Accelerated increases in global and Asian summer monsoon precipitation from future aerosol reductions

. There is a large range of future aerosol emissions scenarios explored in the Shared Socioeconomic Path-ways (SSPs), with plausible pathways spanning a range of possibilities from large global reductions in emissions by 2050 to moderate global increases over the same period. Diversity in emissions across the pathways is particularly large over Asia. Rapid reductions in anthropogenic aerosol and precursor emissions between the present day and the 2050s lead to enhanced increases in global and Asian summer monsoon precipitation relative to scenarios with weak air quality policies. However, the effects of aerosol reductions do not persist to the end of the 21st century for precipitation, when instead the response to greenhouse gases dominates differences across the SSPs. The relative magnitude and spatial distribution of aerosol changes are particularly important for South Asian summer monsoon precipitation changes. Pre-cipitation increases here are initially suppressed in SSPs 2-4


Introduction
Anthropogenic aerosols can affect climate either by scattering or absorbing solar radiation, or by changing cloud properties (Boucher et al., 2013). Overall, aerosols have a global-mean cooling effect, manifested, for example, in a slower rate of global warming in the mid twentieth century concurrently with the rapid increases in aerosol burden (Wilcox et al. (2013) The important role of anthropogenic aerosol in driving precipitation changes has also been documented, including the possible contribution to the spin down in the global water cycle in the mid-twentieth century (Liepert et al. (2004); Wilcox et al. (2013); Wu et al. (2013)). A greater response of global mean precipitation to anthropogenic aerosol changes compared to 15 GHGs is expected since shortwave drivers have a stronger influence on the water cycle (e.g. Liepert et al. (2009);Andrews et al. (2010); Rotstayn et al. (2013); Samset et al. (2016); Liu et al. (2018)). The apparent hydrological sensitivity (% change in precipitation divided by absolute change in temperature; Fläschner et al. (2016)) for anthropogenic aerosol is twice that for GHGs (Kloster et al. (2010); Salzmann (2016); Samset et al. (2016)). This enhanced sensitivity means that anthropogenic aerosol reductions might be expected to play a relatively more important role in future increases in global precipitation for a 20 given temperature change. Several studies using CMIP5 models estimate an increase in global mean precipitation between 0.09 and 0.16 mm day −1 by 2100 from aerosol reductions (e.g. Levy et al. (2013); Rotstayn et al. (2013); Westervelt et al. (2015)).
Their heterogeneous forcing distribution and influence on circulation patterns mean that the effects of future aerosol reductions are likely to be felt more strongly at regional than global scale. Previous work has identified relatively large temperature increases over Europe (Sillmann et al., 2013), the Arctic (Acosta Navarro et al., 2016), and East Asia (Westervelt et al., 2015), 25 compared to the global mean response. For precipitation, the regional response is particularly pronounced for the Asian summer monsoon (Levy et al. (2013); Westervelt et al. (2015); Acosta Navarro et al. (2017); Bartlett et al. (2018); Samset et al. (2018)).
Here, precipitation is sensitivie to changes in remote aerosol through its control on the Interhemispheric Temperature Gradient and Inter Tropical Convergence Zone (ITCZ) location, and to local aerosol changes, which further modify the local monsoon circulation (Polson et al. (2014); Dong et al. (2016); Guo et al. (2016); Shawki et al. (2018); Undorf et al. (2018)). Asia will 30 also undergo the largest anticipated changes in aerosol amounts worldwide (Lund et al. (2018); Scannell et al. (2019)), and is thus a region likely to see an anthropogenic aerosol influence on near-future precipitation trends.
Despite evidence that anthropogenic aerosols influence temperature and precipitation, quantification of the associated changes is hindered by several compounding uncertainties. The degree by which AA reductions enhance future climate change is model dependent, with models featuring weaker historical aerosol forcing generally predicting more moderate global mean warming 35 2 https://doi.org/10.5194/acp-2019-1188 Preprint. Discussion started: 30 January 2020 c Author(s) 2020. CC BY 4.0 License. (Gillett and Von Salzen (2013); Westervelt et al. (2015)) and precipitation increases due to aerosol reductions in future (Rotstayn et al., 2015). The uncertainty in aerosol radiative forcing itself is currently the largest source of uncertainty in estimates of the magnitude of the anthropogenic forcing on climate, with the most recent estimate producing a 68% confidence interval from ∼-1.60 to ∼-0.65 W m −2 (Bellouin et al., 2019). This is comparable to the range simulated by CMIP5 models: -1.55 to -0.68 W m −2 (Zelinka et al., 2014). The effect of this uncertainty on regional climate projections may be further enhanced 5 by feedbacks from circulation changes (Wilcox et al., submitted). The compensating effects from the response to different near-term climate forcers also adds to uncertainty in multi-decadal projections, with future changes in methane and nitrate aerosol having the potential to moderate future temperature enhancements from decreases in anthropogenic aerosol (Bellouin et al. (2011);Shindell et al. (2012);Pietikäinen et al. (2015)).
In opposition to the CMIP5-generation findings summarised above, Shindell and Smith (2019) recently dismissed the pos-10 sibility that future aerosol reductions might lead to rapid increases in the magnitude or rate of global-mean warming, even in scenarios with aggressive clean air policies, based on simulations with a reduced-complexity impulse response model (Smith et al., 2018). Yet, this conclusion may not hold when using a full coupled GCM and when investigating changes beyond global mean temperature Even if the short atmospheric residence time of anthropogenic aerosol possibly makes their effects negligible on centennial timescales, they are likely important for regional and global climate over the next few decades. This is particu- 15 larly the case for Asia where large aerosol emission changes are anticipated, and where aerosol has played an important role in historical changes, in particular for precipitation. In this study, we examine state-of-the-art models and scenarios in CMIP6 and make the case for a potential enhancement of increases in global and Asian temperature and precipitation on a 20-30 year time horizon due to anthropogenic aerosol. Such an effect is of particular importance for adaptation measures.
2 Data and methods 20

Models and experiments
We use data from CMIP6 (Eyring et al., 2016). At the time of writing, data were only available for between 6 and 14 models for the variables and experiments we consider. The data used in this study are summarised in Table 1. All available data is used for each experiment, except where otherwise stated. Many CMIP6 models include improved representation of aerosol microphysics and aerosol-cloud interactions, such as internal mixing and heterogeneous ice nucleation (e.g. Bellouin et al. The SSPs used in CMIP6 sample a far greater range of uncertainty in future aerosol and precursor emissions than the RCPs used in CMIP5 (Scannell et al. (2019); Lund et al. (2018)). Partanen et al. (2018) highlighted the importance of uncertainty in 30 aerosol emission pathways for the potential enhancement of global temperature increases from anthropogenic aerosol reductions. The CMIP5 RCPs 2.6-8.5 sampled only a limited range of this emission uncertainty, with an associated 0.18K difference in global mean temperature. Contrasting aerosol pathways spanning a wider range of emission uncertainty resulted in a 0.86K difference. The SSPs span most of this wider range explored by Partanen et al. (2018). They consider large, rapid reductions in aerosol and precursor emissions in SSP1-1.9, more moderate reductions (comparable to the RCPs) in SSP2-4.5 and SSP5-8.5, and continued increases in the coming decades in SSP3-7.0 (see Figure 1). Much of the spread in global emission pathways comes from diversity over Asia and North Africa (Lund et al., 2018).
In our analysis, we compare the future decadal-mean climate changes to the present day  across SSPs 1-1.9, 5 2-4.5, 3-7.0, and 5-8.5. We focus on the period up to the 2050s, when aerosol emission uncertainty is largest, but the full range of uncertainty in greenhouse gas (GHG) emissions has yet to be emerge ( Figure 1). However, the changes in GHG emissions in this period are not negligible, and the anomalies we show include the effects of changes in both anthropogenic aerosols and greenhouse gases. This precludes the quantification of the respective effects of aerosol and greenhouse gas changes, but does not prevent the identification of the main driver of the anomaly. If the magnitude of the anomaly decreases monotonically from 10 SSP1-1.9, which has the largest aerosol reduction, to SSP3-7.0, which has a moderate aerosol increase (Figure 1), this indicates that aerosol changes are the main driver of the climate response. Further confirmation of aerosol as the main driver is gained from the comparison of the anomalies in SSP2-4.5 and SSP5-8.5, which have similar aerosol pathways, but very different greenhouse gas pathways ( Figure 1). If the response in these two scenarios is similar, then the greenhouse gas influence has yet to emerge over the aerosol signal. As the differences in greenhouse gas emissions between the two scenarios increase, a larger 15 response is expected in SSP5-8.5, which has large increases in global GHG emissions compared to very moderate increases in SSP2-4.5 ( Figure 1). In cases where GHGs are the main driver of the anomaly, the responses will increase monotonically from SSP1-1.9 to SSP5-8.5.
In Section 4.1, we use an additional DAMIP (Detection and Attribution Model Intercomparison Project; Gillett et al. (2016)) experiment, SSP2-4.5-aer. This differs from the companion SSP2-4.5 in that only aerosol emissions are evolving while all 20 other forcings are held constant at their 1850 levels. This scenario allows the response to anthropogenic aerosols to be seen in isolation from the response to greenhouse gas changes and may thus provide support to any conclusion drawn from the analysis of the SSP2-4.5 experiment. Data for this experiment is so far only available for two models, CanESM5 and MIROC6.
In this analysis, decadal mean anomalies are again presented relative to the present day . However, in this case, the present day is necessarily defined based on the historical-aer simulation (a historical simulation where only anthropogenic 25 aerosol and precursor emissions are transient, also included in DAMIP).

Present day model evaluation
Here, we use a number of observation and reanalysis datasets to present a broad evaluation of the performance of CMIP6 models in reproducing present day  climatologies and linear trends in global temperature and precipitation, the interhemispheric temperature gradient, and Asian summer monsoon. Global temperature observations are taken from GIS-  (Figure 3e). This pattern is seen in almost all models, and is highlighted in the comparison of the zonal and meridional components of the 850 hPa wind in CMIP6 and ERA-Interim in Figure 3f. This weak extension of the summer monsoon into 35 eastern China, with an anomalously strong extension into the subtropical west Pacific is consistent with the pattern of differences between CMIP5 models and observations (Sperber et al., 2013), although the CMIP6 models are more skilful in their representation of the Indian summer monsoon compared to CMIP5 (Gusain et al., 2020).
While aspects of the CMIP6 multi-model mean summer monsoon compare well with observations, and the multi-model mean performs better than the individual models, there is a large inter-model diversity in the monsoon characteristics exhibited 5 (summarised in Figure 3f; maps of present -day means, and anomalies compared to observations are shown for individual models in Supplementary Figures S1-S5). Of the models we will consider on an individual basis in Section 4.1, CanESM5 has a small regional mean precipitation bias, but a weak pattern correlation, compared to APHRODITE. It has a particularly large dry bias over India, with less than 3 mm day −1 in the seasonal mean, and a relatively large excess of precipitation over the Tibetan Plateau and into China. MIROC6 performs relatively well over land, but has excessive precipitation west of India.

10
Such biases may affect the pattern of the precipitation anomaly in the SSPs relative to the present day (Wilcox et al., 2015).
Aerosol optical depth (AOD) is a measure of the extinction of solar radiation due to scattering and absorption by an aerosol layer. Comparison of simulated and MODIS (Remer et al. (2008); Platnick (2015)) 550nm aerosol optical depth for the common 2002-2014 period, shows that models underestimate AOD over much of the tropics and NH mid-high latitudes, and overestimate it in the SH midlatitudes (Figure 4a-c). This pattern is common across models, with the exception of CanESM5, 15 which overestimates AOD over Eurasia compared to MODIS (Supplementary Figure S6 shows the comparison between AOD from individual models and MODIS). AOD gives an indication of the amount of aerosol in a given location, but does not give an indication of aerosol forcing. It gives no indication of particle number, as AOD is influenced by the optical properties of those particles, and no indication of aerosol-cloud interactions, which account for the majority of aerosol forcing in models (Zelinka et al., 2014). 20

Future anthropogenic aerosol changes
In all SSPs the largest AOD changes are over Asia (Figure 4d-f), consistent with the changes in aerosol and precursor emissions ( Figure 1). Future changes in AOD are characterised by global decreases in SSP1-1.9, with the exception of an initial increase over South Asia, and positive anomalies relative to 1980-2014 over the Tibetan Plateau and southern Africa, which may be dust (Figure 4d). AOD changes are characterised by regional contrasts in SSP2-4.5, with an overall decrease in the NH 25 contrasted with an increase in the SH, and large decreases over East Asia against large increases over South Asia until the 2030s (Figure 4e). This Asian dipole pattern is persistent and strengthens from the present until the 2040s. In this pathway, the large increases in South Asian AOD are predominantly driven by increases in SO 2 emissions. SSP5-8.5 has similar aerosol changes to SSP2-4.5, consistent with the similar changes in emissions (Figure 1). The final scenario we consider, SSP3-7.0, also contrasts widespread decreases in NH AOD against increases in the SH (Figure 4f). However, this scenario also includes 30 large aerosol and precursor emission and AOD increases over East Asia and particularly South Asia. These increases are driven predominantly by SO 2 over South Asia, but have a BC contribution over East Asia (Figure 1). This pattern persists, but the East Asian increase starts to weaken by 2050 (Figure 1).
Global mean annual mean temperature anomalies for 2025-2034, 2035-2044, and 2045-2054 relative to 1980-2014 are shown in Figure 5a. The boxes show the interquartile range, based on model mean responses, with individual model-mean responses overlaid. For each period, the responses are broadly ordered according to their GHG pathway, and diverge with time in a similar fashion to GHG emissions. This suggests that anthropogenic aerosol plays at most a limited role in the evolution of global mean 5 near-surface temperature on these timescales, supporting the conclusions of Shindell and Smith (2019). However, as discussed later, it does play a role in the pattern and magnitude of regional temperature changes. Importantly, anthropogenic aerosol is the main driver of trends in the interhemispheric temperature gradient until 2050, which has a strong control on ITCZ position and the global monsoon, and thus regional precipitation (Figure 5d). There is a large spread across models, consistent with the large uncertainty in historical trends, but a monotonic increase in the magnitude of the anomaly from SSP3-7.0 to SSP1-10 1.9 is present in all three future periods. The dominant role of anthropogenic aerosol is further supported by the comparable magnitude of anomalies in SSP2-4.5 and SSP5-8.5.
There is a clear aerosol-driven signal in future increases in global mean precipitation and hydrological sensitivity (Figure 5b, c). There is the suggestion of the beginning of a shift towards GHGs as the dominant driver of precipitation increases in 2050, but this is not seen in hydrological sensitivity. where the SSP1-1.9 anomaly is marginally smaller than in SSP2-4.5. GHGs are 15 the main driver of global precipitation change by the end of the 21st century (not shown).

Asian summer monsoon response
The decrease in Asian monsoon precipitation observed in the second half of the twentieth century has been attributed to the global increase in anthropogenic aerosols (Bollasina et 2018)). In contrast to anthropogenic aerosols, where circulation changes are an important component of the response to forcing, GHGs mainly affect (increase) monsoon precipitation by enhancing tropospheric water vapour, and thus increasing moisture transport toward India (Li et al., 2015).

25
Global aerosol reductions in SSP1-1.9 briefly cause this scenario to warm faster than the others considered over Asia and East Asia, but this effect does not persist beyond the 2030s, and is not apparent over South Asia (Figure 6a). However, anthropogenic aerosol does affect the regional pattern of warming (Figure 7), with slower increases in land temperature in many areas in SSP2-4.5, 3-7.0, and 5-8.5 compared to SSP1-1.9. The growing influence of GHGs with time can also be seen in Figure 7 as greater warming in SSP5-8.5 compared to SSP2-4.5 over the ocean, and over land from 2040. Continued increases in anthropogenic 30 aerosol emissions in SSP3-7.0 appear to moderate land warming compared to other SSPs, despite large GHG increases.
As for the global mean case (Figure 5), the influence of aerosol is more clearly seen in regional mean precipitation than regional mean temperature (Figure 6b). Over Asia, the largest mean precipitation increase occurs, for all decades, in SSP1-1.9, the smallest increases in SSP3-7.0, and comparable changes in SSP2-4.5 and SSP5-8.5. The same pattern is found over East Asia for 2035-2044, but the picture is less clear for the other periods. This result could be a reflection of the role of internal variability or the transition to a period where precipitation increases are dominated by GHG changes, with East Asia warmer relative to present day in SSP2-4.5 and SSP5-8.5 compared to SSP1-1.9by the 2050s (Figure 6a, Figure 7. By 2100, GHGs are the dominant influence on the relative magnitude of the future increases in Asian summer monsoon precipitation across the The AOD anomaly compared to 1980-2014 in SSP2-4.5 is characterised by a dipole between increases in South Asian AOD and decreases over East Asia. This pattern is present in observations since 2010 (Samset et al., 2019), and is in contrast to the twentieth century increases in both regions. The relatively small South Asian precipitation increases seen in SSP2-4.5 and SSP5-8.5 compared to SSP1-1.9 and SSP3-7.0, are likely to be due to suppressed warming over India due to continued 15 anthropogenic aerosol increases there (Figure 1, 7). There is potential for this suppression to be enhanced by feedback between  Figures 9 and 10 show that the pattern of Asian precipitation changes are similar, regardless of the emission pathway that is followed, but that the magnitude of the changes are pathway dependent. Figure 9 shows the absolute anomaly compared to 1980-2014, while Figure 10 shows the anomaly relative to the SSP1-1.9 response. In Figure 10, the influence of GHGs can be 20 seen, with greater GHG emissions driving greater drying over the Equatorial Indian Ocean and further increases in precipitation over India (particularly in the comparison between SSP2-4.5 and SSP5-8.5, and in SSP3-7.0 in the 2050s), as for temperature ( Figure 7).

Aerosol only SSP2-4.5
Analysis of SSP2-4.5-aer, in which only anthropogenic aerosol emissions are varying with time following the SSP2-4.5 path-25 way, allows the response to aerosol changes to be isolated from that due to GHG changes. In this case, a dipole in temperature anomalies, with cooling over India and warming over East Asia, and in sea level pressure, with a positive anomaly over India, the Bay of Bengal, and the Indo-China peninsula, and a negative anomaly over the rest of Asia, can clearly be seen in both CanESM5 ( Figure 11) and MIROC6 ( Figure S7). This feature matches the dipole pattern in AOD changes, and is apparent in the SSP2-4.5 response up to 2050 as a moderated GHG-induced warming over South Asian relative to East Asia (Figure 7). 30 Differences in the character of the precipitation anomaly can be seen when comparing the anomalies pre-and post-2050. In the earlier periods, when South Asian aerosol emissions continue to increase, precipitation anomalies are either weakly positive or negative over India, the Bay of Bengal, and the Indo-China peninsula. Post-2050, when anthropogenic aerosols are decreasing throughout Asia, precipitation increases relative to 1980-2014. There are suggestions of this structure in the SSP2-4.5 response, where increases in precipitation are weak over India and the Bay of Bengal compared to the SSP1-1.9 response (Figure 9, 10), but it is not as clear. This is likely partly due to the influence of GHG increases, and partly due to the effects of taking the mean response over models with large differences in their mean precipitation field (Figure 3; a number of the individual models do simulate a tripolar pattern in precipitation change in SSP2-4. 5 ( Figure S8)).

5
There is large uncertainty in future anthropogenic aerosol emission pathways. This is likely to be of limited importance for global-mean temperature, but anthropogenic aerosol does play an important role in changes in regional temperature, and global and regional precipitation until 2050 under the Shared Socioeconomic Pathways. Rapid reductions in anthropogenic aerosol and precursor emissions in SSP1-1.9 lead to larger increases in global and Asian summer monsoon precipitation compared to SSP2-4.5, 3-7.0, and 5-8.5 over East Asia and especially South Asia.

10
In SSP2-4.5 anthropogenic aerosol continues to increase over South Asia, in contrast to decreases over East Asia. This leads to a suppressed precipitation increase over South Asia in SSP2-4.5 over the next 30 years in this scenario relative to SSP1-1.9 and SSP3-7.0. Such a dipole in aerosol trends over Asia has been observed over Asia since 2010, suggesting that SSP2-4.5 (where the current pattern persists) or SSP1-1.9 (where anthropogenic aerosol is reduced in both regions) are more likely than SSP3-7.0 (where anthropogenic aerosol increases in both regions). This presents the possibility of large uncertainty in South 15 Asian summer monsoon precipitation on a 30-50 year time horizon due to uncertainty in aerosol emission pathways.
Data availability. All data used in this work are freely available for research purposes.
Author contributions. All authors designed the analysis and wrote the paper. LJW, ZL, BHS, EH, MTL, KN, and SU performed the analysis.
Competing interests. The authors declare that they have no conflict of interest.
Acknowledgements. This work and its contributors Laura Wilcox, Zhen Liu, and Massimo Bollasina were supported by the UK-China We acknowledge the World Climate Research Programme, which, through its Working Group on Coupled Modelling, coordinated and promoted CMIP6. We thank the climate modelling groups for producing and making available their model output, the Earth System Grid Fed-25 eration (ESGF) for archiving the data and providing access, and the multiple funding agencies who support CMIP6 and ESGF. APHRODITE 9 https://doi.org/10.5194/acp-2019-1188 Preprint. Discussion started: 30 January 2020 c Author(s) 2020. CC BY 4.0 License. precipitation data are available from http://www.chikyu.ac.jp/precip. GISTEMP temperature data, and CMAP, GPCC, and GPCP precipitation data were provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA, from their website at https://www.esrl.noaa.gov/psd/. We also acknowledge the use of ERA-Interim data produced by ECMWF and provided by the British Atmospheric Data Centre and the National Centre for Atmospheric Science. The analysis in this work was performed on the JASMIN super-data-cluster (Lawrence et al., 2012).
JASMIN is managed and delivered by the UK Science and Technology Facilities Council (STFC) Centre for Environmental Data Archival