheric temperature is warmer with increasing resolution particularly at mid-latitudes in both December-February (DJF) and June-August (JJA) seasons. This tendency is also noticed in the ECMWF model (Brankovic and Gregory 2001, hereafter BG2001), Hadley Centre Atmospheric Model version 2 (HadAM2b) (Stratton 1999) and National Center for Atmospheric Research (NCAR) model (Williamson et al. 1995). Stratton (1999) attributes the feature to the increased intensity of the hydrological cycle at higher resolution due to more intense vertical motions. To support this speculation, annual mean global averaged precipitation rates are shown in Table 1. The
precipitation rate increases as resolution increases. The rate of precipitation in T213 is 8% larger than that of T42. One may suspect this increase is due to shorter time step in the higher resolution model. To remove this effect, additional integration was performed at T42 resolution using the shorter time step, which is the same as the maximum time step length of T213. The result of this run is shown as ‘‘T42TST’’ in Table 1. The annual mean global averaged precipitation rate of this run is 4% larger than that of standard T42 experiment. Therefore, the reason for the increase in precipitation in higher resolution models is not only due to increasing resolution but partly due to the effect of shorter time step.
Kobayashi and Sugi: Impact of horizontal resolution on the simulation of the Asian summer monsoon
169
Table 2 Pattern correlations of seasonal mean global precipitation for DJF (top) and JJA (bottom) DJF GPCP
GPCP
T42
T63
T106
T42 T63 T106 T213
0.79 0.81 0.80 0.81
0.95 0.93 0.91
0.97 0.95
0.96
0.79 0.81 0.82 0.81
0.96 0.94 0.92
T213
JJA T42 T63 T106 T213
Fig. 4 Seasonal mean precipitation. a: GPCP climate in DJF, b: T106 in DJF, c: CPCP in JJA and d: T106 in JJA. Contours; 2.0, 4.0, 8.0 mm/day. The areas more than 2.0 mm/day and 8.0 mm/ day are shaded gray and black respectively
3.2 Precipitation and meridional streamfunctions Figure 2 shows the seasonal mean zonally averaged precipitation pattern. The observed pattern is reasonably well simulated with each resolution in both JJA and DJF. However, in DJF, the simulated maximum precipitation peak, which represents ITCZ, is located at around 7°N, although the observed peak position is in the Southern Hemisphere. The amount of overestimation at around 7°N reached 2 mm/day in all resolutions. In JJA, the precipitation peak over the equatorial area is represented well in the Northern Hemisphere. However, there is 2 mm/day overestimate at around 7°S in
0.97 0.96
0.98
all resolutions in the Southern Hemisphere. The precipitation in the tropics tends to be overestimated, particularly in the winter hemisphere. The error patterns are similar among the different resolution models. The resolution dependence is found around the equator. The precipitation amount of high resolution models around the equator is less than that in low-resolution models. We have seen that the most salient feature of the zonal mean precipitation error of the model is the excessive precipitation of winter hemisphere ITCZ in both DJF and JJA. In addition, in DJF the precipitation of Summer Hemisphere ITCZ is excessive, and as a result zonal mean precipitation in the model in DJF shows a distinctive ‘‘double ITCZ’’ structure. This double ITCZ structure is seen at all resolution and becomes more distinctive with increasing resolution. The double ITCZ structure is also clearly seen in the structure of Hadley cell (Fig. 3). In the ERA15 (Fig. 3a), the ascending branch of Hadley cell is located between 15°S and 7°N with the maximum at around 9°S. In the high resolution models (Fig. 3c–e), there are two maxima of ascending motion at the both side of the equator and minimum at around the equator. We have seen a distinctive double ITCZ structure in the zonal mean precipitation of the model in DJF. However, this does not necessarily mean that the double ITCZ structure is seen at all the longitudes. Figure 4 shows the observed and simulated spatial distribution of seasonal mean precipitation. In DJF, there are significant overestimated areas in the equatorial Indian Ocean and in the equatorial eastern Pacific Ocean. The underestimated areas are in North Africa, the eastern North Pacific, Australia, and South America. In JJA, there are overestimated areas around 5°S in Indian Ocean and 5°N in the eastern tropical Pacific. These error patterns in both seasons do not depend on model resolution. Deficiencies in the physical parametrizations of the model, particularly the cumulus convection scheme, may be responsible for these errors. Table 2 shows the pattern correlations of seasonal mean global precipitation. When we compare the simulated precipitation with observed precipitation, the model’s pattern correlations are about 0.8, while the
170
Kobayashi and Sugi: Impact of horizontal resolution on the simulation of the Asian summer monsoon
Fig. 5 Monthly mean precipitation over summer monsoon region in June. a: GPCP climate, b: T42, c: T63, d: T106 and e: T213, respectively. The areas more than 8.0 mm/day are shaded. The rain area denoted A and B in a make up the Meiyu-Baiu front. C is a heavy rain area over Arabian Sea. D is a heavy rain area over Bay of Bengal. The area X in b is a false rain area in the model
pattern correlations are greater than 0.9 when we compare the precipitation among the models. This feature is found in both seasons.
4 Asian summer monsoon Figure 5 shows the monthly mean precipitation over the summer monsoon region in June. The west to east
Fig. 6 Precipitation averaged between 130°E and 140°E in June. a T42, b T63, c T106, d T213. GPCP : heavy solid line models: thin solid line
elongating rain zone from eastern China to Japan, which is denoted A and B in Fig.5a, is associated with a stationary front which have a large impact to eastern Asian climate in summer. As shown in Fig.5a, the rain band corresponding to the Baiu-front marked A is simulated in all resolutions. However, as shown in Fig 6, the peak
Kobayashi and Sugi: Impact of horizontal resolution on the simulation of the Asian summer monsoon
171
Fig. 7 As Fig. 5 but for 100 hPa height field. a ERA15 climate. Contours; 50 m. The areas greater than 16,700 m are shaded
position of precipitation averaged between 130°E and 140°E is located at 25°N in T42, although the observed peak is at 30°N. The maximum precipitation amount in T63 is less than the observed value. T106 and T213 produce reasonable position and maximum precipitation amount of the Baiu-front. Figure 6 also shows systematic underestimation in the north of the Baiu-front and overestimation in the south of the Baiu-front in all resolutions. To check the climate of the Tibetan High as a background of Baiu-front, monthly mean 100 hPa height fields in June are shown in Fig. 7. The strength
Fig. 8a–e As Fig. 5 but for water vapor flux and their divergence. Negative divergence areas are shaded gray. The area with less than –8*10–5 (kg/m2/s) are shaded black
of the Tibetan High affects the location of Baiu-front. As an index of the Tibetan High strength, we show the mean height of 100 hPa. The Tibetan High in the models become stronger with increasing resolution and the strength in T213 is realistic. Although the reason of this intensification is not clear in this study, it is consistent that the tropospheric warming as shown in Fig. 1. Associated with this intensification, the axis of the subtropical jet near Japan is located slightly further
172
Kobayashi and Sugi: Impact of horizontal resolution on the simulation of the Asian summer monsoon
Fig. 9 Latitude-time cross section of the 5 day mean precipitation averaged between 125°E and 135°E. a: CMAP climate, b T106. Contours; 2.0, 4.0, 8.0, 12.0 and 16.0 mm/day. The areas more than 8 mm/day are shaded
north with increasing resolution (not shown). This may be related to the position of the Baiu-front in the models. A false rain area is produced around 30°N 100°E in low-resolution models (heavy rain area X in Fig.5b). However, the amount of this false rain is gradually decreasing with increasing resolution. The center of heavy rain area in central to southeast China slightly moved to southeast coast of China with increasing resolution, although the rain area B is still weaker in the high-resolution models than observation. The reduction of the false rain area with increasing resolution is also seen in July and August (Figures, not shown). Figure 8 shows the monthly mean divergence of water vapor flux in June. It suggests the convergence of vapor flux near false rain area is prevented to form by orography with increasing resolution. Thus, we speculate that the appearance of a false rain area in lower resolution models is due to the lack of horizontal resolution. To represent the climatological position and shape of the Baiu-front extending from China to the south of Japan, it seems necessary to use high-resolution models, at least T106. To check the meridional propagation of Baiu-front with seasonal progress, the latitude time cross section of the 3-year averaged 5-day mean precipitation averaged between 125°E to 135°E is shown in Fig. 9. The northward propagation of the Baiu-front is a part of meridional propagation of trains of cyclonic and anticyclonic flow patterns in the lower troposphere observed over the region between the Arabian Sea and the western Pacific Ocean (Krishnamurti and Subrahmanyam 1982). As shown in Fig. 9a, the precipitation belt associated with Baiu front is propagating northward from 20°N to 30°N during end of May to beginning of July. The models simulated the northward propagation in all resolutions. However the northward propagation of ITCZ that located south of 20°N during June to November is poorly represented in the models.
We show the features of the Indian monsoon briefly. As shown in Fig. 5, the maximum precipitation amount over western India (heavy rain area C) increases as resolution increases. This trend was noticed by BG2001. They speculated that this change is ascribed to the orographic effect of Western Ghats and enhancement of the low-level jet off the Somali coast. The amount of rain on the Bay of Bengal (heavy rain area D) decreases with increasing resolution. This tendency is the opposite of the results obtained in BG2001. To check the seasonal progress of monsoons over the Bay of Bengal, Fig. 10 shows the latitude time cross section of the 3-year averaged 5-day mean precipitation averaged between 85°E and 95°E. The rain area, more than 2 mm/day, propagated northward in late April in observation, while the northward propagation starts in late May in T42. The starting time of the northward propagation is earlier in higher resolution models and becomes more realistic in T213.
5 Tropical cyclones The intensity of simulated tropical cyclones (TCs) is much weaker than the real TCs, although the structure of individual storms is represented more realistically in higher resolution models (Bengtsson et al. 1995). Therefore, the criteria from detecting simulated TCs should be different from that of real TCs. In order to detect TCs from model output, we adopted modified criteria of Bengtsson et al. (1995) as follows: Sea level pressure: candidate for a tropical cyclone center is a grid point at which the sea level pressure shows a local minimum, and the value is less than 1020 hPa. Vorticity: near the cyclone center, 850 hPa vorticity is greater than 3.0*10–5 s–1.
Kobayashi and Sugi: Impact of horizontal resolution on the simulation of the Asian summer monsoon
173
Fig. 10 Latitude-time cross section of the 5 day mean precipitation averaged between 85°E and 95°E. From top to bottom: CMAP climate, T42, T63, T106 and T213, respectively. Contours; 2.0, 4.0, 8.0, 12.0 and 16.0 mm/day. The areas with more than 8 mm/day are shaded
Maximum wind: maximum wind speed at 850 hPa near the cyclone center Vmax is greater than 10 m/s. Warm core: the average temperature difference from the area mean of surrounding region at 300 hPa, 500 hPa, 700 hPa and 850 hPa exceeds 1.5K. Upper level wind speed: maximum wind speed at 300 hPa is less than the maximum wind speed at 850 hPa near the cyclone center. Duration: TCs should continue for least 2 days.
To show the suitability of this criteria, we applied it to ERA15. Table3 shows the number of TCs detected by the criteria in ERA15. The grid point interval of ERA15 is 2.5° in both latitude and longitude. The average number of selected TCs is 80.1 per year, varying between 67 and 94, and their standard deviation is 8.4. There are approximately 80 TCs over the globe per year according to Gray (1979). Also the average number of TCs of the US Navy best track data (observed data) from 1979 to 1988 is 81.2 per year as shown in Table 4. We have noted that the regional number of TCs in ERA15 is somewhat different from
174
Kobayashi and Sugi: Impact of horizontal resolution on the simulation of the Asian summer monsoon
Table 3 The number of detected tropical cyclones in ERA15
1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 Total
NI
NWP
NA+NE
SI
SWP
SA+SE
Total
16 22 13 14 20 14 16 10 12 17 16 21 15 17 11 234
23 31 25 26 12 30 31 34 27 22 30 39 33 29 29 421
11 23 15 11 23 26 18 18 15 13 17 12 6 8 17 233
11 5 7 10 3 12 6 10 8 12 20 3 6 4 7 124
10 13 20 13 9 10 10 14 10 14 6 14 18 12 12 185
0 0 0 0 0 1 1 0 0 0 1 0 1 0 0 4
71 94 80 74 67 93 82 86 72 78 90 89 79 70 76 1201
NI: North Indian Ocean (–105°E), NWP: North Western Pacific (105°E–135°W), NA: North Atlantic, NEP: North East Pacific (135°W-), SI: South Indian Ocean (–105°E), SWP: South West Pacific (105°E–135°W), SA: South Atlantic, SEP: South East Pacific (135°W-)
observations. As compared with observed TCs, the number of TCs in the North Indian Ocean (NI) is much more in ERA15, and the number in the North Atlantic (NA) and the northeastern Pacific (NEP) is less, while the number in the north western Pacific is suitable. When the TC detection criteria are applied to ERA15, we can see that the global number of TCs is acceptable, although the regional numbers of TCs are somewhat different from observations. We may conclude that the TC detection criteria are suitable for detecting TCs in the model. For a fair comparison among the various resolution models, all simulated data are converted to 2.5° latitudelongitude grid data for detecting TCs. Figure 11 shows the simulated tracks of TCs for three years. The number of TCs in each ocean area is shown in Table 4. Although the 3-year integration is not a long enough period to estimate the statistical significance, the monotonic trend with increasing resolution indicates the effect of model resolution. At T42, the number of simulated TCs is only 10.3 per year. The total number of TCs at T63, T106 and T213 is 27.3, 42.3 and 58.3 respectively. There is a clear monotonic increasing trend with increasing resolution. The number of tropical storms at the highest resolution model (T213) is 5.7
Table 4 The number of annual mean tropical cyclones. The area name is same as Table 3. OBS indicate observed values of US Navy best track from 1979– 1988 (Sugi et al. 2002)
T42 T63 T106 T213 OBS ERA15
times larger than that of T42, although the number is only 73% of the observed number even at the highest resolution. Furthermore, there is area dependence. The percentage of simulated TCs in the northwest Pacific is less than the observed rate, while those in the south Indian Ocean are much larger than those observed. This trend exists at all resolutions. The number of TCs in the Southern Hemisphere is well simulated though only half of the number of observed TCs is represented in the Northern Hemisphere. We have seen that the seasonal mean simulated precipitation is not sensitive to the model resolution. In contrast, there is a strong resolution dependence in the number of simulated TCs. The reason for this may be that the mean precipitation in the model are formed not only from TCs but also from ‘‘not organized’’ convective activity like cloud clusters. The effect of resolution significantly appears on the formation of synoptic scale ‘‘organized’’ convective activity such as TCs. The convective activity in higher resolution models is more likely to be organized and develop into a synoptic scale disturbance with strong winds. Then the strong resolution dependence of TCs is found in the result. The number of TCs simulated in the T42 model is only 10.3 per year in the present study. However, Broccoli and Manabe (1990) and Tsutsui and Kasahara (1996) have shown that a reasonable number of TCs are simulated in low resolution models. They pointed out the tropical cyclone frequency in a model is sensitive to physical parametrization, particularly the parametrization of cumulus convection, and cloud and radiation. The model used for the present study is tuned for operational 1-week prediction with high (T106 and T213) resolution. The tuning of the physical parametrization may not be most appropriate to simulate TCs at lower resolution. Figure 12 shows the tropical cyclone frequency distribution as a function of maximum wind speed. Although all resolution models have the same physics, more intense TCs appear in higher resolution models. This result is collateral evidence of the tendency that the higher resolution models are more likely to organize convective activity. However the intensity is not sufficient even in T213 when compared with observations. Also the shapes of distribution plots are different between simulation and observation. The observed number of TCs whose maximum wind speed between 15 and 20 m/s is less than that 20 to 25 m/s, while the largest number of TCs are simulated between 15 and 20 m/s.
NI
NWP
NA+NEP
SI
SWP
SA+SEP
Total
std
0.0 2.3 4.3 3.3 3.9 15.6
2.3 3.7 10.3 18.3 27.4 28.1
0.0 2.3 5.3 10.3 25.2 15.5
4.7 11.7 14.0 12.3 10.6 8.3
3.3 7.3 8.3 13.7 14.0 12.3
0.0 0.0 0.0 1.0 0.1 0.3
10.3 27.3 42.3 59.0 81.2 80.1
12.3 8.4
Kobayashi and Sugi: Impact of horizontal resolution on the simulation of the Asian summer monsoon
175
Fig. 11a–d Simulated tracks of tropical cyclones for three years. a: T42, b: T63, c: T106, d T213
Although this may be due to the detecting scheme, the models fail to simulate this distribution.
6 Summary The main purpose of this work is to evaluate the effect of horizontal resolution on the simulation of the climate of the Asian monsoon and TCs that have a large impact on the East Asian summer climate. The structure and intensity of individual phenomenon are well represented as resolution increases. In common with previous studies, we found that the simulations of seasonal scale climates are basically similar among the models with different resolutions,
although there are some monotonic and systematic differences as resolution increases. The strong double ITCZ structure of the simulated tropical precipitation is not sensitive to horizontal resolution of the model. This seems to suggest that the double ITCZ structure is attributable to deficiency of physical parametrizations. Although the simulated large-scale seasonal mean climate is not strongly dependent on resolution, we have noted that some important local features of monsoon areas are simulated better with higher resolution models. In particular, to represent a realistic position and intensity of the rain band of Baiu (or Mei-yu), it is necessary to use a high resolution model, namely at least T106. This result is consistent with Sperber et al. (1994).
176
Kobayashi and Sugi: Impact of horizontal resolution on the simulation of the Asian summer monsoon
Fig. 12 Tropical cyclone frequency distribution as a function of maximum wind speed. Best track data is represented by a heavy solid line, ERA15 by a heavy dotted line, T42 by dotted, T63 by dot dashed, T106 by dashed and T213 by thin solid line
The total number of simulated TCs increases as resolution increases, though not enough even in T213. As is well known, the simulated tropical cyclone frequency is sensitive to physical parametrization of the model (e.g., Broccoli and Manabe 1990). Our result shows that simulation of TC frequency is also sensitive to model resolution. We consider the effect of resolution appears on the formation of synoptic scale ‘‘organized’’ convective activity. By climate modeling standards at present, T213 may be considered to be a very high resolution. However, it is not enough for realistic simulation of sub-synoptic scale phenomena such as fine structures of Baiu front and TCs. Further studies using higher resolution models are required to estimate the effect of sub-synoptic scale phenomena on climate simulation. Acknowledgements We thank Dr Takano for discussions and valuable suggestions. GrADS was used for the drawings. Thanks are also extended to the anonymous reviewers whose valuable comments and suggestions greatly improved the manuscript.
References Bengtsson L, Botzet M, Esch M (1995) Hurricane-type vortices in a general circulation model. Tellus 47A: 175–196
Bengtsson L, Botzet M, Esch M (1996) Will greenhouse gas-induced warming over the next 50 years lead to higher frequency and greater intensity of hurricanes? Tellus 48A: 57–73 Boyle JS (1993) Sensitivity of dynamical quantities to horizontal resolution for a climate simulation using the ECMWF (cycle33) model. J Clim 6: 796–815 Brankovic C, Gregory D (2001) Impact of horizontal resolution on seasonal integrations. Clim Dyn 18: 123–143 Broccoli AJ, Manabe S (1990) Can existing climate models be used to study anthropogenic changes in tropical cyclone climate? Geophys Res Lett 17: 1917–1920 Gibson JK, Kallberg P, Uppala S, Nomura A, Hernandez A, Serrano E (1997) ERA description. ECMWF Re-Analysis project Report Series 1, ECMWF, Reading, UK Gray WM (1979) Hurricanes their formation, structure and likely role in the tropical circulation. In: Shaw DB (ed) Meteorology over the tropical oceans. R Meteorol Soc pp 155–218 Huffman GJ, Adler RF, Arkin P, Chang A, Ferraro R, Gruber A, Janowiak J, McNab A, Rudolf B, Schneider U (1997) The Global Precipitation Climatology Project (GPCP) combined precipitation data set. Bull Am Meteorol Soc 78: 5–20 JMA (2002) Outline of operational numerical weather prediction at Japan Meteorological Agency. Numerical Weather Prediction Division/Japan Meteorological Agency. pp 157 Krishnamurti TN, Subrahmanyam D (1982) The 30–50 day mode at 850 mb during MONEX, J Atmos Sci 39: 2088–2095 Reynolds RW, Smith TM (1994) Improved global sea surface temperature analysis using optimum interpolation. J Clim 7: 920–948 Sperber KR, Hameed S, Potter GL, Boyle JS (1994) Simulation of the northern summer monsoon in the ECMWF model: sensitivity of horizontal resolution. Mon Weather Rev 122: 2461– 2481 Stratton RA (1999) A high resolution AMIP integration using the Hadley Centre model HadAM2b. Clim Dyn 15: 9–28 Sugi M, Noda A, Sato N (2002) Influence of the Global Warming on Tropical Cyclone Climatology: An Experiment with the JMA Global Model. J Meteorol Soc Japan 80: 249–272 Tibaldi S, Palmer TN, Brankovic C, Cubasch U (1990) Extended range predictions with ECMWF models: influence of horizontal resolution on systematic error and forecast skill. Q J R Meteorol Soc 116: 835–866 Tsutsui J, Kasahara A (1996) Simulated tropical cyclones using the National Center for Atmospheric Research community climate model. J Geophys Res 101(D10): 15,013–15,032 Williamson DL, Kiehl JT, Hack JJ (1995) Climate sensitivity of the NCAR Community Climate Model (CCM2) to horizontal resolution. Clim Dyn 11: 377–397 Xie P, Arkin PA (1997) Global precipitation: a 17-year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs. Bull Am Meteorol Soc 78: 2539–2558