The performance of the CMIP6 Hist-all runs in simulating the climatological mean winter precipitation over Southeast Asia during 1979–2014 is first investigated (Figure 1). In the observation, the distribution of the winter SEAP displays a south-north dipole structure, with more (less) precipitation located east of the Philippines, west of Borneo, over the Indonesia and over the Sumatra (over the Indochina Peninsula and over the northern Philippines) (Figures 1A, B). The CMAP shows a similar pattern to GPCP but with larger (smaller) amount over the eastern Indonesia (east of the Philippines). The spatial correlation between GPCP and CMAP is 0.89 (significant at the 99% confidence level). The Hist-all MME reproduces well the dominant features of the observed SEAP, while the precipitation amount over the eastern Maritime Continent in the model is larger than that in the observation (Figure 1C). The pattern correlation coefficients between the observation in GPCP and CMAP and historical MME simulation over Southeast Asia (10°S–20°N, 90°–140°E) are .85 and .86, respectively, both of which exceed the 99% confidence levels based on the two-tailed Student’s t-test. It indicates that the Hist-all MME can simulate the observed SEAP climatology reasonably.

The distribution of precipitation trends over Southeast Asia during 1979–2014 derived from the observation and simulations is presented in Figure 2. In the observation, significantly positive trends of winter precipitation are observed over most areas of Southeast Asia. They are primarily located around the Philippines, over the South China Sea, the eastern Indochina Peninsula, the Malay Peninsula and the northwestern Sumatra, whereas negative trends are distributed over the southern Indonesia (Figures 2A, B). The precipitation trends are consistent between GPCP and CMAP with the pattern correlation coefficient 99% significant at 0.49. However, there is a conflict of precipitation trends around the Sulawesi between these two gridded datasets. It is a negative trend in GPCP and a positive trend in CMAP. Although the Hist-all MME simulates smaller magnitude of the precipitation trends over Southeast Asia compared with the observations, the spatial pattern of the SEAP trends is well reproduced in the models (Figure 2C). The spatial correlations between the MME simulation and observations are 0.34 and 0.41 for GPCP and CMAP, respectively, both of which are 95% significant even if the effective degrees of freedom are considered (Bretherton et al., 1999). It seems that the historical all-forcing MME simulates a larger reproducibility of the SEAP characteristics in CMAP than those in GPCP. In addition, there are 29 out of 58 realizations (50%) of the Hist-all runs show the same trend pattern over Southeast Asia as those in the historical MME with the spatial correlations exceeding the 90% significance levels, indicating the reliability of precipitation trends simulated in historical all-forcing experiments.

To identify the relative contribution of nature and anthropogenic forcing to the winter precipitation trends over Southeast Asia, the simulated trends in response to Hist-Nat and Hist-Ant forcing are respectively shown in Figure 3. No significant precipitation trends are simulated over the most areas of Southeast Asia in Hist-Nat MME except those significantly negative trends over the southern Indonesia (Figure 3A). In contrast, the Hist-Ant forced pattern, largely resembling that in response to Hist-all forcing (Figure 2C), is characterized by significant increases of precipitation trends over the South China Sea, the eastern Indochina Peninsula and the northwestern Sumatra, east of the Philippines, around the Java Sea and the Sulawesi (Figure 3B). The spatial correlation of the SEAP trends between Hist-Ant (Hist-Nat) and Hist-all MME is 95% significant (insignificant) at .80 (.26). Hence, it can be inferred that the observed and simulated precipitation trends over Southeast Asia during 1979–2014 are markedly affected by anthropogenic forcing. The same analyses were also performed in each of the 12 CMIP6 models (not shown), and the results are basically consistent with those in MME that the effects of anthropogenic forcing are more important than natural variation to drive the historical SEAP trends, indicating the model-independency of such a conclusion. Previous studies suggested that the anthropogenic forcings including the greenhouse gas and anthropogenic aerosols both contribute to the changes in regional precipitation (e.g., Hoerling et al., 2010; Polson et al., 2014; Guo et al., 2022). Thus, we further analyze the relative roles of anthropogenic greenhouse gas and aerosols on human-induced precipitation trends over Southeast Asia. Here, the historical greenhouse gas experiments (named as Hist-GHG) that forced by well-mixed greenhouse gas changes with the same ensemble members as Hist-Nat are used, and the response of aerosol forcing (named as Hist-Aer) is calculated as the difference between the anthropogenic forcing and the greenhouse gas runs (Taylor et al., 2012; Song et al., 2014). It can be found that the Hist-GHG MME explains the human-forced positive precipitation trends primarily over the southern Borneo and the Sulawesi (Figure 3C). While those in the Hist-Aer MME are mainly located over the eastern Indochina Peninsula, east of the Philippines, the South China Sea, and the northern Sumatra (Figure 3D). It indicates that both the greenhouse gas and anthropogenic aerosols contribute to some SEAP trends in Hist-Ant MME. However, the spatial correlation between the Hist-GHG (Hist-Aer) and the Hist-Ant MME is insignificant (significant) at –0.05 (0.77) over the whole of Southeast Asia. As a result, the effects of aerosols forcing are more important than greenhouse gas to contribute to the human-forced SEAP trends during 1979–2014.

In order to obtain a clearer image of the SEAP trends, Figure 4 shows the area-mean trends over Southeast Asia (10°S–20°N, 90°–140°E) during 1979–2014 in the observations and in the simulations. As shown, GPCP displays a significant positive trend over Southeast Asia, with the magnitude at 0.16 mm day⁻¹ decade⁻¹. The trend in CMAP is slightly higher than that in GPCP, with the magnitude 90% significant at 0.26 mm day⁻¹ decade⁻¹. The results are consistent with their spatial distributions as shown in Figures 2A, B. The increasing of precipitation trends over Southeast Asia is partly reproduced by Hist-all MME, whereas its magnitude is smaller than those in the observations, approximately 0.03 mm day⁻¹ decade⁻¹ exceeding the 90% confidence level. It suggests that the SEAP trends simulated in Hist-all MME could explain about 18.8% and 11.5% of the observed trends in GPCP and CMAP, respectively. The Hist-Ant MME accounts for most of the positive SEAP trends (.04 mm day⁻¹ decade⁻¹) modelled in Hist-all MME. On the contrary, the trend in Hist-Nat MME is insignificant and negative (–.01 mm day⁻¹ decade⁻¹). Hence, it reveals that anthropogenic forcing plays a primary role in driving the increasing of the SEAP trend from 1979 to 2014. Further, a separation of anthropogenic forcing indicates that the anthropogenic aerosols rather than the greenhouse gas forcing explain most of the anthropogenic forced SEAP trends, indicating the importance of anthropogenic aerosols forcing. However, these results of separated anthropogenic forcing may have some uncertainties because we adopt the linear assumption to investigate the roles of different external forcings. Whether the responses of climate system in a real world to external forcing, especially the anthropogenic forcings, are strictly linear remains unknown (Song et al., 2014). Since the primary purpose of the study is to distinguish the impacts of anthropogenic forcing and natural variation on the SEAP changes instead of the different roles of anthropogenic contributions, in the following analysis, we only focus on the historical outputs including the Hist-Nat and Hist-Ant simulations. Furthermore, the above results still provide evidence that the anthropogenic forcing definitely impacts the precipitation changes over Southeast Asia.

To evaluate the reliability of the results, the probability density functions (PDF) of area-mean precipitation trends over Southeast Asia (10°S–20°N, 90°–140°E) from the realizations of historical all, natural, and anthropogenic forcings are compared (Figure 5). The observed SEAP trends in GPCP and CMAP (red and blue lines) can possibly happen in an all-forcing world (Figure 5A), and about 15 out of the 58 realizations (25.9%) from Hist-all runs show comparable trend (>.10 mm day⁻¹ decade⁻¹) with the observations. In addition, the observed SEAP trend for GPCP falls inside the 5%–95% confidence interval range (–.147–0.211 mm day⁻¹ decade⁻¹) in a real world simulation. It implies that the increased SEAP trend in the observation can be partly reproduced in Hist-all forcing. On the contrast, the PDF of Hist-Nat simulations can hardly capture the observed trends (Figure 5B), and only three out of 58 realizations (5.2%) displays comparable trend (>.10 mm day⁻¹ decade⁻¹) with the observations. Furthermore, the 5%–95% confidence interval level of SEAP trends in Hist-Nat ranges from –.163 to 0.085 mm day⁻¹ decade⁻¹, and does not contain any observations. The results imply that the positive SEAP trends in the observation is impossible to happen without anthropogenic forcing. The PDF of precipitation trends from Hist-Ant simulations as well as the 5%–95% range (–.196–.210 mm day⁻¹ decade⁻¹) present consistent features with those from Hist-all simulations (Figure 5C), and there are 18 out of 58 realizations (31.0%) showing comparable trend with the observations (>.10 mm day⁻¹ decade⁻¹). The Kolmogorov-Smirnov test shows that the PDF of the area-mean precipitation trends in Hist-all and in Hist-Ant (Hist-Nat) are from the same (different) distribution, statistically significant at the 95% confidence level. Hence, it is suggested that the effect of anthropogenic forcing on the observed SEAP trends is recognizable.

Figure 6 shows the 5-year running mean of the winter precipitation anomalies averaged over Southeast Asia (10°S–20°N, 90°–140°E) during 1979–2014. In the observation, the regional mean SEAP in GPCP (red line) and in CMAP (blue line) both show an increasing trend (Figure 6A), with the increasing rate 95% significant at .17 and .27 mm day⁻¹ decade⁻¹, respectively. The Hist-all MME (black line) simulates the similar precipitation evolution over Southeast Asia to the observations, with an increasing rate 95% significant at .03 mm day⁻¹ decade⁻¹. The correlation coefficients of the SEAP series between the Hist-all forcing and observations in GPCP and CMAP are .66 and .70, respectively, both of which exceeding the 99% confidence levels based on the two-tailed Student’s t-test. The precipitation evolutions simulated in Hist-Nat and Hist-Ant MME show comparable relations with that in Hist-all MME (Figure 6B), and their correlation coefficients are .46 and .49, respectively, exceeding the 95% confidence levels. However, the increased SEAP trend in Hist-Nat MME is near zero during 1979–2014, and that in Hist-Ant MME is 95% significant at .03 mm day⁻¹ decade⁻¹. As a result, the increasing of the winter SEAP trend during 1979–2014 is primarily dominated by anthropogenic forcing. In addition, a close inspection on these historically simulated SEAP series reveals that the Hist-all simulated SEAP evolution before the 1990s and after the 2000s can be well captured by Hist-Ant MME, and their correlation is 95% significant at .72 during the period 1979–1990 combined with 2001–2013, larger than that in Hist-Nat MME with the insignificant correlation coefficient at .22. Nevertheless, Hist-Nat MME presents better simulation of SEAP evolution in Hist-all MME during the 1990s, and their correlation coefficient is .90 for 1991–2000, above the 95% confidence level, while that for Hist-Ant is −.50 below the 90% confidence level. These results imply that the historical evolution of the winter SEAP from 1979 to 2014 simulated in Hist-all MME could be caused by both the natural forcing and anthropogenic activity. Although the effect of anthropogenic forcing may be more important, the contribution of natural forcing to the SEAP evolution should not be ignored.

As shown in Figure 6A in the observation, there is a notable decadal shift of the mean SEAP that occurred after the mid-1990s, in accord with that reported in Dong et al. (2021). A moving Student’s t-test with a 9-year window applied to the observed and simulated SEAP series confirms the decadal shift, and the shift points are detected in 1995 for GPCP and CMAP (Figures 7A, B) and in 1994 for Hist-all MME (Figure 7C), respectively. To better quantify the decadal shift, the whole period 1979–2014 was divided into two sub-periods, 1979–1994 and 1995–2013, to represent negative and positive phases of the winter SEAP, respectively. In the observation, the mean SEAP during 1979–1994 in GPCP (CMAP) is –.168 (–.279) mm day⁻¹, while that during 1995–2013 is .234 (.365) mm day⁻¹. These in Hist-all MME during the two sub-periods are –.032 and .029 mm day⁻¹, which can be well simulated by Hist-Ant MME (–.021 and .024 mm day⁻¹). The differences of their mean values between these two sub-periods for observations, Hist-all, and Hist-Ant MME are 90% statistically significant based on the two-tailed Student’s t-test. In contrast, the Hist-Nat MME fails to simulate the decadal shift of the winter SEAP, and its mean values for 1979–1994 and 1995–2013 are –.011 and .005 mm day⁻¹, respectively, with the difference of mean values below the 90% confidence level. These results indicated that the decadal shift of the winter precipitation over Southeast Asia that transforming from a negative to a positive phase after the mid-1990s causes the significantly increased precipitation trend during 1979–2014, which could be partly reproduced by anthropogenic forcing. To validate the result, the winter precipitation differences between 1995–2013 and 1979–1994 over Southeast Asia in observations and in historical simulations are further shown in Figure 8, and they present a large similarity to those in Figures 2, 3A, B. The spatial correlation between the decadal difference and the linear trend of SEAP in each pair is above .80, exceeding the 99% confidence level based on the two-tailed Student′s t-test, indicating a close relation between the SEAP trend and the precipitation decadal variation. The Hist-all MME (Figure 8C) well reproduces the decadal features of the winter SEAP in the observations (Figures 8A, B), with the spatial correlations 95% significant at .39 and .42 for GPCP and CMAP, respectively. In addition, the decadal increasing of precipitation over Southeast Asia can be mostly captured by anthropogenic forcing (Figure 8E), while natural forcing only captures the decadal increases of precipitation southeast of the Philippines and over the southern Thailand (Figure 8D). Hence, these results highlight the dominance of anthropogenic activity on the decadal changes of the winter SEAP during 1979–2014.

Early studies suggested that the Pacific sea surface temperature (SST) pattern such as ENSO is the most crucial factor in modulating the winter precipitation over Southeast Asia via the Pacific Walker circulation and anomalous ascending/descending motions over the Maritime Continent (Feng et al., 2010; Dong et al., 2021). Hence, to examine the possible causes for the decadal increasing of SEAP during 1979–2014, the role of SST should be investigated. Figure 9 shows the linear trends of skin temperature over ocean from the ERA5 reanalysis data and the historical simulation with different external forcings. In the observation, a warming trend with a La Niña–like pattern has been identified in the Pacific (Figure 9A). The SST warming shows a typical pattern of a horseshoe–like structure, with faster warming in the tropical western Pacific than that in the tropical eastern Pacific, indicating the zonal SST gradient in the tropical Pacific to be stronger (An et al., 2012; Yeh et al., 2012). In the North Pacific, there are some SST signals similar to the Victoria mode, which is the second dominant mode of the North Pacific variability and is closely associated with the ENSO (Bond et al., 2003). The results are consistent with previous studies (e.g., Hua et al., 2018; Luo et al., 2018; Jiang and Zhu, 2020; Zhang et al., 2021), in which a La Niña–like warming pattern in the tropical Pacific and the strengthening trends of Victoria mode over the North Pacific (Bond et al., 2003; Ding et al., 2015) have been observed during recent decades. This non-synchronous warming trend of the Pacific SST has been well reproduced by the Hist-all MME and the Hist-Ant MME in spite of their smaller magnitudes (Figures 9B, D). The spatial correlation of the SST trends between Hist-all and observation (Hist-Ant) is .41 (.95), exceeding the 95% confidence level based on the two-tailed Student’s t-test. In contrast, these features of the SST trends hardly happen in a natural forcing world (Figure 9C). The composite difference of the SST anomalies between 1995–2013 and 1979–1994 (not shown) shows consistent results with those in linear trends. Thus, it is indicated that anthropogenic forcing can facilitate a La Niña–like SST warming pattern in the Pacific during 1979–2014. Previous studies have found that the fast warming in the tropical Indian ocean (e.g., Luo et al., 2012; Zhang et al., 2019; Zhang et al., 2021) and the Atlantic ocean (e.g., McGregor et al., 2014; Li et al., 2016), which could be driven by anthropogenic greenhouse gases forcing and aerosol forcing (Booth et al., 2012; Myhre et al., 2013; Hua et al., 2018), respectively, may contribute to the recent La Niña–like pattern in the Pacific via the strengthening of Pacific trade winds and the equatorial upwelling. In addition, the anthropogenic greenhouse forcing could also weaken the ENSO amplitude directly, which in turn causes a La Niña–like mean state in the Pacific (Kohyama et al., 2017). As a result, the view concluded in the paper that anthropogenic forcing facilitates a La Niña–like SST warming can be supported by the above studies.

In the observation, the zonal SST gradient in the tropical Pacific associated with the La Niña–like SST warming can result in the SLP trends to present a structure of the enhanced east-west pressure gradient during 1979–2014 (Figure 10A), indicating the strengthening of the Pacific Walker circulation. The SST induced SLP trends in the observation can be partly reproduced by Hist-all MME (Figure 10B), and is dominated by anthropogenic forcing (Figure 10D). The spatial correlation of SLP trends in the tropical Pacific between Hist-all and observation (Hist-Ant) is 95% significant at .56 (.96). However, natural forcing plays an unimportant role on the changes in SLP trends (Figure 10C). Therefore, the results indicate that anthropogenic forcing can facilitate the decadal increases of the winter SEAP from 1979 to 2014 via the induced Pacific La Niña–like SST warming and the resultant strengthening of the Pacific Walker circulation.