3.1. Monthly mean precipitation and variability
Figure 1 shows the spatial distribution of mean and standard deviation of rainfall over a period of 30 years, 1984–2014. This paper is not based on evaluating the historical simulations. To this end, we illustrate, through this brief analysis of comparing the historical CMIP6 simulations and observations, how well the models agree with observations, with the only objective of making a better assessment of the projections, in line with the observed records until the recent past.
On a quick spatial comparison of the mean rainfall, CRU and the CMIP6 models show a good agreement, though on a closer look, sub-regions show some differences. Coastal rainfall seen in CRU along Myanmar and Thailand are absent in the models. The other sub-domains show a good agreement in the distribution of rainfall although sub-regional differences are evident in the Philippines and over some areas of the Indonesian islands and Borneo. Notably, the mountain rainfall is not captured well by the models, probably due to the models’ coarse resolutions. Similar inferences can be made on the standard deviations in rainfall though higher variability observed in northern Vietnam and the Philippines is not reproduced by the models. Overall, this general view of the historical simulations helps to obtain an idea of the performance of the models using which their projections can be held credible as appropriate.
3.2. Evaluations of the CMIP6 historical simulations against CRU observations
The time-series of future projections under different scenarios is shown in Fig. 2. Over the entire Southeast Asia (averaged over all land points on SEA), as expected, the high-emission scenario SSP 5-8.5 shows higher magnitudes of rainfall throughout the decades to come, while SSP 3–7.0 shows the least. There are no significant differences in the projections between SSPs 1-2.6 and 2-4.5. It can also be seen the CRU observations edge closely towards the projections only during the recent past when compared to the historical simulations. This implies that the CMIP6 models overestimate rainfall over the entire SEA region during the historical period.
Over sub-domain D01, SSP 5-8.5 stands out as the highest rainfall scenario while SSP 3–7.0 indicates early decades of low rainfall but increasing towards the end of the century, following the trends of the other two scenarios. The SSP 5-8.5 indication of future rainfall could be an early warning for policy makers to prepare for extreme rainfall and flood, given this sub-domain largely covers vulnerable regions of mainland SEA. The projections (SSP 3–7.0) at the beginning of 2020s show some measure of agreement against CRU observations over the recent past, while the other scenarios show higher ranges. This could also mean that the D01 rainfall over the past has been close to the magnitudes of SSP 3–7.0 and that the future trends are also likely to follow a relatively dry trend, though this assumption needs higher scrutiny.
Over sub-domain D02, all three scenarios except SSP 3–7.0 show similar trends, though SSP 5-8.5 suggests a higher rainfall projection by the end of the century. The graphs also strongly indicate that for this region, SSP 3–7.0 remains the dry scenario throughout the 21st century. CRU observations are, however, are far from the historical simulations, showing a higher magnitude of observed rainfall implying that the projections are not in line with the past/recent trends. The sub-domain D03 shows some interesting features, compared to the other domains. While the projections are in line with recent CRU observations, all scenarios are close to each other, with SSP 5-8.5 remaining the high-end scenario. However, SSP 3–7.0 does not follow a dry trend as seen in the other domains and shows increasing rainfall projections towards the end of the century, higher than that of SSPs 1-2.6 and 2-4.5. However, we need to recognize that the CRU observations may not be an accurate representation over domains D01 and D02 due to the lack of gauges at those sites and that the precipitation pattern over these regions is characterized by complex climate types/physical landscape. Furthermore, this discrepancy between observations and historical simulations is expected given CMIP6 historical years are not directly comparable to observations year-to-year. Nevertheless, to reiterate, the intention to include this observational comparison is to better assess whether the bandwidth of the historical simulations contain the observations within them. This provides a better quantification of both the observational and historical (simulations) uncertainty so that the future projections can be considered along with this uncertainty.
3.3. Spatial distribution of precipitation changes
Figure 3 displays the annual and seasonal near-term future (2021–2040) changes in precipitation relative to the present-day average (1995–2014), with the stippling indicating a 90% confidence interval. During the near-term period, the annual (ANN) averaged precipitation changes show markedly drier conditions over Myanmar, parts of Thailand, southern islands of the Philippines and large areas of the maritime Indonesia. The drying signal is most pronounced under the SSP3-7.0 across several regions of Southeast Asia, both mainland and maritime. Marginally wet areas, however, are seen over the rest of the regions on both the mainland and the maritime. The DJF season, which is the Northwest Monsoon dominated seasons of the year, shows a clear gradient in rainfall patterns – a wetter maritime and a drier mainland. The drier regions are notable over areas of Myanmar, Philippines, Vietnam, and the southern maritime islands of Indonesia. What is noticeable amongst the different scenarios is the SSP 3–7.0, which largely indicates a dry near-term future for the entire Southeast Asia. The mountainous areas within the maritime continent, mainly, show relatively higher magnitudes of rainfall (again, except SSP3-7.0). The JJA season, during which the Southwest Monsoon is pronounced, also exhibits widespread dry regions of the maritime continent excluding the northern Borneo Island and parts of Peninsular Malaysia. Overall, amongst the different time scales and scenarios, SSP3-7.0 is seen strongly as a ‘dry’ scenario.
We also analyzed changes in precipitation during MAM and SON seasons but not shown in the figures, due to brevity. During MAM (SON), the analyses on future projections indicated less (more) rainfall over mainland SEA including the Philippines. While most regions in SEA are likely to experience marginally wet changes during SON, the southern part of Indonesia is likely to have reduced precipitation under SSP2-4.5, SSP3-7.0, and SSP5-8.5. On the other hand, MAM manifested widespread dry regions across the mainland SEA.
Figure 4 shows the future changes in precipitation during the mid-term period (2041–2060). The figure shows a similar pattern of changes as that of Fig. 3 but with higher magnitudes of dry and wet changes as well as a higher confidence among the models. It is notable that the areas with exhibit minimal dry conditions in the near-term period (Fig. 3) turn minimally wet in the mid-term period. For instance, in Fig. 4, the ANN change in the northwest SEA region is projected to have a higher precipitation. Although some areas in the mainland SEA including southern Philippines are likely to remain dry, many places in this region could experience marginally wet conditions under SSP3-7.0. Moreover, Malaysia, Brunei, northern Indonesia, and northern Philippines are likely to receive more rainfall than in the near-term period.
Some areas in southern Indonesia, however, could continue to experience drier conditions when compared to the mid-term period. This is suggestive that the increase in precipitation over SEA region in the future relative to the baseline period (1994–2014) is probably linked to the increasing annual global temperature anomaly. However, these marginal increases in rainfall are not seen in the seasonal precipitation changes. In DJF, dry conditions expanding eastward over northwest SEA, including northern Philippines, may be likely with wetter areas expected in the maritime continent except southern Indonesia under a high-end scenario. During JJA, more rainfall is seen across most regions over the mainland SEA, the Philippines, Brunei, and northern Indonesia but strong reduction in rainfall is depicted over southern Indonesia. The spatial distribution of the projected changes over SEA region is also consistent with the results of Tangang et al. (2020). Overall, SSP 3.7-0 continues to be the driest scenario among the others, as seen during the near-term.
Figure 5 shows the future changes in precipitation during the far-term period (2081–2100). A significant increase in rainfall is projected over most SEA regions during the ANN despite reduction in rainfall over some parts of Thailand, Myanmar, and Indonesia. The highest increase in rainfall may be likely under the SSP 5-8.5 scenario, over Myanmar, Kalimantan, and Papua NG. During DJF, decrease in rainfall is mostly seen over mainland SEA under high emission scenarios (SSP 3–7.0, 5-8.5) whereas increases in rainfall are seen under the mitigation scenario (SSP 1-2.6). Overall, the annual changes show strong increases in rainfall across all scenarios over mainland SEA and parts of Borneo, Philippines, and Indonesia, while reduced rainfall is seen over the southern maritime islands of Indonesia. DJF, however, exhibits strong decreases in rainfall over the mainland regions of SEA across all scenarios except SSP 1-2.6.
We also examined the future changes zonally, for a quick assessment on the entire SEA region across time-slices. This is shown in Fig. 6. In the near-term period, reduced precipitation can be clearly seen during JJA, in the southern regions of SEA, below the equator. While the ANN shows no significant change, DJF shows marginally wetter northern latitudes, except SSP 3.7-0, which shows drier conditions (as reflected in the spatial changes, Fig. 3). The mid-term shows clear patterns on the changes, with an overall increase on the northern regions above the equator, during ANN. DJF shows larger increases over all regions except decreases under SSPs 2-4.5 and 3.7-0, above 10°N. The JJA season exhibits a clear signal of decrease below the equator while increases above the equator can be seen under all scenarios. The signals of change emerge distinctly during the far-term. The ANN shows minimal decreases over the southern islands below 10°S. DJF shows stronger increases (decreases) of up to 10% and higher over the southern (northern) latitudes, although the increases in the northern latitudes are seen under SSPs 1-2.6 and 2-4.5. JJA signals show a strong dipole with large decreases below the equator and large increases above. The clear patterns of tendency increased and decreased rainfall in the southern (northern) hemispheres of SEA during DJF and during JJA, by the end of the century, is probably an indication of an enhanced monsoon precipitation and intensified double-Intertropical Convergence Zone (double-ITCZ). This is not further examined in this study, however, systematic biases of the double-ITCZ were examined by Tian and Dong (2020) and noted a reduction of biases from CMIP5 to CMIP6 models.
3.4. Changes in wind field and moisture content during far-term period
Two prominent monsoon seasons affect the weather and climate in the Southeast Asia: Northeast monsoon during November-March and Southwest monsoon during June-September. These monsoon seasons are projected to intensify and suppress in the different SEA regions during the far-term period as illustrated in Fig. 7. Higher increases in precipitation are associated with higher increases in moisture content and enhanced monsoon flow. For instance, in DJF, the northeasterlies from South China Sea funneling within Borneo Island, Malaysia Peninsular and Indonesia are projected to enhance with a significant increase of moisture band between 9°S and 9°N where high increases in rainfall are also found. Conversely, it is projected that Myanmar, Laos, southern Thailand, southern Cambodia, and southern Vietnam are likely to experience drier condition under all scenarios except SSP1-2.6 (Fig. 5). The drying condition is due to the minimal intensification of monsoon flow with none to minimal increase in moisture content. In JJA, enhanced monsoon flow with significant increases in moisture content is apparent over the Mainland SEA and the Philippines. The robust drying condition over southern Indonesia under all scenarios, as also seen in Fig. 6, is caused by the suppressed easterlies accompanied with minimal increases in moisture content during JJA season. Increases in Asian monsoon precipitation is also analogous to the study of Chen et al. (2020) and Wang et al. (2020). Furthermore, the projected increase in summer monsoon precipitation is linked to the projected large changes on the land and ocean thermal contrast (Wang et al., 2020).
3.5. Future changes examined using Box Plots
Figure 8 summarizes spatially averaged future changes. In the SEA region, the future changes among models indicate a strong median increase up to 4%, 5%, 6% and 9% towards the end of the century under SSP1-2.6, 2-4.5, 3–7.0, and 5-8.5, respectively. The smallest median change is projected under the SSP 3–7.0 during the near- (0.3%) and mid-term (2.4%) periods. Possible extreme changes show a substantial increase of 22% (far-term of SSP 5-8.5) and a decrease of about 8% (near-term of SSP 3–7.0) as represented by the 90th and 10th percentiles, respectively. The continued increase over the different domains in the SEA region in the future is apparent. In D01, D02, and D03, rainfall changes are projected with increases up to 8%, 3.5%, 8%, respectively, during the far-term period.
However, extreme changes are apparently higher in the sub-domains than in the averaged SEA region. Of interest, during the far-term period, upper extreme changes are projected to increase by 33%, 32%, and 22% at the 90th percentile while lower extreme values might decrease by 25%, 30%, 11% at the 10th percentile over D01, D02 and D03, respectively. All the lower extreme changes in the sub-regions are projected under the SSP 3–7.0 whereas the upper extreme changes are exhibited by the SSP 5-8.5. This may be suggestive that under a warmer climate, extreme changes could be exacerbated. It is apparent that a large spread in the interquartile range represents larger uncertainty during the far-term period across all domains.
3.6. Changes in the annual cycle
The multi-model ensemble precipitation annual change is briefly evaluated in Fig. 9. In the SEA region, future precipitation during December-February shows a strong increase in comparison to other seasons, indicating a likely strengthening of the Northeast Monsoon’s influence on rainfall (Fig. 8a). The months of March-May exhibit the least increase in precipitation in the near-term period. Moreover, the months of June-November show relatively wetter conditions but with a significant increase in the far-term period. The highest change in precipitation is during December with an increase of about 14% under the SSP5-8.5 while May shows a decrease of about 2% during the near-term under SSP3-7.0.
Furthermore, the annual changes display distinct patterns in the different domains. Markedly, D01, under SSP 5-8.5, shows strong increases during most of the months in the year by the end of the century, while maintaining the October peak rainfall. The high precipitation change during this month may be attributed to the withdrawal of the summer monsoon affecting future change of climate extremes. Ha et al. (2020) reported that during the month of October, the summer monsoon showed a delayed retreat in the mid- to the late-21st century, thus, this may increase precipitation extremes over the Indo-China Peninsula where D01 is located.
Sub-region D02 shows noticeable decreases during the March-May months (near-term) which is strengthened progressively during the mid-term and intensified during the far-term. The rainfall change also indicates higher proportions of the Northeast Monsoon rainfall during the end of the year. This is also consistent with the downscaled projections simulations undertaken by Villafuerte et al. (2020) who reported that the projected Northeast Monsoon rainfall in the Philippines indicated a robust increase in most areas in the mid-21st century under the high-emission scenario.
While the annual rainfall changes in the near-term of sub-region D03 has a minimal increase, the increases in the mid- and far-term are apparent. The strengthening of the Northeast Monsoon rains by November-January is also evident by the end of the century. These changes in the annual cycle imply strong future changes in the two main monsoon seasons (Northeast and Southwest) which influence the precipitation regime of Southeast Asia. The policy options, depending on which region is dependent on which monsoon season for rainfall, therefore need to be planned based on these plausible changes that are relevant to water resources management or extreme events such as floods/droughts/agricultural practices.
3.7. Response of precipitation to global mean temperature
Furthermore, we analyzed the sensitivities of SEA precipitation on global mean temperature (GMT) under different emission scenarios (Fig. 10). Sensitivity of precipitation changes is found to have a linear dependence on the changes in GMT. Seasonal mean precipitation exhibits robust increases with warming over SEA region (Fig. 10a,b). The calculated correlations for JJA are 0.52, 0.48, 0.55, and 0.64 whereas correlation coefficients during DJF are 0.79, 0.60, 0.62, and 0.60 under SSP1-2.6, 2-4.5, 3–7.0, and 5-8.5, respectively. Majority in the time series yielded positive rainfall change for both seasons. It is apparent that the near-term period resulted in a lower precipitation change in a relatively low warming scenario while the far-term period resulted generally higher future changes in rainfall in an extremely warm climate. However, we found that the annual response rate of precipitation to GMT under SSP1-2.6 is found highest among the scenarios whereas SSP5-8.5 yielded lowest (Fig. 10c). The median precipitation rate of increase with warming is at 8.9%, 6.3%, 3.6%, and 2.7% under SSP1-2.6, 2-4.5, 3–7.0, and 5-8.5, respectively. This means that the rate of change in precipitation per degree Celsius of GMT is more sensitive in SSP1-2.6 than in higher emission scenarios. Moreover, a slight increase in GMT is likely to trigger higher changes in precipitation under SSP1-2.6. The higher climate sensitivity of SSP1-2.6 is also found by Jiang et al. (2020) over the semiarid region of Central Asia.