Future changes in summertime precipitation amounts associated with topography in the Japanese islands
Disaster Risk Research Unit, National Research Institute for Earth Science and Disaster Prevention, Tsukuba, Japan
Corresponding author: N. Tsunematsu, Disaster Risk Research Unit, National Research Institute for Earth Science and Disaster Prevention, 3-1, Tennodai, Tsukuba, Ibaraki, 3050006, Japan. (email@example.com; firstname.lastname@example.org)
 This study investigates future changes in summertime precipitation amounts over the Japanese islands and their relations to the topographical heights by analyzing data from 20 km resolution regional climate model downscalings of MIROC3.2(hires) 20C3M and Special Report on Emission Scenarios A1B scenario data for the periods of 1981–2000 and 2081–2100. Results of the analyses indicate that future increases in June-July-August mean daily precipitation amounts are noticeable in the west and south sides (windward sides) of the mountainous regions, especially in western Japan where heavy rainfall is frequently observed in the recent climate. The large precipitation increases are likely to occur not only in high altitude areas but also at low altitudes where many urban areas are located. In such areas, the occurrence frequencies of precipitation amounts greater than 100 mm d−1 would also increase under the future climate scenario (A1B). One of the main causes of these precipitation changes appears to be the intensification of southwesterly moist air flows in the lower troposphere, which is likely to be associated with future increases in the north-south atmospheric pressure gradient, especially at latitudes south of 35°N. The intensified southwesterly moist air flows that impinge on the western and southern slopes of the mountains can generate stronger upslope flows and well-developed clouds, leading to increased precipitation. In contrast, future precipitation changes in the lee sides of the mountainous regions would be comparatively small. These results indicate that future precipitation changes strongly depend on the topography and prevailing wind direction.
 General circulation model (GCM) projections indicate that future global warming leads to increases in mean and extreme precipitation amounts in East Asia. Shiogama et al.  analyzed 10-member ensemble runs of an atmosphere-ocean coupled GCM (MIROC3.2) [K-1 Model Developers, 2004] for the periods 1951–1970 and 2011–2030. They showed future increases in mean and extreme precipitation at higher latitudes including Japan and in the tropics and decreases in the subtropics. This spatial pattern is consistent with long-term predictions for the periods up to 2100, shown in Meehl et al. . Kusunoki et al.  showed that rainfall caused by the Meiyu/Baiu front [e.g., Ninomiya and Murakami, 1987] increases over the Yangtze River valley, the East China Sea, and western Japan under the future climate (2080–2099) using a 20 km mesh GCM developed at the Japan Meteorological Agency (JMA) and the Meteorological Research Institute (MRI) of JMA. Using long-term climate scenarios simulated by MIROC3.2, Kimoto et al.  suggested that mean precipitation in the future (2071–2100) over Japan increases more than 10%, especially in warm seasons, indicating future increases in the frequency of heavy rainfall exceeding 30 mm d−1. These projections agree that summertime precipitation amounts in the Japanese islands increase under future climate conditions.
 More detailed spatial distributions of future precipitation increases in Japan have been investigated mainly by the use of regional climate models (RCMs). The RCMs allow greater topographic realism and finer-scale atmospheric dynamics to be simulated, thereby better representing precipitation with their higher spatial resolution [Monette et al., 2012]. Kurihara et al.  conducted numerical simulations for long-term periods (1981–2000 and 2081–2100) using a 20 km mesh RCM (RCM20 developed at MRI) with boundary conditions produced by a double-nesting technique for a 280 km mesh coupled GCM (MRI-CGCM2) [Yukimoto et al., 2006] and a 60 km mesh RCM (RCM60 developed at MRI). They showed that the future precipitation increases can be recognized over western Japan during summer. Yasunaga et al.  showed that the frequency of rainfall greatly increases with its intensity in July in the Meiyu/Baiu frontal zone in a warmer climate by performing climate simulations with a non-hydrostatic model-based 5 km mesh RCM developed at JMA and the lateral boundary conditions derived from the 20 km mesh GCM. Moreover, they showed that precipitation systems with an area larger than 90,000 km2 are more frequently seen in July in a warmer climate, especially in the vicinity of Kyushu. Iizumi et al.  indicated high probabilities of increases in heavy rainfall in spring and summer over Japan in the period of 2081–2100, relative to 1981–2000, using results of simulations by three RCMs (NHRCM, NRAMS, and T-WRF; see section 2) nested into MIROC3.2 for Special Report on Emission Scenario (SRES) A1B.
 Annual precipitation amounts exceed 1000 mm in almost all parts of the Japanese islands. In the Kyushu, Shikoku, and Kii districts (Figure 1a), there are areas where annual precipitation amounts reach 4000 mm. Annual maximum daily precipitation amounts in Japan are usually recorded in summer when the Meiyu/Baiu front formation and tropical cyclones promote the northward advection of warm moisture-laden air masses originating from subtropical areas. Also, high-altitude areas occupy the greater part of the Japanese islands as shown in Figure 1a. The topographical heights rapidly change between the highlands and plains, forming steep slopes. The warm moisture-laden air masses ascend the steep slopes as strong upslope flows after reaching the Japanese islands. The strong upslope flows, which induce well-developed convective cloud formation, lead to heavy orographic rainfall [e.g., Lin, 2007]. The heavy orographic rainfall events sometimes cause large-scale flood disasters. For instance, in the summer of 2000, the Tokai district including the Nagoya metropolitan area (see Figure 1a) experienced a historical heavy rainfall event [e.g., Ushiyama and Takara, 2002], which caused rivers to overflow and several levees to break. The government of Japan estimated the economic damage due to this flood disaster to be approximately 270 billion yen (2.5 billion dollars). Recently, abnormally high levels of torrential rain induced by the orographic effect fell in the mountainous regions of Northern Kyushu in July 2012, causing extensive flooding.
 Recently, Ishizaki et al. [2012b] analyzed the relationship between precipitation amounts and topography in the Japanese islands in the past 20 years by the use of re-analysis data and simulation results of three RCMs (NHRCM, NRAMS, and T-WRF; see section 2). Their results demonstrated that precipitation amounts over mountainous regions are sensitive to topography. The simulated precipitation amounts over land in the Kyushu district reflect orographic precipitation and can be attributed to more realistic representation of topography in the RCMs. However, there are few studies on the relationship between “future” precipitation changes and the topographical heights in Japan. Future increases in the total precipitation amounts in the mountainous regions (high altitude regions) are likely to cause larger discharges not only to the upper reaches of rivers but also to the lower reaches, which can enhance the risk of large-scale flooding. Higher frequency of heavy rainfall in the plains (low-altitude regions) including many urban areas would enhance the risk of inundation disasters inside levees and urban flooding. Investigations of the geographical distribution of future precipitation increases and its relation to the topographical heights, therefore, contribute to the assessment of vulnerability and adaptation to climate change in water hazard.
 This study thus investigates future changes in summertime precipitation amounts associated with topographical heights in the Japanese islands using data from climate simulations performed by three different 20 km resolution RCMs. For instance, Gao et al.  indicated that higher resolution simulations by RCMs do not only show more detailed geographic features and topographical forcing on circulation and moisture flux, but they also provide a much better reproduction of precipitation patterns over East Asia. They suggested that high-resolution models are needed to better investigate future climate projections over the region.
2 Data and Models
 To obtain the geographical distributions of simulated daily precipitation amounts in Japan during the periods of 1981–2000 (hereafter “recent climate”) and 2081–2100 (future climate), we analyzed results of long-term numerical simulations performed by three RCMs: Non-Hydrostatic Regional Climate Model (NHRCM) [Saito et al., 2006; Ishizaki and Takayabu, 2009; Ishizaki et al., 2012a], Regional Atmospheric Modeling System V 4.3 [Pielke et al., 1992] modified by National Research Institute for Earth Science and Disaster Prevention (NRAMS) [Dairaku et al., 2008a, 2008b], and Weather Research and Forecasting model [Skamarock et al., 2008] V 3.1.1 modified by the University of Tsukuba (T-WRF) [Kusaka et al., 2012a, 2012b]. Each simulation was carried out with a 20 km horizontal grid resolution, as part of the Japanese research project of Multi-Model Ensembles and Downscaling Methods for Assessment of Climate Change Impact (S-5-3) [e.g., Ishizaki et al., 2012a].
 The twentieth century climate (20C3M) and SRES A1B scenario data, produced by MIROC3.2(hires), were used as the boundary conditions of each simulation through dynamical downscaling that has been applied to many previous studies on climate change and its influence on atmospheric phenomena [e.g., Sato et al., 2007; Tsunematsu et al., 2011]. The MIROC3.2(hires) model has 23 vertical levels and T106 horizontal resolution (approximately 120 km). Note that only one driving GCM (MIROC3.2(hires)) had been used for carrying out the three RCM simulations. Future changes in atmospheric circulation, which would be strongly associated with atmospheric pressure changes, somewhat differ between MIROC3.2(hires) and the other GCMs [Meehl et al., 2007]. Differences in atmospheric circulation would influence spatial patterns of precipitation. The MIROC3.2(hires) projection for the A1B scenario shows that the summertime (June-July-August; JJA) sea level pressure (SLP) pattern in the vicinity of Japan is similar to that of the CMIP3 (Coupled Model Intercomparison Project phase 3) multi-model ensemble mean, although north-south gradients of the SLPs are larger than the multi-GCM mean [Meehl et al., 2007, supplementary material]. Hence, the geographical distributions of future changes in JJA precipitation simulated by the three RCMs are expected to be qualitatively similar to the case where the multi-GCM mean data are used as the boundary conditions. However, the larger north-south SLP gradients in the MIROC3.2(hires) projection could cause larger water vapor fluxes and then increase JJA precipitation in Japan.
 Calculation domains of all the RCM simulations cover the whole area shown in Figure 1a. The location and elevation of each grid point differ slightly between the RCMs owing to the different grid projections and the smoothing for the common topographic data (GTOPO30) [U.S. Geological Survey, 1996]. The differences in topographical heights between the RCMs can be seen in Figure 1 of Ishizaki et al. [2012a]. In addition to the differences of the grid projections and the smoothing for the topography, the number of grid points, initial times of numerical simulations, and parameterizations used in each RCM is specified in Table 3 of Iizumi et al.  and Table 1 of Ishizaki et al. [2012b].
 We evaluated precipitation biases of the RCMs in the JJA period because the evaluation of model biases can be useful for the discussion of precipitation changes simulated by each model. In this study, the biases in precipitation amounts were estimated using surface observation data obtained from Automated Meteorological Data Acquisition System (AMeDAS) and a model bias detection program based on Tanaka et al. . The AMeDAS is a fine-resolution surface meteorological observation network established by JMA, which has about 1300 observation sites over the Japanese islands at intervals of approximately 17 km (Figure 1a). In the bias detection program, the simulated precipitation amounts are averaged for all model grid points corresponding to each region that divides the Japanese islands into seven areas presented in Figure 1b. The observed precipitation amounts are averaged for all AMeDAS sites corresponding to each region after seeking the AMeDAS site nearest to each model grid point. The averaged values of the simulated precipitation amounts are then divided by the observed values to calculate the model biases.
 Figure 2 shows matrices of the model biases in JJA mean and 95th percentile precipitation amounts for the seven regions in Japan. The T-WRF model overestimates the precipitation in northern and eastern Japan and Okinawa (Figure 2a). The NHRCM also tends to overestimate the precipitation in Japan except for northern Japan. As for eastern and western Japan, the NRAMS mean biases are less than 10%. In all RCMs, the 95th percentile precipitation amounts are largely underestimated (Figure 2b), although the NHRCM biases are comparatively small. Each RCM might not well simulate heavy rainfall amounts associated with orographic effects in the Japanese islands because of the topographic smoothing. This can be one of the main causes of the large underestimation of 95th percentile precipitation.
3 Analyses Results of Future Changes in Precipitation
 The GCM (MIROC3.2(hires)) simulation results indicate that the JJA mean daily precipitation amounts in the Japanese islands are 2–12 mm in the recent climate (Figure 3a) and 4–14 mm in the future climate (Figure 3b). Although the relationship between the JJA precipitation and the topography is unclear in the GCM simulations due to the coarser spatial resolution, the three RCMs (Figure 4) simulate large precipitation amounts at the west and south sides of the mountainous regions, especially in western Japan (see the topographical heights in Figure 1a). Rainfall distributions similar to those in Figure 4 can be recognized in all RCM projections for the future climate scenario (figures not shown).
 Figure 5 shows differences in the JJA mean daily precipitation amounts between the future climate and the recent climate. We conducted the Lepage test [Lepage, 1971] for analyzing the differences in the simulated JJA precipitation amounts between the recent and future climate scenarios. The Lepage test is a nonparametric test that can investigate significant differences between two samples, even if the distributions of the parent populations are unknown, like precipitation amounts [Kwon et al., 2005]. We therefore consider that the Lepage test is more suitable for detection of precipitation changes than the Student's t test. If the Lepage statistic is greater than 5.99, the mean change between the two samples is significant at a 95% confidence level. For instance, Yonetani  used the Lepage test for analyzing the geographical distributions of changes in annual precipitation between two different 25 year periods. In this study, each sample consists of simulated daily precipitation data on 1840 days (92 days in JJA times 20 years).
 All RCMs show future increases in JJA mean precipitation in the greater part of the Japanese islands (Figures 5a–5c). The future precipitation increases in the west and south sides of the mountainous regions where precipitation amounts are large in the recent climate (Figure 4) tend to exceed 3 mm d−1, but are comparatively small in the east and north sides of the mountains, e.g., the Kanto district. Also, the precipitation increases greater than 5 mm d−1 are noticeable in the west and south sides of the mountainous regions in the Kyushu, Shikoku, Kii, and Tokai districts. In some parts of Western Kyushu, the precipitation increases exceed 10 mm d−1 in all the RCM projections.
 The model grid points where the future increases in the JJA mean daily precipitation exceed 3 mm and 5 mm are shown in Figure 6 after dividing the topographical heights at every grid point into several elevation zones at an interval of 300 m. In the west and south sides of the mountainous regions, the precipitation increases of more than 3 mm d−1 can be seen not only in high-altitude areas but also at low altitudes below 300 m above mean sea level (amsl) (Figures 6a–6c). Note that the precipitation increases exceeding 5 mm d−1 are widely distributed at the low-altitude areas in the western part of Kyushu (Figures 6d–6f).
 Figure 7 shows east-west vertical cross sections of the future changes in the daily precipitation amounts and the topographical heights in the Kyushu district. The precipitation amounts and topographical heights were averaged in the range 32.5°N–33.0°N, as denoted by the box with thick broken lines in Figure 1a. In the NHRCM and T-WRF projections (Figures 7a and 7c), future increases in the precipitation amounts are less than 1 mm d−1 in the eastern slope of the Kyushu Mountains (Figure 1) but exceed approximately 5 mm d−1 in the west side of the mountaintop. The largest precipitation increases exceeding 10 mm d−1 can be recognized in the vicinity of downtown Kumamoto where the topographical heights are lower than 300 m. Also, the precipitation increases in the vicinity of downtown Nagasaki reach 8–10 mm d−1.
 The NRAMS shows a similar spatial pattern of the future precipitation increases (Figure 7b), although the precipitation increases in the vicinity of downtown Nagasaki are larger than those in downtown Kumamoto. However, the precipitation increases in the eastern slope of the mountains are relatively large, compared with the NHRCM and T-WRF projections (Figures 7a and 7c). We speculate that this is due to the differences in the topographic smoothing mentioned in section 2. The NRAMS simulation adopted the wavelength filter in topographic smoothing [Iizumi et al., 2011]. The lower weighting function in the RAMS topography smoothing routine, which is favorable to the model stability, has the effect of reducing the topographical heights [Papineau et al., 1994]. Hence, the average height of the Kyushu Mountains in NRAMS is lower than the top heights in the other two RCMs by approximately 150 m. There is a possibility that the lower height of the mountains weakened the “rain shadow” effect [e.g., Whiteman, 2000], as discussed in section 4, resulting in the larger precipitation in the eastern slope of the mountains under the future climate.
 Concerning the relationship between the future precipitation changes and the topography in each RCM simulation, another east-west vertical cross section including the Nagoya and Yokohama metropolitan areas is shown in Figure 8. In all RCMs, the future increases in the precipitation are noticeable in the west side of the Japanese Alps (Figure 1) where downtown Nagoya is located. The maximum value of the future precipitation increases is nearly 8 mm d−1 (NRAMS; Figure 8b). In contrast, the precipitation increases in the east side of the top of the Japanese Alps including downtown Yokohama are less than approximately 3 mm d−1. The NHRCM shows future “decreases” in the precipitation amounts at the eastern end of the mountains (Figure 8a). The topographic smoothing method adopted in the NRAMS appears to cause the slightly larger precipitation increases in the east side of the mountaintop (Figure 8b) in comparison with the other two RCMs (Figures 8a and 8c). The large precipitation increases in the west side of the mountaintop and the smaller precipitation changes in the east are the same as the spatial pattern of the precipitation changes in the Kyushu district (Figure 7).
 In addition to the JJA mean precipitation amounts (Figures 4-8), we investigated the geographical distributions of heavy rainfall frequencies, using output data from the RCM simulations. Figure 9 shows the occurrence frequencies of the simulated precipitation amounts exceeding 100 mm d−1 in JJA in the recent climate and also future changes in the frequencies. All RCMs show high frequencies of heavy rainfall in the western part of Kyushu, the southern part of Shikoku, Kii, and Tokai in the recent climate (Figures 9a–9c). This corresponds to the geographical distributions of large increases in the average daily precipitation amounts between the recent and future climate scenarios (Figure 5). Figures 9d–9f indicate that future increases in heavy rainfall frequencies are likely to be noticeable in the same areas, i.e., the west and south sides of the mountainous regions, especially in western Japan.
 Regarding the NHRCM simulation, the frequency of precipitation amounts exceeding 100 mm d−1 under the recent climate is in the range of approximately 3–10% in almost all parts of the western part of Kyushu, the southern part of Shikoku, Kii, and Tokai (Figure 9a). Large future increases in the frequencies (>3%) are widely distributed in the western part of Kyushu, the southeastern part of Shikoku, the western part of Kii, and Tokai (Figure 9d). The maximum value of frequencies of the precipitation amounts greater than 200 mm d−1 in those areas is approximately 1% under the recent climate, and its future value reaches 3% (figures not shown). In contrast, the frequencies of precipitation greater than 100 mm d−1 are less than 3% in the east and north sides of the mountainous regions under the recent climate (Figure 9a). Future changes in the frequencies in almost all parts of the east and north sides of the mountainous regions are less than approximately 1% (Figure 9d). The NRAMS and T-WRF simulations also show a similar spatial pattern for heavy rainfall frequencies (Figures 9b, 9c, 9e, and 9f), but absolute values of the frequencies are relatively small, compared to the NHRCM simulation. The differences in absolute values of the heavy rainfall frequencies in the recent climate between NHRCM and the other two RCMs reflect the model biases in the 95th percentile precipitation shown in Figure 2b, which indicates less underestimation of the 95th percentile precipitation in the NHRCM simulation. Iizumi et al.  mentioned that one possible reason for the difference in simulated precipitation between NHRCM and the other two RCMs is a modified version of the Kain-Fritsch scheme [Kain and Fritsch, 1990], which is only used by NHRCM. The Kain-Fritsch scheme is one of cumulus parameterization schemes. Different versions of the Kain-Fritsch scheme are used in the NRAMS and T-WRF simulations. The topographic smoothing is another possible reason for it. The topography in NHRCM is closest to the real topography [Ishizaki et al., 2012a, Figure 1].
 The summertime heavy rainfall events in the Japanese islands are usually induced by the northward transport of warm moisture-laden air masses originating from subtropical areas, as mentioned in section 1. Figure 10 depicts the JJA mean water vapor fluxes (WVFs) and wind vectors at the 850 hPa pressure level in the recent and future climate scenarios, simulated by the RCMs. The moisture-laden air masses can be recognized as the high WVFs. All RCMs predict future increases in the WVFs in almost all parts of the Japanese islands, which should basically reproduce the driving GCM pattern. Although the future changes in the WVFs in the Kanto and Hokkaido districts are comparatively small, the WVFs in other districts in the future climate are approximately one-and-a-half times as large as those in the recent climate.
 As shown in Figure 10, southwesterly winds prevail over Japan in the JJA period. Judging from the geographical pattern of the simulated SLPs presented in Figures 11a–11c, the southwesterly winds prevail along the northwestern rim of North Pacific High. Absolute values of the WVFs are influenced by the southwesterly wind velocities that are basically proportional to the pressure gradients. Figure 10 indicates that future increases in the wind velocities are noticeable in the areas where the WVFs increase markedly. Therefore, the future changes in the WVFs would depend on the changes in the pressure gradients around the Japanese islands. Figures 11d–11f show differences in the JJA mean SLPs between the future climate and the recent climate. These figures indicate future increases in north-south gradients of SLPs, especially at latitudes south of 35°N. The larger SLP gradients can encourage the moisture-laden air masses that intrude into the Japanese islands with the southwesterly winds (Figures 10d–10f). Also, all three RCM projections indicate future increases in specific humidity over Japan, influenced by the GCM projection (figures not shown). This is one of the main causes of the future increases in the WVFs. The future increases in the values of specific humidity and the wind velocities averaged in the range of 26°N–36°N and 126°E–146°E are estimated as 2.46 g kg−1 and 0.98 m s−1, respectively, for the NHRCM simulation, 3.00 g kg−1 and 0.87 m s−1 for NRAMS, and 2.70 g kg−1 and 0.95 m s−1 for T-WRF.
 Future increases in WVFs in the lower troposphere, which are associated with the increases in the north-south pressure gradient, wind velocity, and water vapor content around Japan (Figures 10 and 11), are likely to be an important factor of the large increases in the precipitation in the west and south sides of the mountainous regions (Figures 5-9). Recently, Kusunoki and Arakawa  indicated future increases in precipitation intensities in almost all regions of East Asia using the CMIP3 multi-GCM ensemble predictions. They showed that the changes in precipitation climatology and daily precipitation intensities can be interpreted as moisture convergence changes associated with changes in horizontal transport of moisture. According to Kurihara et al. , the intensification of the anticyclonic circulation of the subtropical high pressure system to the south of Japan (North Pacific High) in future climate will induce a strong water vapor flux along the rim of the anticyclonic anomaly. They indicated that the strong water vapor flux will converge over the western part of Japan from the southwest and may bring about increased precipitation around western Japan. The future increases in the north-south pressure gradient (Figures 11d–11f) are considered to be caused by the anticyclonic anomaly. Kanada et al.  also presented the large westward protrusion of the subtropical high to the south of Japan in the future climate.
 It is widely known that moist flows ascending a mountain barrier (upslope flows) will typically enhance precipitation along the windward slope of that barrier and that the amount of precipitation that falls is related to the magnitude of the upslope flows [Neiman et al., 2002]. Over a meso-α/β (200–2000 km/20–200 km scales) or large-scale mountain range, precipitation is triggered or enhanced on the windward slope of a prevailing wind due to orographic lifting on the upwind slope [Lin, 2007]. Also, when a moist airflow impinges on a mountain, the dynamical and cloud microphysical characteristics of the airflow are modified by orographic lifting and blocking which may modify and/or trigger cloud and precipitation systems in the vicinity of the mountain [Lin, 2007]. The future increases in the southwesterly moist air flows intruding into Japan (Figures 10d–10f) would invigorate these dynamical mechanisms. The upslope flows driven by the intensified moist air flows can lead to generation of stronger updrafts and formation of well-developed clouds in the western and southern slopes of mountains, resulting in large precipitation increases in the west and south sides of the mountainous regions (Figures 5-9).
 The RCM projections indicate that the maximum rate of future increases in the total number of heavy rainfall days in the JJA period could reach approximately 10% in the west and south sides of the mountainous regions in western Japan (Figure 9). Furthermore, Figures 6-8 indicate that large precipitation increases will occur not only at high altitudes on the windward slopes of mountains but also at low altitudes in the upwind sides where many urban areas are located. Figure 12 shows the topographical heights at every model grid point where future increases in simulated heavy rainfall (>100 mm d−1) frequencies are greater than 1%. This figure indicates that future increases in heavy rainfall frequencies are relatively large at lower altitudes rather than high altitudes, in all RCMs. These results suggest possible future increases in the risk of large-scale floods including inundations inside levees of urban areas, especially in the west and south sides of the mountainous regions.
 In contrast, the small precipitation changes and the lower frequencies of heavy rainfall in the eastern slopes and the east side of mountains (Figures 5-9) are considered to be due to the rain shadow effect. As westerly airflow descends the lee sides of the mountain barriers, it warms and dries, creating a rain shadow on the east side of mountain ranges [Whiteman, 2000]. On the lee side of the mountain, there is little or no rain due to the depletion of moisture over the upwind slope and the adiabatic warming associated with the descending air [Lin, 2007]. The influence of mountains on climatological precipitation is orographic precipitation enhancement and suppression in the windward and lee sides of mountains, respectively. For instance, the rain shadow effect appears to prevent the WVFs and precipitation from increasing in the Kanto district in the future climate (Figures 5-10). If this is the case, then future changes in precipitation in the Kanto district including the Tokyo metropolitan area might largely be influenced by artificial variations, such as land use changes and not natural climate variability, although this is only one possibility.
5 Summary and Conclusion
 Future changes in the JJA precipitation amounts associated with the topography in the Japanese islands were investigated using output data from three different 20 km resolution RCMs (NHRCM, NRAMS, and T-WRF) driven by MIROC3.2(hires) 20C3M and SRES A1B simulations for the periods of 1981–2000 and 2081–2100. Prior to discussing the analysis results of the future changes in the JJA precipitation, we evaluated the model biases of the RCMs. The NHRCM and T-WRF tended to overestimate the JJA mean precipitation in the Japanese islands, whereas the NRAMS mean biases were comparatively small. The model biases in the 95th percentile precipitation were underestimated in all RCMs.
 Differences in the JJA mean daily precipitation amounts between the recent climate (1981–2000) and the future climate (2081–2100) in all the RCM simulations showed that the future increases in the precipitation are noticeable in the west and south sides of the mountainous regions in the Kyushu, Shikoku, Kii, and Tokai districts where heavy rainfall is frequently observed in the recent climate. In some parts of Western Kyushu, the future daily precipitation increases exceed 10 mm. In these areas, the noticeable precipitation increases would appear not only at high altitudes but also at low altitudes below 300 m amsl where many urban areas are located. Furthermore, the occurrence frequencies of the precipitation amounts greater than 100 mm d−1 would increase markedly in those areas under the future climate scenario. In contrast, the precipitation increases are comparatively small in the east and north sides of the mountains. For instance, in the Kanto district, the precipitation changes are relatively small, especially in the NHRCM and T-WRF simulations. The RCMs thus simulated large precipitation increases at the west and south sides of the mountainous regions, especially in western Japan, although the relationships between the precipitation changes and the topography were unclear in the GCM simulations, owing to the coarser spatial resolution that cannot resolve the topography in Japan sufficiently.
 All the RCM projections showed future increases of WVFs in the lower troposphere in the greater part of the Japanese islands, excluding the Kanto and Hokkaido districts. This is likely related to the future increases in the north-south atmospheric pressure gradients at latitudes south of 35°N, which would intensify velocities of the southwesterly winds prevailing along the northwestern rim of North Pacific High and intruding into the Japanese islands. The future WVF increases are considered to be an important factor of large precipitation increases in the windward sides, i.e., the west and south sides of the mountainous regions, because the intensified southwesterly moist air flows that impinge on the western and southern slopes of the mountains can generate stronger upslope flows and well-developed clouds, leading to the larger precipitation amounts.
 These results indicate that future precipitation changes strongly depend on the topography and prevailing wind direction. Also, the results of this study suggest possible future increases in the risk of large-scale floods that occur in the west and south sides of the mountainous regions, especially in western Japan. Precipitation increases at the low-altitude regions including the urban areas could increase inundation disasters inside levees and urban flooding.
 The three RCM simulations had been performed using only one driving GCM (MIROC3.2(hires)), as mentioned above. We would like to carry out long-term RCM simulations driven by other GCMs in future studies for more investigations of future precipitation changes over the Japanese islands and their relations to the topography. Also, results of detailed analyses of the precipitation changes in the other seasons will be shown in future studies.
 This study was conducted as part of the research subject “Vulnerability and Adaptation to Climate Change in Water Hazard Assessed Using Regional Climate Scenarios in the Tokyo Region” of Research Program on Climate Change Adaptation (RECCA) and was supported by the SOUSEI Program, funded by Ministry of Education, Culture, Sports, Science and Technology, Government of Japan. We thank the regional climate modeling groups (MRI/NIED/Univ. Tsukuba) for producing and making available their model output. Their work was supported by the Environment Research and Technology Development Fund (S5-3) of the Ministry of the Environment, Japan.