3.1 Observed precipitation trends and related physical processes
Figure 1 exhibits the spatial distribution of TP precipitation trends in summer during 1979–2013 based on the meteorological stations, CMFD, GPCC, and ERA5 datasets. Results from all the four datasets display predominantly positive rainfall trends over the vast western TP and northeastern part of the TP, with negative trends over the southeastern TP.
To reveal the physical processes related to the TP precipitation trends, a moisture budget analysis is performed using the ERA5 reanalysis data. The spatial patterns of the trends in moisture budget components are shown in Fig. 2. A close resemblance is observed among precipitation (Fig. 1d), the vertical moisture transport term (Fig. 2d), and the dynamic component (Fig. 2f) over the TP, while the thermodynamic component displays a wetting trend (Fig. 2e). The horizontal moisture advection terms contribute partly to the increasing trend of precipitation over the northeastern TP (Figs. 2b and 2c). The drying trend of precipitation over the southeastern TP is dominated by the dynamic component related to atmospheric circulation changes. The evaporation change features a widespread decrease over the TP (Fig. 2a) possibly owing to the combined effects of decreasing net radiation and wind speed (Yao et al. 2021; Zhang et al. 2018).
Based on the above analysis, the corresponding circulation characteristics have been investigated to reveal the underlying mechanism responsible for the observed hydrological changes over the TP. The summer-mean trends in atmospheric circulation and moisture at 500hPa during 1979–2013 in ERA5 are presented in Fig. 3a. Anomalous easterly flows are seen over the TP, accompanied by increased specific humidity. It is clear that the easterly wind anomalies are generally larger to the east than to the west of the TP. Such significant contrast in wind anomalies is favorable for moisture flux convergence over the TP (Sun et al. 2020; Zhou et al. 2019). Thus, water vapor carried by the westerlies is restrained over the TP.
Can the observed atmospheric circulation trend potentially be attributed to the observed boundary conditions over the oceans together with historical anthropogenic forcing? Indeed, much of the observed precipitation and 500hPa wind trend pattern is reproduced by the ensemble mean of the GOGA simulations (Figs. 3b and 4b). Can the observed hydroclimate trends be attributed to anthropogenic forcing? The CESM-LE ensemble mean represents an estimate of the forced response to radiative forcing. The spatial pattern of externally forced trends in precipitation and 500hPa wind (Figs. 3d and 4d) bears close resemblance to that of the ensemble mean of the GOGA (Figs. 3b and 4b) but with reduced amplitude, suggesting the combined influences of SST and anthropogenic forcing. The observed SST trend pattern resembles closely the interdecadal Pacific oscillation (IPO), and cannot be attributed to anthropogenic forcing according to the CESM-LE and other models (figures not shown). To what extent are the trends in precipitation and 500-hPa wind in GOGA EM forced by tropical Pacific SST and radiative forcing? The TPAC EM, which include both effects, also displays easterly wind anomalies over the TP but with a slightly weaker magnitude (Fig. 3c), in accordance with the weaker wetting over the central TP (Fig. 4c), compared to GOGA EM. Precipitation and atmospheric circulation responses in TPAC EM are a result of both internal variability driven by tropical Pacific SST anomalies and radiative forcing. It is noted that the simulated moisture trend (Figs. 3b-d) is much weaker than the observations (Fig. 3a), which could be related to systematic error in the model climatology over the TP (Fig. 5). The drier background in the model may limit the thermodynamic changes and in turn atmospheric circulation changes.
The above analysis suggests that the TP precipitation changes in recent decades are closely related to atmospheric circulation changes, and tropical Pacific SST changes may play an important role in the circulation changes around the TP. Hence, we will discuss the relationship between the tropical Pacific SST and TP summer rainfall changes in the subsequent analysis.
3.2 Role of tropical Pacific SST
In this section, we first examine the atmospheric teleconnection associated with the TP precipitation changes. Figure 6a shows the trend in 200-hPa eddy geopotential height in the observations. The eddy geopotential height is defined as the difference of geopotential height from its zonal mean. As can be seen in Fig. 6a, a wave train appears in the extratropics, with negative anomalies over central-eastern Atlantic Ocean and western Asia and positive anomalies over central Europe and the Lake Baikal. This large-scale wave train resembles closely the “Silk Road Pattern” (SRP). The SRP is a summer teleconnection pattern in the upper troposphere along the subtropical westerly jet over the Eurasian continent (Enomoto et al. 2003; Lu et al. 2002; Kosaka et al. 2009). This result indicates that the cyclonic circulation anomaly near the Lake Baikal is regional manifestation of SRP on the decadal time scale.
The ensemble mean of the GOGA simulations (Fig. 6b) shows a series of high pressure and low pressure systems over Eurasia but with reduced amplitude (Fig. 6a). The anomalies over North America-North Atlantic Ocean-Europe display differences from the observations. Note, however, that the anomalies around the TP are similar to the observations. In addition, the ensemble mean of the Pacemaker simulations shows a similar pattern of geopotential height trends as GOGA EM over much of the extratropical Eurasia (Fig. 6c), indicating that the atmospheric teleconnection pattern could be forced by tropical Pacific SST anomalies. Figure 6d displays the trend in 200-hPa geopotential height from the CESM-LE EM. Although the spatial pattern is similar to that of GOGA EM, the anticyclones over the Lake Baikal and the North Pacific are much weaker (Fig. 6d).
Can the tropical Pacific SST anomalies alone reproduce the observed anomalies on decadal time scales? To address this question, we compute 35 year trends from nonoverlapping segments of the 1800 year control integration of the fully coupled CESM, for a total 51 segments. Figure 7 illustrates the trend differences in SST and 200h-Pa eddy geopotential height between the segments with SST trends averaged over 5°S-5°N, 170°-90°W below one negative standard deviation and over one positive standard deviation. The SST trend difference features a cooling in the tropical central-eastern Pacific and a warming in the western mid-latitude North Pacific (Fig. 7a), which resembles the negative phase of the IPO pattern in the observations (Dong and Dai 2015). As shown in Fig. 7b, the IPO cold phase induces a series of anticyclone and cyclone systems over the Eurasian continent at 200hPa, which is similar to the observations (Fig. 6a).
Thus, these results imply that the IPO may be one of the key internal variability modes influencing the near-term simulation of TP summer rainfall. To examine whether the IPO-related SST anomalies are responsible for the uncertainty in the simulated TP rainfall trend, the CESM1 large ensembles (CESM-LE) are used. With the same external forcing, the 40 ensemble members simulate diverse TP rainfall changes, accompanied by a variety of atmospheric circulation trend patterns (figures not shown). To identify the leading modes of the inter-member spread in TP circulation changes, the MV-EOF analysis is applied to a set of five circulation variables, including the JJA-mean 500- and 200-hPa wind vector and the 500-hPa vertical pressure velocity over the TP and adjoining regions (10°-45°N, 60°-110°E).
The first and second MV-EOF modes account for 18.4% and 12.4% of the total variance, respectively. The two leading modes can be separated from each other and from other modes based on the criterion of North et al. (1982). The regression patterns of the 500-hPa atmospheric circulation anomalies onto the normalized first principal component (PC1) show two pronounced anticyclones, with centers located to the west of TP and over eastern China, respectively (Fig. 8a). The corresponding 200-hPa non-zonal geopotential exhibits an anomalous wave train over mid- and high latitudes of Eurasia, with positive (negative) anomalies over the northwestern Europe, the east of the Caspian Sea and the eastern China (the eastern European Plain-western Siberia-Mongolia), respectively (Fig. 9a). This pattern highly resembles the typical spatial pattern of the summer North Atlantic oscillation (SNAO; Folland et al. 2009).
The general characteristics of the 500-hPa atmospheric circulation anomalies associated with the EOF2 shows an anomalous cyclone located to the west of TP and an anomalous anticyclone over the eastern China (Fig. 8b). The 200-hPa circulation anomalies associated with the EOF2 display a wave train-like teleconnection pattern along the Northern Hemisphere westerly jet, with its centers located over the high-latitude North Atlantic, the central Europe, the Caspian Sea, the eastern China and the North Pacific (Fig. 9b). This zonal teleconnection pattern is similar to the typical pattern of the SRP.
Previous studies (Hong et al. 2017; Piao et al. 2017; Wang et al. 2017; Wu et al. 2016) have suggested that the AMO may play an import role in regulating the SRP variability on decadal time scales. However, both the EOF1 and EOF2 patterns appear to be significantly correlated with the IPO-like SST trend pattern (Fig. 10), indicating a potential link between different members of the IPO phase and the spread in the simulated TP summer rainfall in the CESM-LE. To investigate the respective contributions of SST anomalies over tropical central-eastern Pacific and western mid-latitude North Pacific to the large-scale atmospheric circulation, we compute the trend differences in 200-hPa geopotential between the realizations with SST trends averaged over 30°-60°N, 130°E-160°W (5°S-5°N, 170°-90°W) over positive 1 standard deviation and below negative 1 standard deviation in the CESM-LE. As shown in Fig. 11, the SNAO- and SRP-like atmospheric teleconnection patterns may be modulated by the SST trend anomalies over western mid-latitude North Pacific and tropical central-eastern Pacific, respectively.