Impact of anthropogenic aerosols on Indian summer monsoon

Authors


Abstract

[1] Using an interactive aerosol-climate model we find that absorbing anthropogenic aerosols, whether coexisting with scattering aerosols or not, can significantly affect the Indian summer monsoon system. We also show that the influence is reflected in a perturbation to the moist static energy in the sub-cloud layer, initiated as a heating by absorbing aerosols to the planetary boundary layer. The perturbation appears mostly over land, extending from just north of the Arabian Sea to northern India along the southern slope of the Tibetan Plateau. As a result, during the summer monsoon season, modeled convective precipitation experiences a clear northward shift, coincidently in general agreement with observed monsoon precipitation changes in recent decades particularly during the onset season. We demonstrate that the sub-cloud layer moist static energy is a useful quantity for determining the impact of aerosols on the northward extent and to a certain degree the strength of monsoon convection.

1. Introduction

[2] The influence of anthropogenic aerosols (particularly absorbing aerosols) on tropical precipitation has become evident in recent years. Studies using different general circulation models (GCMs) all indicate that direct radiative forcing (DRF) of absorbing black carbon (BC) aerosols can lead to northward shift of precipitation in the Intertropical Convergence Zone (ITCZ) over the Pacific Ocean [Wang, 2004; Roberts and Jones, 2004; Chung and Seinfeld, 2005]. Recent modeling studies also suggest that DRF of aerosols could have a significant impact on another critical tropical precipitation system, the Indian summer monsoon [Lau et al., 2008]. Wang [2004, 2007] noticed an enhancement of the Indian summer monsoon circulation by BC DRF (the analysis was generally done on an annual-mean base). Ramanathan et al. [2005] found that an increase in the BC DRF over Indian Subcontinent and surrounding regions in their model leads to a reduction of monsoon precipitation while an enhancement to the pre-monsoon precipitation of March–April–May (MAM). Meehl et al. [2008] also found an enhancement of BC on pre-monsoon circulation and precipitation. However, during the monsoon months the effect of BC is likely to reduce the precipitation over India, although it might enhance the precipitation over the elevated Tibetan Plateau. The study of Meehl et al. demonstrated that BC DRF could weaken the surface temperature gradient between the tropical waters and the land of the Indian Subcontinent. This could serve as the forcing mechanism of BC on the monsoon circulation and precipitation. Lau and Kim [2006] suggested that the absorbing carbonaceous and dust aerosols accumulating along the slope of Tibetan Plateau could heat the air in these elevated places in April and May. This would initiate a positive feedback by drawing water convergence from oceans first and then form condensation and thus further heating over the slope of the Plateau. Based on this hypothesis, monsoon precipitation would be suppressed over central India due to aerosol-induced surface cooling. However, precipitation would come earlier and be enhanced over northern India and the southern slope of the Tibetan Plateau. The monsoon rainfall in July and August over the entire India would also be enhanced. Prive and Plumb [2007a, 2007b] indicated, however, that the equatorial oceanic air would have low moist static energy so that the low-level convergence might lead to a negative feedback. There is still a range of opinions about the reasons behind the impact of aerosols on the monsoon [Lau et al., 2008]. Detailed mechanisms of the aforementioned hypothetical impacts still remain to be examined.

[3] A considerable fraction of the world's population lives in the south Asia regions, where much of the agricultural activity and water resource is dependent on the summer monsoon evolution. A better understanding of the aerosol effect on monsoon circulation and rainfall has an important practical meaning. In this study we force an interactive aerosol-climate model with the DRF of several different combinations of carbonaceous and sulfate aerosols as well as their mixtures and then examine the effects of these different aerosols on the Indian summer monsoon circulation. Compared to the majority of previous similar studies [e.g., Lau and Kim, 2006; Wang, 2007; Meehl et al., 2008], prognostic aerosol processes and structures included in this study, especially the core-shell structured BC-sulfate mixture in addition to the external BC particles, are more sophisticated while arguably more realistic.

2. Model Configuration and Results

[4] We have used in this study a three-dimensional aerosol-climate model based on the Community Climate Model version 3 (CAM3) of the National Center of Atmospheric Research (NCAR) [Kim et al., 2008] (see also auxiliary material). Four model simulations have been carried out, namely a reference run (REF) that excludes the radiative effects of any aerosol, and three aerosol-forcing runs that include respectively the DRF of: 1) only scattering aerosols (i.e., all external sulfate modes, external organic carbon (OC), and the mixture of OC and sulfate (MOS); hereafter SCA); 2) only absorbing aerosols (i.e., external BC and the mixture of BC and sulfate (MBS); hereafter ABS); and 3) all sulfate and carbonaceous aerosols (hereafter COM). Note that we include OC and MOS in the SCA run because their optical properties resemble those of sulfate more than BC [Kim et al., 2008]. For this reason the results of the SCA run should reflect the effect of a weak solar absorption by OC and MOS. Note also that all four model runs actually include the same prognostic processes of all carbonaceous and sulfate aerosols, they differ only in the selection of aerosols to be included in the radiation module of the model. Each model run was driven by a slab ocean model and lasted 60 years to reach equilibrium. The last 20-year means are used in the analyses unless otherwise indicated. The climate changes caused by the aerosol DRF are isolated by comparing the results of SCA, ABS, or COM run with those of the reference run. We have also derived the statistical significances for all the diagnostic quantities [Wang, 2007]. Unless otherwise indicated, the results discussed below are statistically significant based on a paired t-test.

[5] A key factor in many of the hypotheses of aerosol-induced monsoon effects is the strong thermodynamic effect resulting from solar absorption of aerosols. This absorption induces a local heating of the atmosphere while it cools the Earth's surface. The local atmospheric heating could affect the atmospheric stability depending on the vertical location of the aerosol layer. The surface cooling would create a temperature gradient between the area underneath the aerosol layer and surrounding regions. Including the aerosol-climate interaction in our model produces a strong seasonality in the modeled three-dimensional distributions of aerosol absorbing strength (auxiliary material) [also Wang et al., 2009]. A high center of aerosol absorbing optical depth (AAOD), a quantity describing the solar extinction by aerosols through absorption in a given layer [Seinfeld and Pandis, 1998], is located over the Indian Subcontinent during the pre-monsoon season. However, during the summer monsoon season (June through August, JJA), the value of AAOD over the Subcontinent is greatly reduced due to the enhanced precipitation scavenging. The highest values of AAOD appear west of the Subcontinent. The high zonal-mean AAOD center over land is mainly located along the southern slope of the Plateau, with a relatively reduced value during summer compared to that in the pre-monsoon season.

[6] Our results indicate that the inclusion of absorbing aerosols (COM and ABS runs) affects convective precipitation both during the onset months (May and June) and in the summer monsoon season (June–July–August). Such changes in convective precipitation are reflected as an increase over the land areas north of the Arabian Sea, extending towards the north and also towards the southeast along the south slope of the Plateau, and a decrease over most of the land areas of the Subcontinent south of 20N (Figure 1). The similarity between the results of COM and ABS run suggests the dominant role of absorbing aerosols in causing such an alteration of convective precipitation. The changing pattern of convective precipitation in COM and ABS run is in agreement with that of the observed precipitation between the 20-year periods of 1981–2000 and 1946–1965 (CRU TS 2.1 land dataset) [Mitchell and Jones, 2005], and the pattern of 1951–2002 precipitation trend derived using the same data by Chung and Ramanathan [2006]. The use of 1950s as the reference is based on the estimated historical emissions of BC [Ramanathan et al., 2005]. The similarity between modeled and observed precipitation change patterns particularly during the onset season suggests an important role of anthropogenic absorbing aerosols in influencing the summer monsoon system. Note that we do not assume the effect of absorbing aerosols to be the only reason behind the observed change of monsoon precipitation. For instance, the observed reduction of JJA precipitation appears in a larger area and extends further north than our results, and this cannot be only explained by the effect of aerosols. We have also noticed some differences in the details of precipitation change pattern between our results and those of previous modeling studies [e.g., Chung and Ramanathan, 2006; Lau and Kim, 2006; Wang, 2007; Meehl et al., 2008], likely attributed to differences in modeled properties of absorbing aerosol.

Figure 1.

(top) May-June and (bottom) June-July-August average changes in convective precipitation (dm/season) derived from: COM, ABS, and SCA run for India and surrounding regions, and the observed precipitation change (land-only; dm/season) derived from the data of the Climate Research Unit (CRU) at the University of East Anglia. Model results shown are based on year 41–60 mean differences with REF run. CRU results are derived from differences between the 20-year means of 1981–2000 and 1946–1965, and based on the version 2.1 dataset with 0.5 degree resolution.

[7] Previous studies suggested that the summer monsoon system is a northward extension of the tropical ITCZ [Chao, 2000; Chao and Chen, 2001] and that the onset of the monsoon could be a result simply of a longitudinal SST gradient, not necessarily the traditional belief of land-ocean temperature gradient. It has been demonstrated [Prive and Plumb, 2007a, 2007b] that the poleward boundary of the monsoon circulation is co-located with the maximum sub-cloud layer moist static energy (or entropy; hereafter MSE), corresponding to the minimum of vertical meridional wind shear [Emanuel, 1995]. Such a location and extent of the monsoon would also be influenced by the position of sub-tropical thermodynamic forcing as well as the advection of MSE. The MSE is defined as:

equation image

Here cp is the specific heat at constant pressure, Lv the latent heat of vaporization, T the air temperature, q the specific humidity (water vapor mixing ratio), and ϕ the geopotential. Its anomaly caused by incorporating aerosol DRF in the model compared to the reference run can be derived using the anomalies of T, q, and ϕ defined similarly at any model grid. In our analysis, we have applied a topography correction to the value of h. We use a mean over the 3 lowest model layers to represent the sub-cloud layer anomaly of h in the analyses.

[8] We find that during the monsoon onset season, anthropogenic aerosols indeed can introduce an anomaly to the sub-cloud MSE distribution (Figure 2), however, such forcing and corresponding anomalous wind in the planetary boundary layer differ by case for different types of aerosols. During the May–June period, in all three aerosol-forcing cases the highest anomalous thermodynamic forcing caused by aerosols appears in the northwestern corner of the Subcontinent north of the Arabian Sea. These anomalous centers are located meridionally across the regions with a strong gradient of mean MSE so that they extend the maximum MSE gradient centers further north. The forcing anomaly is the strongest in the ABS case, followed by the COM case and then SCA (caused by a weak absorption of organics aerosols; note that a different color scale for SCA case is used due to its low values compared to the two other cases). The SCA run produces widespread negative values of anomalous MSE over the analyzed domain. The aerosol-caused anomalous advection that brings air from the tropical Indian Ocean into the Subcontinent is also much weaker in SCA run than in the two other runs, owing to a weaker aerosol solar absorption. The strongest northward advection of MSE caused by aerosols occurs on the west side of the Subcontinent. The gradient of anomalous MSE along this air path is the smallest in ABS run, followed by COM run. These differences in aerosol-caused anomalous MSE distribution and circulation lead to different modeled strength in monsoon convective precipitation in various runs. The evident similarity between COM and ABS run results can be also seen in several other analyses including solar heating rate, temperature, and convective cloud cover during the monsoon onset (auxiliary material), which also largely coincides with the recent finding of a warming trend from 1979 to 2007 along the south slope of the Plateau [Gautam et al., 2009].

Figure 2.

May-June (MJ) mean of wind and moist static energy in the reference run (REF), which excludes the aerosol radiative effect, and anomalies of MJ mean wind and moist static energy derived from 3 model runs (ABS, SCA, and COM). Data shown are averaged values for the lowermost 3 atmospheric layers based on year 41–60 means. Unit wind vector = 1 m/s. Moist static energy is in 103 J/kg. Note that in the 3 anomaly plots, a different color scale is used in SCA. A terrain correction has been applied to the REF result.

[9] The thermodynamic forcing by absorbing aerosols could affect local atmospheric stability. However, it is the alteration made by absorbing aerosols to the distribution of MSE, the key factor determining the location of convection based on the statistical equilibrium theory that exhibits the most critical impact of these particles on large-scale summer monsoon circulation.

[10] The influences of different types of aerosols on monsoon structure are identified in the monsoon season analyses (Figure 3). The highest anomaly of sub-cloud MSE appears around 28N in all runs. However, the sign of this anomaly is different in the three model runs, where only COM and ABS produced a positive anomaly, reflecting partially the heating of absorbing aerosols to the boundary layer air. The positive anomaly in ABS and COM runs also causes an enhanced northward transport of water vapor to reach ∼25N and this further enhances sub-cloud MSE in the north. In combination, these effects create a stronger convective precipitation in the north and a significant reduction in the southern part of the Subcontinent. The maximum convective precipitation enhancement occurs just south of the location of the maximum anomaly of sub-cloud layer MSE, appearing as a two-peak structure in agreement with Prive and Plumb [2007b] although in our analysis we used anomaly rather than actual value of the MSE.

Figure 3.

Regional-zonal-mean (60E-90E) anomalies of: (a) moist static energy, lowermost 3 layer means (dHb; 103 J/kg;); (b) vertical column integrated meridional water vapor transport (dVQ; 106 kg/s); and (c) convective precipitation over land (dPRECC; mm/yr). Data shown are June-July-August means of year 41-60 and derived as differences between 3 aerosol-forcing runs and the reference run.

3. Discussions and Conclusions

[11] We have conducted three interactive aerosol-climate simulations, driven respectively by the direct radiative forcings of: only scattering aerosols, only absorbing aerosols, and both aerosols. The aerosol-caused changes particularly in the circulation and convective precipitation of Indian summer monsoon regions have been isolated by comparing the results of these aerosol-forcing runs with a reference simulation that excludes entirely the aerosol effect in the model.

[12] Among different types of anthropogenic aerosols, scattering aerosols have only a very limited impact on monsoon circulation and precipitation. On the other hand, absorbing aerosols, with or without the co-existence of scattering aerosols, have a strong influence on the monsoon circulation and the development of convective precipitation.

[13] We find that the influence of absorbing aerosols is reflected in a perturbation to the sub-cloud layer moist energy structure, initiated with a heating by absorbing aerosols of the planetary boundary layer, mostly over the land areas north of the Arabian Sea and also along the south slope of the Tibetan Plateau. This is then enhanced by the import of airmass with high water vapor concentration. The corresponding anomalous sub-cloud layer airflow that brings relatively humid air primarily originated from the Arabian Sea. The anomalous advection from the tropical Indian Ocean entering the Subcontinent does exist, although comparatively weaker. As a result of the perturbation in moist static energy by absorbing aerosols during the summer monsoon season, modeled convective precipitation clearly moves northward, represented by a significant reduction over the Indian Subcontinent south of 20N and a large enhancement north of this zone. The elevated areas with concentrated aerosol forcing would assist the enhancement but not serve as the driver of the discussed changes in both monsoon circulation and convective precipitation.

[14] Our analyses have demonstrated the usefulness of using sub-cloud layer MSE to determine the northward extent and to a certain degree the strength of monsoon convection [Prive and Plumb, 2007a, 2007b]. Despite being developed based on radiative-convective equilibrium [Emanuel, 1995], it seems to provide a better insight to changes in the Indian monsoon system than some of the traditional diagnoses. The MSE perturbation caused by absorbing aerosols, however relatively small compared to the reference level, has been demonstrated to be strong enough to cause a northward shift of the convective precipitation center.

[15] Note that our analysis in this study is focused on the Indian summer monsoon regions. Our simulations, however, do include aerosol effects all over the globe. It is likely that aerosol-caused perturbations in other regions (e.g., the west tropical Pacific) could affect Indian summer monsoon circulation and precipitation [Meehl and Arblaster, 2002; Wang, 2007]. Actually, this remote aerosol impact can be detected from the sharply differing distributions of sub-cloud layer MSE anomalies in the various simulations (Figure 2). This “teleconnection” issue surely demands further investigation, however, exceeds the scope of this study, so does the interesting issue of separating the effect of land-based from that of ocean-based aerosols.

[16] Previous studies suggest a potential impact of dust aerosols on the Indian summer monsoon [Kim et al., 2006; Lau et al., 2008]. Heavy dust loading normally appears over or west of the Arabian Sea during the monsoon season and might overlap with anthropogenic aerosols in some cases [Wang et al., 2009]. The issue of solar absorption by dust and the mixture of dust and anthropogenic aerosol constituents will be considered in a future study.

Acknowledgments

[17] This research was supported by the NSF (ATM-0329759), the NASA (NNX07AI49G), and the MIT Joint Program on the Science and Policy of Global Change. We thank Alan Plumb, Steve Ghan, William Lau, and Heiner Körnich for discussion and comment. The Climate and Global Dynamics Division (CGD) of the National Center for Atmospheric Research (NCAR) provided computer codes and related datasets of the CAM3. The NCAR is operated by the University Corporation for Atmospheric Research under the sponsorship of the National Science Foundation. We thank T. Mitchell and the Climate Research Unit (CRU) at the University of East Anglia for making the CRU TS 2.1 data available to us. We thank the two anonymous reviewers for providing constructive comments.

Ancillary