4.1 Climatological temperature and moisture response
The JJA mean temperature at 850 hPa was 12–18 °C and the JJA mean specific humidity at 850 hPa was 6–11 g kg−1, from Hokkaido to Kyushu in the HIST run of the AGCM experiments (Fig. 2). Because moisture is advected from the southeast of Japan, the isopath of the specific humidity along the Japanese Archipelago crossed isotherms. Abundant moisture intruded along the western fringe of the Bonin high. The pentad temperatures averaged over the Hokkaido domain almost followed the normal distribution with an average of 13.0 °C and a standard deviation of 2.8 °C, whereas those averaged over the Kyushu domain were positively skewed with a small variation (Fig. 3a). This is probably because of the contrast between the dominant fluctuation caused by the passage of extratropical cyclones in the subarctic zone and the uniformly warm temperature that prevails in a subtropical downdraft area. Similar to the probability density function of the pentad temperature, the specific humidity followed the normal distribution of the average at 7.5 g kg−1 in Hokkaido and the positively skewed distribution of 5–15 g kg−1 in Kyushu (Fig. 3b). These climatological features in the HIST run were similar to the observed features.
The 2K run showed an increase in climatological temperature of 1.5–2 K and an increase in climatological specific humidity of 10–15% (Fig. 4a), both with a statistical significance of 5% over almost all of the target domain (not shown). Because the 1 K increase in temperature corresponds to a 7% increase in water vapor amount, this almost followed the CC relation as well. Moreover, the 4K run showed a temperature increase of 4–4.5 K over Japan (Fig. 4b). The specific humidity increased by approximately 28% in Kyushu and approximately 35% in Hokkaido, almost following the CC relation. The temperature and moisture increase exceeded the statistical significance level of 5% (not shown). This result suggested that hot, humid environment in Hokkaido in the +4 K climate was comparable to that in Kyushu in present climate (Fig. 3).
4.2 Climatological circulation and precipitation response
The d4PDF DDS simulation captured the observed precipitation over Japan in JJA well (Figs. 5, 6). Hokkaido had average precipitation of about 3 mm day−1, whereas Kyushu had an average of about 10 mm day−1. The total amount of precipitation in JJA exceeded 1000 mm along the Pacific side of the Japanese Archipelago due to the effect of topography. The summertime monsoonal flow converged abundant moisture from the southwest. Overall, the MFC showed the flux convergence over Japan, and the second term of the left-hand side of Eq. (1) was positive for the windward side around central Japan and negative on the leeward side. This was related to the persistent Baiu front in June and July, but the front did not affect the climate in Hokkaido greatly.
The summertime precipitation difference between the 4K and HIST runs is shown in the shading of Fig. 7. This result was not sensitive to the sea surface temperature anomaly imposed in the AGCM (not shown). The precipitation decreased by ~1 mm day−1 (10–20%) almost over central Japan including eastern Kyushu. However, in the 4K run, the precipitation in Hokkaido increased slightly, especially in the south, and the precipitation in western Kyushu increased by >2 mm day−1 (20–30%) (Fig. 7). The change in precipitation approximated the MFC changes (see Section 3.2), which can be divided into the effects of moisture content, climatological flow, and transient eddy activity. Based on the positive MFC on the windward side of Japan and negative MFC on the leeward side near high mountains, we explain why the precipitation increased slightly in Hokkaido, decreased in eastern Kyushu, and increased in western Kyushu.
Figure 8a displays the total MFC difference between the 4K and HIST runs. Despite a considerable error in the moisture budget around the steep slope area, the vertically integrated MFC between 4K and HIST runs decreased over central Japan including eastern Kyushu, but showed a slight difference in Hokkaido and in western Kyushu (Fig. 8a, with the same shading as Fig. 7). Figure 8b shows the stationary response of MFC, which was positive over Hokkaido and western Kyushu and negative over southern Japan including eastern Kyushu. The response was divided into the contributions from moisture change (Fig. 8c) and circulation change (Fig. 8d). Consistent with the CC relation (Fig. 4b), the 4K run provided an almost uniform increase in MFC due to the increase in humidity. A patchy decrease in the MFC over Japan was caused by the horizontal convergence/divergence pattern that emerged below 850 hPa around the mountain region. In contrast, the MFC due to circulation change decreased significantly over central Japan including eastern Kyushu, but we found a patchy increase in the MFC in western Kyushu. This result was probably related to the change in monsoonal flow direction in the 4K run (Fig. 7). The climatological summertime wind over Kyushu was southwesterly and the MFC was positive. The wind response to global warming was westerly or northwesterly over Kyushu, which strengthened the wind convergence in western Kyushu and weakened it in eastern Kyushu.
Figures 9a, c show the seasonal march of the climatological jet stream at 500 hPa averaged over longitudes of 130–140°E. The HIST run reproduced the jet stream axis in June and July, although it had a southward bias (Fig. 9a). The Baiu front was mostly located on the southern edge of the jet core in the mid-troposphere. In the observations, the Baiu front gradually moved northward from June and ended at the end of July. The jet stream tended to be shifted southward in the Baiu season in the global warming simulations (Fig. 9c), consistent with previous studies with a different model or multiple model comparison (Kawase et al. 2009; Horinouchi et al. 2019). The shift was 200 km in the 4K run and left the rainband far from the central Japanese islands. This was closely related to an eastward shift of the Bonin high (Figs. 9b, d).
In summary, the precipitation increase in western Kyushu explained the MFC increase by strengthening the horizontal wind convergence and increasing the amount of moisture following the CC relation. The precipitation decrease in eastern Kyushu explained the MFC decrease by weakening wind convergence, which surpassed the effect of moisture enrichment. However, we could not explain the slight increase in precipitation over Hokkaido in the limited analysis of the climatological field.
4.3 Effect of transient eddies
Figure 10 shows HIST’s climatology and the difference between the 4K and HIST runs. The difference between the total MFC and the MFC caused by the climatological field (Fig. 10d) is the MFC caused by transient fluctuations, including extratropical cyclones and TCs, which was negative over Hokkaido and positive off the Pacific coast of Japan. However, in Hokkaido, this transient response had the opposite sign to the stationary response, so that the response of the total MFC (Fig. 8a) was slightly weaker than that of the MFC caused by the climatological field (Fig. 8b). In the HIST run, storm activity estimated with the band-pass-filtered meridional wind variance reached its maximum around 50°N (Fig. 10b). The difference between the 4K and HIST runs decreased significantly, presumably related to the southward shift of the jet stream (Figs. 9, 10a). The northward moisture flux caused by band-pass eddies also decreased in response to global warming at 35–40°N. This decrease weakened the eddy moisture flux to the north, and the MFC difference caused by transient eddies was negative over Hokkaido and positive over the south of Japan (Fig. 10d). In summary, the band-pass-filtered eddies like extratropical cyclones contributed to the decrease in MFC over Hokkaido, which counteracted the increase in MFC caused by the moisture enrichment following the CC relation.
The transient eddies contained TCs, some of which approached Japan along the western fringe of the Bonin high. The number of TCs, including the stage after extratropical transition, approaching Hokkaido was 0.4–1.0 per JJA in the HIST run, which was consistent with the observations. The contribution of TCs to the total precipitation in JJA was 1–3% in Hokkaido (Fig. 11a). The percentage was higher in eastern Hokkaido, whereas Kyushu had more than two TCs per JJA on average in the HIST run, and the TC effect was greater in eastern and southern Kyushu (Fig. 11b). Because the TCs were much more intense when they passed near Kyushu, the contribution to the total precipitation exceeded 7% in eastern Kyushu and was ~3% even in western Kyushu (Fig. 11b). This west–east contrast has been already reported by Tamaki et al. (2018) and Kawase et al. (2019).
Overall, the 2K and 4K runs showed an increase in the JJA precipitation over Hokkaido and western Kyushu. The number of TCs approaching Hokkaido per JJA was 0.4–0.8 in the 2K run and <0.5 in the 4K run, and this decrease was consistent with Yoshida et al. (2017). Despite the decrease in the number of TCs, the rate of change of TC-related precipitation remained almost constant in southern Hokkaido and increased toward the north in the 2K and 4K runs (Figs. 12a, c). The increase in precipitation reached 50% in the northernmost part of Hokkaido. This is because a single TC brought precipitation of 11.4 mm/day over Hokkaido in the 4K run, much larger than precipitation from a single TC at 6.9 mm/day in the HIST run (Fig. 13). In contrast, the TC-related rainfall decreased substantially, by 10–20% in the 2K run and >30% in the 4K run over Kyushu. The number of TCs approaching Kyushu decreased to 1.5–2 in the 2K run and 0.5–1 in the 4K run (Figs. 12b, d), whereas the precipitation from a single TC increased by 20% in the 2K run and by 47% in the 4K run following the increase in temperature (Fig. 13). Although the TCs passing through Kyushu became more intense, the TC contribution to total precipitation decreased because the TCs were less frequent. The decrease in total precipitation in eastern Kyushu (Fig. 7) was partly attributed to the decrease in the frequency of TCs passing through Kyushu. However, the decrease in TC frequency was not the primary reason why the total precipitation in JJA decreased, even in eastern Kyushu, because the contribution of TCs to the total precipitation was only 10% (Fig. 11). Instead, the primary reason was the climatological wind response to global warming (Figs. 7, 8d) associated with the eastward shift of the Bonin high (Figs. 9b, d).