Atmospheric aerosols play a key role in driving recent climate changes through their complicated radiative effects1. They can perturb the radiative balance and cloud microphysical properties, inducing remarkable impacts on atmospheric circulation and water cycle in both regional and global scales2. Meanwhile, the complexity of aerosol composition leads additional uncertainties to the responses of climate system, since scattering and absorbing aerosols have significantly different or opposite effects3. Therefore, additional to the effects of total aerosol loading change, the change in the fraction of different aerosol types would also change the climate. Yet the effect of changes in scattering-absorbing aerosol fraction has been less emphasized by previous studies on aerosol effects in climate changes, such as precipitation.
Northern India is one of the typical regions where aerosols closely interact with precipitation. Located at the south of the Himalayan foothills, northern Indian’s climate is controlled by the South Asian summer monsoon system4. Rainfall during June to September (JJAS) provides about 80% of the annual precipitation5,6, which is essential for agricultural, industrial and domestic use of more than seven hundred million people. Observations of precipitation from multi-sources, including ground-based rain gauge and satellites, reveal a drying trend in the monsoon season over northern India, especially the Gangetic Basin (GP, Fig. S1), in the past few decades7–10, which has caused adverse impact on the social and economic activities in this most populated region all over the globe. At the same time, the intensive human activities in northern India released large amounts of pollutants to the atmosphere, causing an increase in the observed aerosol optical depth (AOD) over the past few decades11,12. A number of studies have been focusing on the aerosol-South Asian monsoon interactions13–18, many of which have noted distinct effects of different types of aerosols. Generally, the cooling effects of aerosol-radiation interaction and aerosol-cloud interaction can significantly perturb the atmospheric circulation and precipitation patterns by inducing land-ocean thermal contrast and hemispheric asymmetry of radiative forcing4,14,19,20, offsetting the precipitation increase due to greenhouse gases21,22. On the other hand, the heating effects due to absorbing aerosols such as black carbon during pre-monsoon season can induce an increase of moisture convergence feedback and shift the monsoon precipitation from central to northern India23. Dust minerals from the Middle East and the Arabian Sea can also induce an atmospheric heating over northwestern South Asia, enhancing meridional thermal contrast and strengthening monsoon circulation24. Although the effect of total aerosol changes as well as the individual effects of sulfate and carbonaceous aerosols have been considered25–28, whether aerosol changes are responsible for the observed GP drying is unclear. Moreover, no study has explicitly investigated the role of aerosol compositional change, namely the change in the fraction of scattering and absorbing aerosols, in driving the observed GP monsoon precipitation trend.
In this study, we analyze the contribution of changes in total aerosol mass (represented by aerosol optical depth, AOD) and in aerosol composition (represented by single scattering albedo, SSA) to the precipitation trend in the GP since the start of the satellite era. A set of aerosol control experiments with detailed arrangements of aerosol mass concentration fields are performed using the Community Earth System Model version 1.2.2 (CESM) with the Community Atmosphere Model version 4 (CAM4) as the atmosphere component, in which the change of AOD and SSA are controlled separately. We also evaluate the precipitation and aerosol optical property changes in the Coupled Model Intercomparison Project phase 6 (CMIP6) simulations to identify potential linkage between precipitation and SSA trends.
Observed and simulated precipitation and aerosol optical property trends in the Gangetic Plain
Since the start of satellite era around 1980, a significant decreasing trend in monsoon precipitation as large as 100 mm/year per decade has been observed over the GP. The trend is the most remarkable in the Global Precipitation Climatology Centre (GPCC) precipitation dataset (Fig. 1a) but also evident in multiple other datasets (Fig. S2). The average precipitation trend over GP (red contour in Fig. 1a) reaches up to -138.1 mm/year per decade according to the GPCC data for the 1979–2019 period (Fig. 1b). In particular, after 1999, an even stronger decrease is observed, reaching − 317.0 mm/year per decade.
The GP region is also one of the regions with the heaviest aerosol pollution in the world. Many studies indicated significant increases in AOD over the past two decades11,12. Surface sunphotometer measurements at the Kanpur station (26.5°N, 80.2°E, see method) reveal a significant increase in AOD at 440nm, reaching 0.083/decade (Fig. 1c). Meanwhile, significant increase in aerosol SSA at 440nm reaching 0.023/decade is also observed (Fig. 1d). These positive AOD and SSA trends suggest remarkable changes in both aerosol total mass and aerosol composition, in particular, a decrease of aerosol absorption likely caused by reduced black carbon fraction, which can also be inferred from the decrease of absorbing AOD at both 440 nm and 870 nm and increased SSA at 870nm (Fig. S3). Unfortunately, no long-term aerosol optical property observations are available before 2000 and no other stations in this area has measurements as complete as those at Kanpur. Nonetheless, the simultaneous decrease in precipitation and increase in AOD as well as SSA from 2000 to 2020 imply possible links between these parameters, which drive us to further investigate whether the aerosol changes are responsible for the precipitation trends, and whether it is AOD or SSA change that plays the major role, through climate model simulations.
CAM4 experiments are performed with only aerosol forcing and adjusting the AOD and SSA changes according to those at the Kanpur station (see method). The simulated AOD and SSA changes well agree with observation (Fig. 1f&g). Consistent with observation, the monsoon precipitation response to the aerosol changes also indicates a negative trend over the GP region, albeit with a relatively weaker amplitude of -75.3 mm/year per decade (Fig. 1e). There is also a slight southward shift in the center of the precipitation change, which might be due to the impact from topography. Overall, the CAM4 results confirm that aerosol changes, including both total loading and composition, can induce a drying trend over the GP region.
Roles of AOD and SSA in driving monsoon precipitation changes over the GP
We further examine the separate impact of total aerosol loading (represented by AOD) and aerosol composition (represented by SSA) changes on monsoon precipitation in the study area. Two sets of experiments are performed by changing AOD and SSA respectively, with the other parameter fixed at the 1980s level (Fig. S4, see method). Markedly, both AOD and SSA changes result in significant decreases of monsoon precipitation over the GP, whose magnitude is comparable to that in observation. Moreover, increasing SSA (referred as SSA experiment hereafter) appears to induce even stronger precipitation reduction than increasing AOD (referred as AOD experiment hereafter) (Fig. 2a-c). On average, decadal precipitation trend due to AOD change is ~-79.6 mm/year, whereas it is ~-96.2 mm/year in response to the SSA increase.
In response to the AOD increase, a strong surface cooling trend is observed over most of the Indian subcontinent (Fig. 2e). This cooling causes descending air anomalies corresponding to decreased omega in GP, which forms a high-pressure center there (Fig. 2h&k). Surrounding this high-pressure center, southeast wind anomalies are observed. The increase of SSA produces a slight and insignificant increase of surface temperature over the study area, which is due to the increase of surface shortwave radiation associated with increased aerosol scattering. The GP region also exhibits strong anomalous descending motions and high-pressure anomalies (Fig. 2i&l), which produces similar but stronger circulation pattern anomalies than the AOD experiment. The all-aerosol experiment (referred as AER experiment hereafter) shows similar changes in the meteorological variables to those in AOD and SSA experiments.
Because SSA increase slightly increases surface radiation but causes a strong decrease of vertical velocity, it is also necessary to examine the meteorological changes in the vertical dimension (Fig. 3). Increase of AOD and SSA clearly result in different changes in the vertical temperature profile. AOD increase causes the strongest cooling at surface and the lower atmosphere, but slight and insignificant warming in the upper atmosphere (Fig. 3b). In contrast, increase of SSA induces strong cooling in the upper atmosphere but weakly heats the surface (Fig. 3c). These different temperature profile changes thus result in different changes in the vertical velocity, that is, weaker omega decrease in the lower atmosphere for the AOD experiment (Fig. 3e), but stronger decrease throughout the atmosphere column for the SSA experiment (Fig. 3f).
In response to the AOD and SSA changes, the southeast wind anomaly tends to bring more moisture from the ocean surface to the GP region, whereas decreased vertical velocity indicates weaker convection and less chances for rainfall. To clarify and quantify their respectively impacts, we perform the moisture budget analysis by decomposing precipitation changes into the contributions of evaporation, horizontal moisture advection, vertical moisture advection and residual term changes (see method). In all three experiments, aerosol radiative effect causes a slight decrease in evaporation, relatively weak increase in horizontal water vapor advection but mostly confined to the eastern Indian coast, and much stronger decreases in the vertical moisture advection (Fig. S5). Clearly, vertical advection plays the dominant role in driving the precipitation change, accounting for − 76.8, -82.6 and − 120.0 mm/year per decade (102%, 104% and 125%) for AER, AOD and SSA experiments, respectively. This dominant role of vertical advection change is reasonable as previous studies highlighted the essential role of topography-related uplifting in precipitation formation in this region29. The above analysis altogether indicates that by cooling the atmosphere, an increase of aerosol SSA can significantly suppress convection and therefore reduces precipitation, an effect exceeding that by changing aerosol loading alone, but largely neglective by previous studies. Note that precipitation responses to AOD and SSA increase both appear stronger than that in the AER experiment, implying a non-linear relationship between aerosol scattering and absorption effects. This might be related to the height of the aerosol forcing, and will be investigated in future study.
Aerosol and precipitation trends in CMIP6 models
In addition to single model simulation, it is also necessary to examine whether multiple models in the CMIP6 ensemble capture the monsoon precipitation trends, in particular, the effect of SSA changes in GP. Unfortunately, the multi-model-mean historical simulations indicate a positive JJAS precipitation trend over the GP region, as well as most parts of South Asia (Fig. 4a), which is opposite to that in observation (Fig. 1a). The atmospheric model experiments (AMIP) also indicate positive precipitation trends in the GP (Fig. S6a). The aerosol forcing simulations in the CMIP6 ensemble nonetheless show a negative decadal precipitation trend of ~-50 mm/year over northeastern India (Fig. S6d), whereas greenhouse gas forcing induces positive precipitation changes (Fig. S6c). This suggests that CMIP6 models do capture the effect of aerosols on monsoon precipitation in the GP region, but with a serious underestimation.
All models simulate positive AOD trends in the study area, with different magnitudes and an averaged magnitude of ~ 0.1/decade, comparable to observation (Fig. 4b). However, seven models simulate weak positive SSA changes while three reports negative SSA trends, which combined results in an averaged SSA trend of ~ 0.0006/decade. By regressing modeled AOD trends against precipitation trends, a weak negative correlation is found (Fig. 4d), indicating that models with larger AOD increase also tend to simulate stronger precipitation decrease. The observed AOD and precipitation trend, however, appears to be an outlier from the simulated AOD-precipitation trend relationship. All models underestimate precipitation trends, although some show reasonable or even higher AOD trends. This means that other mechanism than AOD increase alone must be responsible to reproduce the observed precipitation trends. On the other hand, the simulated SSA trends and precipitation trends also exhibit a significant negative correlation, meaning that increased SSA tends to be associated with decreased precipitation. The observation also well fits the multi-model SSA-precipitation trend regression line (Fig. 4e). The analysis of CMIP6 results further supports that SSA change is more likely responsible to induce the downward monsoon precipitation trends over the GP region, and the underestimation of precipitation trends by CMIP6 models might be attributed to their incorrect representation of SSA trends, at least in the aerosol forcing experiment.
Implications of aerosol forcing and climate change
Aerosol radiative forcing have long been recognized to play important roles in driving local and global climate anomalies. Aerosols are also the most uncertain factor in anthropogenic forcing of climate change, due to their complicated composition and spatio-temporal heterogeneity. The forcing of aerosols arises from the changes of both their loading and their composition. While these two factors are implicitly embedded in climate change simulations, our study explicitly investigates the role of changing aerosol composition and highlighted its importance in perturbing the energy balance and the hydrological cycle. This compositional change, reflected as a change in the scattering and absorption properties, may significantly changes the vertical temperature structure as well as regional circulation, and further impact local climate such as precipitation. This effect can exceed that of changing aerosol loading alone.
The change of aerosol scattering and absorption is yet to be captured by most climate models, whose simulated SSA trends are largely apart from those in observation. This bias in SSA trend estimation might be the reason for the bias in their simulated precipitation trends, at least over the GP region. This result emphasizes the importance of accurate representation of aerosol composition and thus its optical property by the models, in additional to its total amount. Unfortunately, till now there is still lack of accurate SSA observations over the globe3, let alone its trends. A first step would be incorporating the limited long-term surface SSA observations into climate models. Next generation satellites, typically featured multi-angle, polarization measurements30 will likely provide SSA information globally31. Another factor that is even more difficult to estimate is the change of aerosol properties vertically, especially the scattering and absorption properties. As shown above, this might have caused the nonlinearity between the AER simulation and AOD-SSA experiments combined. Joint passive and active sensors with advanced technical design are required to depict a 3-D global aerosol picture3, in order to more accurately understand and predict climate change.