Changes in precipitation intensity over East Asia during the 20th and 21st centuries simulated by a global atmospheric model with a 60 km grid size

Authors


Abstract

[1] We conducted three-member ensemble simulations using a global atmospheric model with a high horizontal resolution of a 60 km grid size for the period 1872–2099 (228 years). Between 1872 and 2005, the model was forced with observed historical sea surface temperatures (SST), while between 2006 and 2099, the boundary SST data were estimated using the multimodel ensemble of the Coupled Model Intercomparison Project Phase 3 models and assuming A1B emission scenario. Annual mean precipitation (PAVE), the Simple Daily Precipitation Intensity Index (SDII), and the maximum 5 day precipitation total (R5d) averaged over East Asia increase almost monotonically through the 21st century. The statistically significant area of precipitation intensity increase is larger for 2080–2099 than for 2046–2065. In particular, intense rainfall will increase over northern and southern China during 2080–2099. The conversion rate from water vapor to precipitation per 1°C rise in surface air temperature for SDII and R5D is much larger than that for PAVE during the 21st century. This suggests that extreme rainfall events will occur more frequently than moderate rainfall events even if the amount of temperature rise is same. Future changes in the horizontal transport of water vapor also lead to more intense precipitation over East Asia. In particular, the increase in clockwise water vapor transport due to intensification of the subtropical high contributes to increased intense precipitation over southern China.

1 Introduction

[2] Observational studies based on past historical long-term data indicate an increase in heavy precipitation over many regions of East Asia, such as Japan [Iwashima and Yamamoto, 1993; Fujibe et al., 2006], Korea [Kim et al., 2006; Ha et al., 2012], mainland China [Zhai et al., 2005], and Taiwan island [Hsu and Chen, 2002; Tu and Chou, 2013]. However, the observed changes in precipitation extremes are much less spatially coherent and less statistically significant than the observed changes in temperature extremes [Alexander et al., 2006; Intergovernmental Panel on Climate Change (IPCC), 2007, 2012].

[3] Most of the Atmosphere-Ocean General Circulation Models (AOGCMs) involved in the Coupled Model Intercomparison Project Phase 3 (CMIP3) underestimated the frequency of heavy precipitation in the tropics [Dai, 2006], East Asia [Kusunoki and Arakawa, 2012], and southern China [Tu et al., 2009]. The reproducibility of heavy precipitation during the East Asian summer rainy season is improved by the use of global models [Kusunoki et al., 2006] and regional models [Gao et al., 2006; Kanada et al., 2008; Kitoh et al., 2009] with a higher horizontal resolution.

[4] The AOGCMs of CMIP3 [Kimoto et al., 2005; Emori and Brown, 2005; Kripalani et al., 2007; Li et al., 2011] and also regional models [Boo et al., 2006; Su et al., 2009; Im et al., 2008, 2011; Kitoh et al., 2009] project future increases in the precipitation intensity of the East Asian summer monsoon season. A series of global warming projections using an atmospheric global model with a 20 km grid size also indicated future increases in precipitation intensity over East Asia [Kamiguchi et al., 2006; Kusunoki et al., 2006; Kusunoki and Mizuta, 2008; Kim et al., 2010; Endo et al., 2012]. However, these studies had to restrict the length of the target period to a few decades at the end of the 21st century, because the 20 km model requires huge computer resources. Consequently, we have been developing a 60 km mesh version of the 20 km model to conduct ensemble simulations that will quantify the uncertainty of future climate projections, as the computer time required for the 20 km model is 30 times larger than that of a 60 km model. For the purpose of investigating long-term climate change over the 228 years from 1872 to 2099, we conducted additional continuous simulations with the 60 km model. Focusing on changes in tropical cyclone activity, Sugi and Yoshimura [2012] reported a continuous decrease in the number of tropical cyclones at the global scale throughout the whole 228 year period. However, they did not investigate changes in extreme rainfall events over East Asia. The aim of this study is to reveal future changes in heavy precipitation over East Asia for the whole of the 21st century.

[5] As extreme precipitation events tend to be concentrated in the warm season over East Asia, many observational and modeling studies focus on the summer season (June to August). The average number of typhoons generated in the Northwestern Pacific each year was about 25.6 during the 30 years between 1981 and 2010 (Japan Meteorological Agency, http://www.data.jma.go.jp/fcd/yoho/typhoon/index.html). However, if we concentrate only on the summer, the average number of typhoons occurring falls to 11.2 (43.8% of the annual average). Typhoons often generate severe rainfall events; consequently, if the target season is limited only to the summer, many extreme rainfall events associated with typhoons are excluded from the analysis. Therefore, in this present study, we have consistently calculated annual statistics.

2 Model and Experimental Design

[6] The model used in this study was the Meteorological Research Institute–atmospheric general circulation model, version 3.2 (MRI-AGCM3.2), which was jointly developed by the Japan Meteorological Agency and MRI [Mizuta et al., 2012]. The model has a 60 km horizontal grid spacing and 60 levels with a 0.01 hPa top (an altitude of about 80 km). For cumulus convection, we implemented the “Yoshimura scheme” developed by Hiromasa Yoshimura [Yukimoto et al., 2011] which is based on the scheme of Tiedtke [1989]. This Yoshimura scheme considerably improved the climatology of tropical convection [Mizuta et al., 2012]. This 60 km model is the same as that used in previous studies that investigated changes in tropical cyclone [Murakami et al., 2012; Sugi and Yoshimura, 2012] and precipitation over Asia [Endo et al., 2012].

[7] A time-slice experiment [Bengtsson et al., 1996] was conducted, in which the high-resolution AGCM was forced by prescribed external boundary conditions and forcing. For the 134 years from 1872 to 2005, the model was integrated with the monthly mean of observed historical sea surface temperature (SST) and sea ice concentration from HadISST1 [Rayner et al., 2003]. For the 94 years from 2006 to 2099, the boundary SST data were created by superposing (i) future change in the multimodel ensemble (MME) of SST projected by the CMIP3 multimodel data set, (ii) the linear trend in the MME of SST projected by the CMIP3 multimodel data set, and (iii) the detrended observed SST anomalies for the period 1979–2003. Future change in the MME of SST was evaluated using the difference between the 20th Century Experiment in the Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report [IPCC, 2007] and the future simulation for the IPCC Special Report on Emission Scenario A1B emission scenario [IPCC, 2000]. In this procedure, the 25 year interannual variation of SST anomalies between 1979 and 2003 was repeatedly added to the CMIP3 SST projection for consecutive future 25 year periods. Future sea ice concentration was obtained in a similar fashion. Mizuta et al. [2008] describe the method in more detail. As for greenhouse gases, such as carbon dioxide and methane, observed historical concentrations were prescribed from 1872 to 2000. After 2001, concentrations based on the A1B emission scenario were prescribed. We used three-dimensional natural and anthropogenic aerosol distributions calculated by the MRI-Earth System Model [Yukimoto et al., 2011] based on historical and A1B scenario aerosol emission data. Aerosols from volcanic eruptions were included only for the Mt. Pinatubo eruption of 1991. The three-dimensional distributions of stratospheric ozone calculated by the MRI-Chemical Transport Model [Shibata et al., 2005] based on historical and A1B scenario aerosol emission data were prescribed. To evaluate uncertainty originating from the internal variability of the model atmosphere, three-member ensemble simulations were executed for three different atmospheric initial conditions. The experiment design was identical to that adopted by Sugi and Yoshimura [2012], but some errors in the description of the experimental design in Sugi and Yoshimura [2012] are corrected in the present paper. We mainly analyzed daily precipitation data archived for the whole 228 year period from 1872 to 2099.

3 Present-Day Climate Simulations

3.1 Verification Data

[8] To verify the simulated precipitation, we used the One-Degree Daily (1DD) data from the Global Precipitation Climatology Project (GPCP) V1.1 compiled by Huffman et al. [2001], which has a horizontal resolution of 1° in longitude and latitude, corresponding to a grid spacing of about 90 km over Japan. These GPCP V1.1 data have a higher resolution in time and space than conventional 2.5° data, but the period is limited to 1997 onward. We terminated GPCP data at 2005, because the model was forced with observed SST from 1872 to 2005.

[9] We also used the Asian Precipitation Highly Resolved Observational Data Integration Towards the Evaluation of Water Resources (APHRODITE) Version 1101 data compiled by Yatagai et al. [2009]. APHRODITE is a daily gridded precipitation data set that is based on rain gauge observations over Asia region, and has a horizontal resolution of 0.25° in longitude and latitude, corresponding to a grid spacing of about 25 km over East Asia. These data cover the whole period of the present-day climate simulation (1986–2005) but are restricted to land areas.

3.2 Indices of Precipitation Intensity

[10] We adopted several of the 10 indicators of precipitation intensity proposed by Frich et al. [2002]. The Simple Daily precipitation Intensity Index (SDII) is defined as the total annual precipitation divided by the number of rainy days (precipitation ≥ 1 mm/d). Unit is mm/d. If there are no rainy days at a grid point, we applied a missing flag to that grid point. SDII is widely used in model studies; e.g., Dai [2006] and chapter 10 entitled Global Climate Projection in IPCC [2007]. We also used the maximum 5 day precipitation total (R5d), as well as the similar R3d and R10d. Unit is millimeter. We further calculated the annual mean precipitation (PAVE) as a basic measure of model performance. Unit is mm/day.

3.3 Precipitation Climatology

[11] Figure 1 compares the observed climatology from GPCP (1997–2005) and APHRODITE (1986–2005, land only) with the simulated climatology (1986–2005) for the present-day climate. The observed PAVE from GPCP (Figure 1a) is large over the southern part of mainland China, Taiwan island, the East China Sea, and Japan. During the warm season, rainbands stagnating over East Asia and passing tropical cyclones mainly contribute to this rainy area tilting from south-west to north-east. The observed SDII from GPCP (Figure 1b) is large over the southern part of mainland China, Taiwan island, the Korean Peninsula, Japan, and the ocean to the south of Japan. In terms of R5d from GPCP (Figure 1c), heavy rainfall is observed over Taiwan island, the Korean Peninsula, the East China Sea, and the ocean to the south of Japan. The spatial distribution of R5d (Figure 1c) is almost the same as that of SDII (Figure 1b) with differing magnitudes. The spatial distribution of APHRODITE (Figures 1d–1f) also shows similar characteristics to those of GPCP; but in the southern part of China, SDII from APHRODITE (Figure 1e) is much smaller than that from GPCP (Figure 1b).

Figure 1.

The present-day climatology of precipitation. (a) Observed PAVE (mm/d) from the GPCP 1DD v1.1 data for 1997–2005 (9 years). R is the regional average. (b) Same as Figure 1a but for SDII (mm/d), (c) same as Figure 1a but for R5d (mm). (d) Observed PAVE (mm/day) from the APHRODITE Version 1101 data for 1986–2005 (20 years). (e) Same as Figure 1d but for SDII. (f) Same as Figure 1d but for R5d. (g) Annual precipitation simulated by the model for 1986–2005 (20 years). three-member ensemble average. B is the regional averaged bias (%) against GPCP observations normalized by R value in Figure 1a. C is the spatial correlation coefficient (%) between simulated precipitation and GPCP observations. (h) Same as Figure 1g but for SDII. (i) Same as Figure 1g but for R5d.

[12] The East Asia Monsoon has large seasonality [Chang et al., 2011; Zhou et al., 2011; Huang et al., 2012], but severe rainfall events are mostly concentrated in the summertime. However, the spatial distribution of the precipitation intensity index for summer (June to August) is very similar to the annual statistics. The spatial correlation coefficients between the summer and annual distributions of GPCP observations exceed 0.9 for all precipitation intensity indices (figures not shown), although the magnitude of the summer intensity differs from the annual statistics due to extreme events, such as typhoons, which occur outside summer.

[13] The simulated PAVE (Figure 1g) reproduces well the observed local precipitation maximum near Japan. The regional average annual precipitation from the model (Figure 1g) is 3.6 mm, which is almost equal to the observed value of 3.5 mm (Figure 1a). The spatial correlation coefficient between observations (Figure 1a) and the simulation (Figure 1g) is as high as 92%. The simulated SDII (Figure 1h) reproduces the intense precipitation over Japan but underestimates this over other regions. The regional average SDII from the model (Figure 1g) is 8.9 mm/d, which is 15% smaller than the observed value of 10.5 mm/d (Figure 1b). The spatial correlation coefficient between observations (Figure 1b) and the simulation (Figure 1h) is as low as 74%. The model (Figure 1i) reproduces the observed R5d distribution reasonably well (Figure 1c) but slightly overestimates it over the ocean to the east of Taiwan. The regional average R5d from the model (Figure 1i) has a positive bias of 12%. The spatial correlation coefficient between observations (Figure 1c) and the simulation (Figure 1i) is 89%, which is higher than that for SDII (Figures 1b and 1h). Consequently, we have confirmed that the model has the ability to reproduce the observed annual precipitation and precipitation intensity of the present-day climate, although the model performance largely depends on the measure of precipitation intensity used. Underestimation of precipitation intensity by the 60 km model for some areas, and for some intensity indices, may be attributed partly to the low horizontal resolution of the model, because the 20 km model enhances reproducibility of severe rainfall events when compared with lower-resolution models [Kusunoki et al., 2006].

4 Long-Term Variation of Precipitation

[14] Figure 2 shows the time series of simulated precipitation indices averaged over East Asia. The target area is the same as Figure 1. In the case of PAVE (Figure 2a), precipitation is almost constant from the 1870s to the 1970s, but later, it increases almost monotonically. Simulated precipitation agrees well with observations (red line) for the period 1997–2005. The amplitude of the year-to-year variability in the 21st century seems to be the same as that in the twentieth century. As for SDII (Figure 2b), precipitation intensity is almost constant from the 1870s to the 1970s, but later, it increases almost monotonically in a similar manner to PAVE. Simulated precipitation intensity is underestimated when compared with the observations, which is consistent with Figures 1c and 1d. The amplitude of the year-to-year variability increases slightly after the 2070s. The standard deviation of the ensemble average increases from 0.16 mm/d for the present-day period (1986–2005) to 0.22 mm/d for the future period (2080–2099). This increase is statistically significant at the 95% level based on F test using variances [Storch and Zwiers, 1999]. In the case of R5d (Figure 2c), the long-term trend resembles to PAVE and SDII. The simulated precipitation intensity is slightly overestimated compared with observations. The amplitude of the year-to-year variability increases after the 2070s. The standard deviation of the ensemble average increases from 6.3 mm for the present-day period (1986–2005) to 11.2 mm for the future period (2080–2099). This increase is also statistically significant at the 95% level. In summary, precipitation intensity increases in the 21st century and is associated with an amplification of year-to-year variability toward the end of the century.

Figure 2.

Time series of precipitation indices averaged over East Asia (100–150°E, 20–50°N; Figure 1) from 1872 to 2099 (228 years). Red line shows observations from GPCP 1DD v1.1 data for 1997–2005 (9 years). Black line shows the three-member ensemble average. Maximum and minimum ranges from the individual simulations are shaded. Green lines show three 20 year target periods (1986–2005, 2046–2065, and 2080–2099) for analyses in Figures 1, 3, and 4. (a) PAVE (mm/d). (b) SDII (mm/d). (c) R5d (mm).

[15] The precipitation distributions in Figures 1g–1i indicate that precipitation over the sea is much larger than that over land. Figure 3 shows the time series of precipitation averaged over sea and land separately. For PAVE (Figure 3a), precipitation over the sea is larger than over land, but with a similar increasing trend. Simulated precipitations over sea and land reproduce well observations from GPCP, but precipitation over land is somewhat overestimated compared with observations from APHRODITE (land only). For SDII (Figure 3b), precipitation over the sea is also larger than over land. The simulated precipitation over land reproduces well observations from APHRODITE, but it is underestimated when compared with observations from GPCP. The standard deviation of precipitation over the sea increases from 0.30 mm/d for the present-day period (1986–2005) to 0.41 mm/d for the future period (2080–2099). This increase is statistically significant at the 95% level. R5d (Figure 3c) shows similar characteristics to PAVE (Figure 3a). The standard deviation of precipitation over the sea increases from 12.5 mm for the present-day period (1986–2005) to 20.6 mm for the future period (2080–2099). This increase is also statistically significant at the 95% level. As for precipitation over land, the standard deviation changes from the present-day period (1986–2005) to the future period (2080–2099) are not significant for PAVE, SDII, or R5d.

Figure 3.

Time series of precipitation averaged over sea (blue) and land (red) for East Asia (100–150°E, 20–50°N; Figure 1). Black line indicates the all area average. Format is similar to Figure 2. Model values are three-member ensemble averages. Orange line indicates observations from APHRODITE (land only). Dotted line indicates observations from GPCP.

5 Geographical Distribution of Future Changes

[16] The geographical distribution of future changes is illustrated in Figure 4 for the periods 2046–2065 and 2080–2099. In the case of PAVE for 2046–2065 (Figure 4a), precipitation increases over almost all regions except for the oceans to the south of Japan. For the period 2080–2099 (Figure 4d), the statistically significant regions are much larger than for 2046–2065 (Figure 4a). As for SDII between 2046 and 2065 (Figure 4b), spatial pattern is very similar to PAVE (Figure 4a), but the statistically significant regions over the northern part of China is smaller than for PAVE (Figure 4a). For the period 2080–2099 (Figure 4e), precipitation intensity increases over all regions, with the statistically significant regions being much larger than for 2046–2065 (Figure 4b). In the case of R5d (Figure 4c), precipitation intensity generally increases over the whole region, but the statistically significant regions are small when compared with PAVE (Figure 4a) and SDII (Figure 4b). For the period 2080–2099 (Figure 4f), precipitation intensity increases over almost all regions except for the Japan Sea and oceans to the south of Japan. Precipitation intensity in terms of SDII and R5d generally increases over East Asia for 2046–2086 and 2080–2099. The degree of increase and the area of the statistically significant regions are much larger for 2080–2099 than for 2046–2065. In particular, an increase in precipitation intensity is evident across both southern and northern China. The increase of precipitation intensity in southern China is consistent with a previous study by Feng et al. [2011] that used a 40 km mesh global atmospheric model. The geographical distribution of changes in precipitation intensity based on R3d and R10d is very similar to that based on R5d (figures not shown).

Figure 4.

Future changes in precipitation indices (%) from the present-day climatology (1986–2005). Change is normalized by the present-day climatology. Hatched regions show the 95% significance level based on Student's t test. Three-member ensemble averages. (a) PAVE for 2046–2065. (b) Same as Figure 4a but for SDII. (c) Same as Figure 4a but for R5d. (d) Same as Figure 4a but for 2080–2099. (e) Same as Figure 4d but for SDII. (f) Same as Figure 4d but for R5d.

6 Mechanism of Precipitation Change

[17] The mechanisms driving increases in PAVE, SDII, and R5d indicated by Figures 2–4 can be roughly divided into two elements: thermodynamic environmental change and dynamical circulation change.

6.1 Thermodynamical Effect

[18] In principle, the increase in precipitation amount and precipitation intensity can be ascribed to the increased availability of water vapor caused by the rising air temperature of the atmosphere [IPCC, 2007, 2012]. The annual surface air temperature will rise by about 2 to 3° over East Asia for 2080–2099 (Figure 5a). This leads to increase in precipitation and precipitation intensity (Figures 4d–4f). Figures 5b and 5c show the rate of precipitation change per 1° rise in surface air temperature. This quantity can be regarded as a kind of “precipitation efficiency,” which measures the conversion rate from water vapor to precipitation. This precipitation efficiency is referred to as the “hydrological sensitivity” in IPCC [2001, section 9.3.4]. The precipitation efficiency of R5d (Figure 5c) is much larger than that of PAVE (Figure 5b), suggesting that strong rainfall events occur due to the very effective conversion from water vapor to precipitation when compared with moderate or weak rainfall events. The red contour in Figures 5a and 5b shows the rate of precipitation increase (7.5%/C) theoretically expected from the Clausius-Clapeyron (C-C) relationship [Vecchi and Soden, 2007]. The precipitation efficiency of PAVE (Figure 5b) does not reach the level of the C-C relationship in most areas, whereas precipitation efficiency of R5d (Figure 5c) exceeds the level of the C-C relationship in most areas.

Figure 5.

Precipitation efficiency. Three-member ensemble average. (a) Change in annual mean surface air temperature (°C) for 2080–2099 relative to 1986–2005. (b) Precipitation efficiency (%/°C) for PAVE defined as PAVE change (Figure 3d) divided by surface air temperature change. Red contour indicates the rate of precipitation increase (7.5%/°C) expected from the Clausius-Clapeyron relationship [Vecchi and Soden, 2007]. (c) Same as Figure 5b but for R5d.

[19] Figure 6 displays precipitation (PAVE, SDII, R3d, R5d, and R10d) changes averaged over East Asia as a function of surface air temperature change. Temperature changes are also averaged over the same region. For example, the increase in PAVE during the near future decade of 2030–2039 ranges from 1.6% to 4.4% for individual simulations (2.6% for the ensemble average) associated with a surface air temperature increase of about 1.1°, resulting in the precipitation efficiency rate ranging from 1.4 to 3.8%/°C for individual simulations (2.2%/°C for the ensemble average). During the last decade of 21st century (2090–2099), changes in temperature and PAVE change are much larger than during the period 2030–2039, but the precipitation efficiency rate of the ensemble average is 2.3%/°C, which is similar to that for 2030–2039. The precipitation efficiency rate of PAVE, calculated using only ensemble average values estimated by linear regression, is 2.5%/°C. This is consistent with findings from previous global warming studies that the conversion from water vapor to precipitation is less effective than that expected from the C-C relationship (7.5%/°C) between water vapor and temperature [Held and Soden, 2006; Vecchi and Soden, 2007]. In the case of SDII, the precipitation efficiency rate estimated by a similar procedure is 4.2%/°C, which is larger than that for PAVE. This indicates that conversion from water vapor to precipitation will become much more effective for intense rainfall than for moderate and weak rainfall in a future warmer climate. This occurs because convective precipitation is much more sensitive to temperature increases than is stratiform precipitation [Berg et al., 2013]. The precipitation efficiency rates measured by indices of heavy precipitation become larger in the following order: SDII (4.2), R10d (5.1), R5d (5.5), and R3d (5.7). This suggests that these indices represent much more extreme rainfall events in the order given. Therefore, extreme rainfall events will occur more frequently than moderate rainfall events, even if the amount of temperature rise is the same.

Figure 6.

Change in precipitation (%) averaged over East Asia (100–150°E, 20–50°N; Figures 1 and 3) relative to the present-day climatology 1986–2005 as a function of surface air temperature change (°C) for nine successive decadal periods (2010–2019 up to 2090–2099). The cross indicates the individual simulations. Closed circles show change calculated using three-member ensemble average for the present-day period and future decades. Black slanted line denotes the rate of precipitation increase (7.5%/°C) expected from the Clausius-Clapeyron relationship [Vecchi and Soden, 2007]. Linear regressed lines fitted only to ensemble averages (closed marks) are shown.

6.2 Dynamical Effect

[20] The thermodynamic effect alone cannot be used to determine the cause and character of future changes in precipitation amount and intensity. Dynamical effects also play an important role in the midlatitudes through horizontal advection [Meehl et al., 2005] and in the tropics through vertical motion [Emori and Brown, 2005]. Here we highlight the contribution of the combined effect of dynamical circulation change and water vapor transport on the changes in precipitation amount and intensity projected in our simulations. Figure 6a illustrates future changes in the vertically integrated annual mean water vapor flux for 2080–2099 relative to the present-day climatology (1986–2005) and its convergence. Eastward water vapor flux increases over most areas to the north of 30°N, resulting in the increase of PAVE (Figure 4d), SDII (Figure 4e), and R5d (Figure 4f) over a wide area of East Asia. Clockwise water vapor flux change is evident to the south of Japan. In addition, north-astward water vapor flux increases over the Indochina Peninsula. As a consequence, water vapor flux converges over southern China. Therefore, increase in PAVE, SDII, and R5d over southern China can be interpreted as this convergence of water vapor flux (Figure 7a). However, correspondence between the spatial distribution of convergence and divergence patterns in Figure 7a, and those of changes in Figures 4d–4f is not so evident. This is partly because we used the ensemble average from the simulations, which leads to a smoothing of spatial patterns.

Figure 7.

(a) Future change in vertically integrated annual mean water vapor flux (arrow; kg/m/s) for 2080–2099 relative to the present-day climatology (1986–2005) and its convergence (shading; mm/d). Flux is calculated from historical monthly data of specific humidity and wind. The unit of convergence is converted to mm/d assuming the density of liquid water as 1 g/cm3. Note that the displayed region is extended toward south and east by 10° to cover the subtropical high area compared with Figures 1 and 4. Red arrow denotes confidence level exceeding 95% based on Hotelling's T2 statistics [Storch and Zwiers, 1999; Wilks, 2011]. Three-member ensemble average. (b) Future changes in 500 hPa height relative to the present-day climatology averaged for latitudes 15–25°N. Target region is the area between the two green lines in Figure 7a. In each period, the longitudinal average for 100–160°E is subtracted to highlight the longitudinal gradient of height. Three-member ensemble averages are shown.

[21] The clockwise water vapor flux change found to the south of Japan (Figure 7a) is due to intensification of the subtropical high. Changes in the zonal profile of 500 hPa height over the subtropical high area are shown in Figure 7b. For the period 2046–2065 (black line in Figure 8b), the east-west height gradient is larger than that for the period 1986–2005 (blue line), and the gradient becomes even larger in the period 2080–2099 (red line). Consequently, the geostrophic balance relationship between the wind and height field leads to strengthening of the clockwise geostrophic wind. This intensification of the subtropical high can be attributed mainly to the change during summer, as is indicated by future projections focusing on summer by the CMIP3 models [Kimoto et al., 2005; Kripalani et al., 2007] and by higher horizontal resolution atmospheric models [Kusunoki et al., 2006, 2011; Kusunoki and Mizuta, 2008]. The intensification of the subtropical high can be interpreted as an atmospheric response of El Niño-like ocean change, as shown in Kusunoki et al. [2006, Figure 22]. The intensification of the subtropical high leads to the suppression of tropical cyclone generation in the area, which is consistent with the decreasing trend in the number of tropical cyclones shown by Sugi and Yoshimura [2012].

Figure 8.

Contribution of water vapor change and wind change to water vapor flux change at the 850 hPa level. Annual mean. Unit is g/kg m/s. The first member of the ensemble simulation is used. (a) Present-day climatology (1986–2005). Flux is calculated from historical monthly data of specific humidity and wind. (b) Same as Figure 8a but for future climatology (2080–2099). (c) Change; Figure 8b − Figure 8a. (d) Contribution of water vapor change: water vapor flux change from the present-day climatology (Figure 8a) calculated using present-day climatological wind and future climatological water vapor. (e) Contribution of wind change: water vapor flux change from the present-day climatology (Figure 8a) calculated using future climatological wind and present-day climatological water vapor.

6.3 Contribution of Water Vapor and Wind to Changes in Water Vapor Flux

[22] The contributions of changes in water vapor (thermodynamic effect) and wind (circulation effect) are both included in the analyses in Figure 7a. To separate the relative contributions of water vapor and wind to the change in water vapor flux precipitation, we conducted additional analyses (Figure 8). As water vapor is mainly concentrated in the lower troposphere, we selected the 850 hPa level as an example. In Figure 8d, the future vapor flux was calculated using the wind field of the present-day climatology and the water vapor of the future climatology. Then, the present-day climatology (Figure 8a) was subtracted from the calculated future vapor flux. On the other hand, in Figure 8e, the future water vapor flux was calculated using the wind field of the future climatology and the water vapor of the present-day climatology. Then, the present-day climatology (Figure 8a) was subtracted from the calculated future vapor flux. The close similarity between Figures 8c and 8d indicates that the change in water vapor mainly contributes to the change in the water vapor flux. This result is opposite to the finding shown in Kusunoki et al. [2006, Figure 23] in which the contribution of wind change is larger than that of water vapor. The difference between Kusunoki et al. [2006] and the present study may be attributed to the differences in the horizontal resolution of the models, cumulus convection schemes of the models, prescribed future SST distributions, and target seasons.

7 Summary

[23] We conducted three-member ensemble simulations using a high horizontal resolution global atmosphere model with a 60 km grid size for the period from 1872 to 2099 (228 years). For the 134 years from 1872 to 2005, the model was forced with the observed historical SST and observed concentration of greenhouse gases. For the 94 years from 2006 to 2099, the boundary SST data were estimated using the MME of the CMIP3 models. The A1B emission scenario was assumed for future concentrations of greenhouse gases. The model reproduced the observed annual precipitation and precipitation intensity of the present-day climate (1986–2005) over East Asia reasonably well. PAVE, SDII, and R5d averaged over East Asia (100–150°E, 20–50°N) increase almost monotonically during the 21st century. The amplitude of the year-to-year variability in SDII and R5d increases after the 2070s. The statistically significant area of precipitation intensity increase is larger for the period 2080–2099 than for 2046–2065. In particular, intense rainfall will increase over northern and southern China during the period 2080–2099. The conversion rate from water vapor to precipitation per 1° rise in surface air temperature for SDII and R5d is much larger than that for PAVE during the 21st century. This suggests that extreme rainfall events will occur more frequently than moderate rainfall events due to the more effective conversion from water vapor to precipitation associated with active deep convection. Future change in the horizontal transport of water vapor is also responsible for increases in intense precipitation over East Asia. In particular, the increase of clockwise water vapor transport due to intensification of the subtropical high contributes to an increase in the frequency of intense precipitation in southern China.

Acknowledgments

[24] This work was conducted under the framework of the Development of Basic Technology for Risk Information on Climate Change supported by the SOUSEI Program of the Ministry of Education, Culture, Sports, Science, and Technology of Japan. We thank anonymous reviewers for their valuable comments and suggestions, which have greatly improved the manuscript.