Observed daily precipitation datasets were supplied by the National Meteorological Information Center of the China Meteorological Administration (CMA). A total of 250 stations in WC (28˚–36˚N, 100˚–108˚E, Figure 1) were selected in late summer (August) from 1961 to 2021, in which the missing stations with less than 5% of the total missing time series were estimated using the inverse distance method (Eischeid et al., 2000). The ERA5 atmospheric reanalysis data used in this study, including potential height, temperature, horizontal winds, vertical velocity, and specific humidity, were provided by the European Center for Medium-Range Weather Forecasting (Bell et al., 2021). The horizontal resolution of the variables was 0.25° × 0.25°, and the vertical layer was 1000–1 hPa with 37 layers. To investigate the combined effects of SST anomalies on the interannual variability of extreme precipitation in WC, global monthly SST data with a 1° × 1 ° horizontal resolution obtained from the Met Office Hadley Centre Sea Ice and Sea Surface Temperature (HadISST) (Rayner et al., 2003) were also used. Notably, linear trends in all datasets were removed before analysis.

Extreme precipitation indices were provided by the World Meteorological Organization (WMO) Expert Team on Climate Change Detection and Indices (ETCCDI; http://etccdi.pacificclimate.org/list_27_indices.shtml). In this study, the 99% percentile of wet days in August was selected as extreme precipitation. The total precipitation accumulation with daily precipitation exceeding the 99th percentile was defined as R99TOT, and the number of occurrences was R99FRE. The Generalized Extreme Value (GEV) distribution was applied to describe the distribution of extreme values, and the return period used to investigate the occurrence of particular extreme events was calculated using the GEV. Empirical Orthogonal Function (EOF) analysis (Lorenz, 1956) was proposed to exhibit the leading mode of variability of extreme precipitation in August over WC from 1961 to 2021. The interannual variability of extreme precipitation was calculated by removing the 9-year running mean from the original time series. Consequently, linear regression and Pearson correlation were used for statistical analysis, and statistical significance was determined using Student’s t-test.

The wave activity fluxes (WAFs) were calculated according to Takaya and Nakamura (2001) as follows: W = 1 2 U ¯ u ¯ ψ x ′ 2 − ψ ′ ψ x x + v ¯ ψ x ψ y ′ − ψ ′ ψ x y u ¯ ψ x ′ ψ y ′ − ψ ′ ψ x y ′ + v ¯ ψ y ′ 2 − ψ ′ ψ y y ′ where ψ denotes the stream function, U = u , v means the horizontal wind, and W represents the two-dimensional Rossby WAFs. The barotropic Rossby-wave source (RWS) was formulated by Sardeshmukh and Hoskins (1988):33 R W S = − ∇ ∙ V χ ξ a = − V χ ∙ ∇ ξ a − ξ a ∇ ∙ V χ where V χ and ξ a are the divergence wind and absolute vorticity, respectively.

The vertically integrated water vapor transport flux (WVT) was defined by Trenberth (1991). The vertical integrated < Q > was calculated from 1000 to 300 hPa following: < Q > = 1 g   ∫ p 0 p Q   d p

The climatological spatiotemporal characteristics of late summer precipitation and extreme precipitation in WC during 1961–2021 are shown in Figure 2. Both precipitation (Figure 2A) and extreme precipitation (Figure 2B) exhibited an inhomogeneous spatial distribution, which increased from the northwest to the southeast, and the values were higher in the western Sichuan Basin. Spatial distributions are influenced by the orography and altitude of the WC (Nie and Sun, 2021). As shown in Figure 2C, the R99FRE exceeded 10 times by more than 97% of the stations, and the stations in southeastern WC largely reached 18 times. Apart from southeastern Sichuan, increasing precipitation trends prevailed in southern Gansu, southwestern Shaanxi, and the Sichuan Basin (Figure 2D). Compared with precipitation, R99TOT and R99FRE occurred more frequently in most WC, except in the southeastern regions (Figures 2E, F).

The regional precipitation, R99TOT, and R99FRE in WC and the eastern regions during late summer are further discussed. There were no significant trends or decadal changes in precipitation and extreme precipitation indices (Figures 2G–I). However, the non-linear trends of R99TOT and R99FRE showed a transition around 2000 from a decreasing to an indistinctive increasing trend (Figures 2H, I), further implying that the variability of extreme precipitation in WC has different features.

Under the background of a slightly linear increasing trend in WC at a rate of 0.01 mm decade⁻¹. The observed rainfall in August 2020 in WC was a rare event, occurring with a return period of 60 years, based on the GEV fits (Figure 3A). The intensity reached 2.46 standard deviations (Qian et al., 2022). Furthermore, comparisons between the R99TOT anomalies reconstructed from different periods suggest that climate change is likely to increase the occurrence of extreme precipitation events, especially in recent decades (Figure 3B). In addition, the KS test in Figure 3A suggests that the observed extreme precipitation in August 2020 was not an outlier and can be explained within the context of historical records. Therefore, an in-depth understanding of extreme precipitation from large-scale circulation is crucial for predicting and mitigating disasters.

The first EOF mode of R99TOT displayed a dipole pattern with an increase in the northwestern WC and a decrease in the southeastern WC, accounting for 22.12% of the total variance (Figure 4A). The maximum positive values were located in the western Sichuan Basin and southeastern Gansu Province. PC1, which corresponds to the first mode, as shown in Figure 4B, presents an asymmetrical oscillation. The interannual and interdecadal components accounted for 88% and 12%, respectively, of the total variance in PC1. Therefore, this study further explored the atmospheric processes associated with the interannual variability in late summer extreme precipitation.

Changes in interannual late summer extreme precipitation in WC were closely linked to consistent changes in atmospheric circulation. Thus, Figure 5 illustrates the anomalous atmospheric circulation patterns regressed onto the interannual PC1 component from 1961 to 2021.

There are three notable anti-cyclonic systems in the Western Siberian Plain, Iranian Plateau, and Korean Peninsula from 200 hPa (Figure 5A) to 500 hPa (Figure 5C). Simultaneously, three cyclonic systems were located over the Caspian Sea, Mongolia, and the Chersky Mountains. The cyclonic system over Mongolia stretched southward, and the WC was sandwiched between two high-pressure centers at 500 hPa (Figure 5C). It is worth noting that the 12,500 gpm at 200 hPa and 5,880 gpm at 500 hPa isolines abnormally strengthened. Ma et al. (2022) emphasized that the eastward SAH and westward WPSH provided favorable conditions for extreme events in August 2022. Furthermore, the westerly jet at 200 hPa was also strengthened, and the WC was located on its south side (Figure 5B). In the lower troposphere, the cyclonic system over Mongolia further stretched southward, and an anti-cyclonic system over the western Pacific caused the WC to be affected by the low-pressure trough and southerly water vapor (Figure 5D). As shown in Figure 5E, abundant anomalous southerly water vapor was transported to WC and mainly converged along 30°–36°N, 104°–106°E, which is consistent with the maximum R99TOT. In addition, the regressed vertical motion enhanced significant upward vertical motion and sustained extreme precipitation in WC (Figure 5F).

Under these atmospheric circulation patterns, the SAH and WPSH tended to extend further east and west, respectively. This favored the formation of an upper-level divergence. The enhanced jet stream shifted northward, and the cyclonic system over Mongolia stretched southward, providing beneficial dynamic conditions (Yokoyama et al., 2017). Correspondingly, the anti-cyclonic system over the western Pacific guaranteed sufficient water vapor. Finally, a high R99TOT occurred in cooperation with enhanced updrafts.

To illustrate the formation of quasi-barotropic circulation anomalies, summer SST fields associated with the interannual component of PC1 are plotted in Figure 6. Figure 6A shows the SST regression of the interannual component of PC1. A significant positive SST anomaly was observed in the western Pacific Ocean. The areas with the highest relevance (at the 95% confidence level) were selected as key SST regions. The SST index (SSTI) is defined as the year-to-year regionally averaged SST over 0°–25°N, 110°–150°E. In Figure 6B, the SSTI significantly increased with 0.014°C/year (at the 99% confidence level), and the correlation coefficients with the interannual component of PC1 reached 0.48 (at the 99% confidence level). In addition, the SSTI experienced significant interdecadal changes around the mid-1980s.

To validate the speculation that increasing SST could influence interannual R99TOT by adjusting atmospheric teleconnections. Figure 7 shows the regressed circulation and precipitation against the SSTI. Figure 7A shows the regressed zonal wind at 200 hPa, the strengthened westerly jet stream, and the southern WC. At 500 hPa, three notable anti-cyclonic systems were centered over the Western Siberian Plain, the Indian peninsula, and southeast China (Figure 7B). Against the backdrop of increasing SST in the western Pacific, the SAH and WPSH also tended to strengthen. The cyclonic system over Mongolia stretched southward and placed the WC between two high-pressure centers. The anomalously strengthened anticyclone over the western Pacific at 850 hPa continuously transported water vapor into WC (Figure 7C). Simultaneously, the strong upward vertical velocity further enhanced precipitation (Figure 7D). In Figures 7E, F, the significantly increased precipitation and R99TOT further prove that the warming of the western Pacific might be closely related to precipitation and extreme precipitation in WC.

Figure 8 shows the responses of WAFs and RWS to interannual R99TOT and increased SSTI. In Figure 8A, the regressed WAF with respect to the interannual components of PC1 originated from the West Pacific, was transmitted across the Atlantic, and was enhanced over Europe before arriving and converging in the WC. Figure 8B shows the SSTI regressed WAFs and RWS. It is clear that the WAFs separated from the western Pacific, and an obvious wave train crossed the mid-high latitudes, strengthened over western Europe, and converged in the WC, which further proved that the stationary Rossby-wave train was derived from the Pacific.

In summary, the warming SST in the western Pacific could induce a barotropic Rossby-wave train, which provides beneficial conditions for extreme precipitation in WC.