Impact of aerosols on tropical cyclones: An investigation using convection-permitting model simulation



[1] The role of aerosols' effect on two tropical cyclones over the Bay of Bengal is investigated using a convection-permitting model with a two-moment mixed-phase bulk cloud microphysics scheme. The simulation results show the role of aerosol on the microphysical and dynamical properties of the cloud and bring out the change in efficiency of the clouds in producing precipitation. The tracks of the tropical cyclones (TCs) are hardly affected by the changing aerosol concentrations, but the intensity exhibits significant sensitivity due to the change in aerosol concentration. It is also clearly seen from the analyses that higher heating in the middle troposphere within the cyclone center is in response to latent heat release as a consequence of greater graupel formation. Greater heating in the middle level is particularly noticeable for the clean aerosol regime which causes enhanced divergence in the upper level, which, in turn, forces lower level convergence. As a result, the cleaner aerosol perturbation is more unstable within the cyclone core and produces a more intense cyclone as compared to the other two aerosol perturbations. This study, along with previous simulations, shows the robustness of the concept of TC weakening by storm ingestion of high concentrations of cloud condensation nuclei (CCN). The consistency of these model results gives us confidence in stating that there is a high probability that ingestion of high CCN concentrations in a TC will lead to weakening of the storm but has little impact on storm direction. Moreover, as pollution is increasing over the Indian subcontinent, this study suggests that pollution may be weakening TCs over the Bay of Bengal.

1 Introduction

[2] Land-falling tropical cyclones (TCs) are notorious for their destructive power. Well-known factors affecting TC intensity are heat and moisture surface fluxes, which, in turn, are determined by the sea surface temperature (SST) and surface wind speed [e.g., Anthes, 1982; Khain, 1984; Emanuel, 2005]. During the past decade, it was found that aerosols (including anthropogenic ones) substantially affect the cloud microphysics and, consequently, the rate of latent heat release, the dynamics, and the precipitation [van den Heever et al., 2006; van den Heever and Cotton, 2007; Levin and Cotton, 2009; Khain et al., 2009; Rosenfeld et al., 2008; Carrió et al., 2010]. It is therefore important to study the effect of aerosols on cloud processes and precipitation [Ramanathan et al., 2001] in a convective environment.

[3] The increasing industrial and vehicular pollution has manifested in the enhanced spatiotemporal distribution of atmospheric aerosol. The enhanced concentration of aerosol eventually leads to increased scattering and absorption of shortwave and longwave radiation. The modulation of radiation in the troposphere influences the life cycle of cloud and, finally, the precipitation. It is well documented that the cloud-radiative feedback and the effect of aerosol are the major source of uncertainty in climate models [Ramaswamy et al., 2001]. The question of how the aerosol distribution changes the heating distribution of the atmosphere has only been partly resolved. Although it has been demonstrated that small hygroscopic aerosols invigorate tropical convection, increasing vertical velocities and cloud top heights of deep convective clouds [van den Heever et al., 2006; van den Heever and Cotton, 2007; Khain et al., 2008b, 2009; Koren et al., 2005; Lee et al., 2008], the microphysical effects of aerosols on cloud clusters such as TCs have only been explored on a limited basis.

[4] There is some recent evidence of aerosols altering the microphysical properties of deep convective clouds [Khain et al., 2005; van den Heever et al., 2006; van der Heever and Cotton, 2007; Seifert and Beheng, 2006; Tao et al., 2007; Lee et al., 2008; Li et al., 2008; Carrió et al., 2010; Carrió and Cotton, 2011]. Zhang et al. [2009] indicated the possible effect of aerosol on TC development through the influence on the cloud microphysics. This was an idealized simulation in which it is concluded that hurricanes weaken because convective intensity in outer bands increased while that in the eyewall (which did not get many aerosols) decreased. Upper level divergence from the outer bands then suppressed eyewall convection. Rosenfeld et al. [2007] proposed a method of TC mitigation by seeding of clouds at the TC periphery near their cloud base with small aerosol particles of 0.05–0.1 µm in radius, thereby modifying the cloud microphysics in an ad hoc manner to estimate its effect. Likewise, Cotton et al. [2007] independently proposed that seeding TCs with small hygroscopic aerosol could weaken the storm. Simulations of the evolution of an idealized TC using the Regional Atmospheric Modeling System [Zhang et al., 2007, 2009] supported the conclusion that aerosols (e.g., Saharan dust) can substantially weaken the intensity of TCs. Carrió and Cotton [2010] simulated aircraft seeding of TCs in the outer rainband of an idealized TC. They found consistent weakening of the TC until the cloud condensation nuclei (CCN) concentrations reached very large values, wherein a “tipping point” was represented in which further increases in CCN concentrations lead to lesser reductions in storm intensity. Khain et al. [2008b, 2011] and Khain and Lynn [2011] used a spectral bin microphysical scheme [Khain et al., 2004] and simulated the effect of the increase in CCN concentrations on Hurricane Katrina during landfall. The enhanced CCN concentrations led to a reduction of maximum wind speeds by 10–15 m/s as well as a reduction of the area of strong winds.

[5] Simulations by Cotton et al. [2012] of Typhoon Nuri, which propagated into widespread pollution from the Asian mainland, revealed that during the early period of pollution ingestion, the storm intensified, but later on, the storm weakened in strength in accordance with the basic hypothesis. The reason the storm intensified during the early period of pollution ingestion was because the pollution plume invaded the eyewall region of the storm, resulting in intensified convection in the eyewall with little interference from downdrafts and cold pools in that nearly saturated region. Later, the pollution aerosol was prevented from reaching the storm interior due to scavenging, and thereby, the outer rainband convection was mainly altered, leading to the weakening of the storm as in the basic hypothesis. No change in storm direction was simulated for varying aerosol amounts.

[6] TCs are the most important weather systems which affect the socioeconomic condition of large coastal populations of India. The importance of deep convection and its large-scale organizations such as monsoon convection, Madden-Julian oscillation (MJO), etc., on the global hydrological cycle has been well documented [Houze, 1993], but the role of aerosols has not. So it is pertinent to establish the role of aerosol on the track and intensity of tropical cyclones, particularly in the backdrop of the anthropogenic pollution. The objective of this study is to demonstrate the role of the changing distribution of aerosol during two evolving cyclones over the Bay of Bengal and examine the impact of aerosol on the microphysical and dynamical properties of the cloud, as well as the change in efficiency of the clouds in producing precipitation.

2 Details of the Model and Experimental Setup

[7] For this study, we have utilized the fifth-generation Pennsylvania State University/National Center for Atmospheric Research mesoscale model (known as MM5) with modifications by Cheng et al. [2007] to include the two-moment warm cloud scheme of Chen and Liu [2004] (hereinafter the CL scheme) and by Cheng et al. [2010] to couple it with the mixed-phase scheme of Reisner et al. [1998] for ice-phase processes (hereinafter the CLR scheme). The CL scheme consists of a series of empirical bulk formulas for masses (the third moment) and number concentrations (the zeroth moment) of liquid water condensates. These formulas are derived based on statistical analyses of a parcel model simulation using the detailed cloud microphysics of Chen and Lamb [1994]. In addition, the CL scheme also provides diagnostic equations to calculate the mean terminal velocities and the effective radius of condensates, which are critical to precipitation process and radiation heating/cooling, respectively. This scheme requires a prediction of the concentration of atmospheric cloud condensation nuclei (CCN), which are assumed to be composed of ammonium sulfate and have a trimodal lognormal size distribution, and their activation into cloud drops following the Köhler theory. The Köhler-curve critical radius, which depends on the degree of supersaturation, of the last time step is retrieved from prognostic dry aerosol and total aerosol masses. When the Köhler-curve critical radius of the present time step is smaller than that of the previous time step, CCN with radius values in between are activated. In addition, the masses of aerosols inside clouds and precipitation are two new prognostic variables to account for the aerosols recycled from the evaporation of cloud drops. The CL scheme also considers the creation of rain embryos directly from giant cloud condensation nuclei (GCCN), so autoconversion is not the only mechanism of warm-rain production. More details on the aerosol activation treatment can be found in Cheng et al. [2007]. Other processes considered include condensation, cloud drop collision-coalescence, autoconversion of cloud drops into raindrops, accretion of cloud drops by rain, raindrop collision-coalescence and breakup, as well as deactivation of raindrops into cloud drops and cloud drops into CCN, thus allowing aerosol recycling.

[8] The ice microphysics of Reisner et al. [1998] includes three ice classes: cloud ice, snow, and graupel/hail. Their initiation, the growth by vapor deposition and riming, as well as the interchange between classes are fully described. Some of these processes, including ice nucleation and snowmelting, as well as size-dependent riming efficiency, were modified from the original scheme [cf. Cheng et al., 2010]. Condensation nuclei are initialized with trimodal lognormal size distributions, assuming an ammonium sulfate chemical composition. GCCN are included in the coarse mode, but only those with size greater than 10 µm turn into rain embryo during activation.

[9] There is no emission in the current version; only initial and boundary conditions are specified for CCN and ice nuclei (IN). For an initial sensitivity test, we selected three types of aerosols following Whitby [1978]: (1) clean continental, (2) average continental background, and (3) urban. We have added one more type, i.e., “maritime clean,” as the TC is over the ocean and demonstrated a few important findings. The total numbers of aerosols for each of these three types are shown in Table 1. The aerosol concentrations are varied from a “maritime clean” to a much polluted (urban) concentration. The high end of concentrations (urban) examined here is possibly out of the range of what has generally been observed over the oceanic region. However, a paper based on a recently carried out field experiment, i.e., the Cloud Aerosol Interaction and Precipitation Enhancement Experiment, over the Indian subcontinent shows variation of aerosol concentration in the range of 10–1500 cm−3 [Kulkarni et al., 2012]. In another study based on the Indian Ocean Experiment, Krishna Moorthy et al. [1998] showed a high aerosol concentration over the Arabian Sea. These studies show that in rare occasions, urban-type aerosol concentrations may be present over land and sea and may affect the atmospheric circulation over the whole Southeast Asia, as shown by Ramanathan et al. [2005]. For an idealized study such as this, the goal here is to examine the cloud process responses to a wide range of aerosol concentrations in a cyclone environment. We recognize that a large-scale homogeneous aerosol layer, particularly at urban levels, surrounding a tropical cyclone is not realistic but is included here to examine the response in this extreme scenario. A similar study to see the response of cloud processes in tropical deep convection under a wide range of aerosol perturbations has been recently reported by Storer and van den Heever [2013]. Aerosols are initialized in the present study as horizontally homogeneous and exponentially decreasing with height, except below the 850 hPa level, where the aerosol concentration is homogeneous [Cheng et al., 2010]. Although the CLR scheme has the capability of simulating ice nuclei (IN) effects, we applied implicit heterogeneous nucleation formulas of Huffman [1973] and Cooper [1986] for simplicity so as to focus on cloud condensation nuclei (CCN).

Table 1. Numbers, Means, and Geometric Widths of Nucleation (Nu), Accumulation (Ac), and Coarse (Co) Modes of Three Aerosol Size Distribution Types Over Continent, Including Clean, Average, and Urban Backgrounds
Mode (Factor)Number (cm−3)Mean (µm)Geometric Width
Nu (×103)Ac (×103)CoNu (×10−3)Ac (×10−2)CoNuAcCo
Maritime clean0.1330.0663.14.00.1332.90.530.690.77

[10] The modified MM5 was run with a four-level nesting structure, with the outermost and innermost domain resolutions set as 81 and 3 km, respectively, with 31 terrain-following sigma coordinates with the model top at 10 hPa. The first 24 h simulation was considered as the spin-up and was not analyzed. The 1° × 1° National Centers for Environmental Prediction final analysis meteorological data were applied to initialize the model. In this simulation, Grell cumulus parameterization [Grell, 1993] is used for the first two outer domains, and the remaining two inner domains are run with an explicit moist scheme. A medium-range forecast [Hong and Pang, 1996] planetary boundary layer (PBL) scheme is considered here for the PBL process. A five-layer soil model [Dudhia, 1996] and a cloud radiation scheme [Grell et al., 1994] are used for land surface and radiation processes, respectively.

3 Results

3.1 Tropical Cyclone Aila

[11] A tropical depression formed over the southeast Bay of Bengal at 0600 UTC on 23 May 2009. Under favorable conditions like warmer SST, low to moderate vertical wind shear, and upper level divergence, the observed storm intensified into cyclonic storm Aila at 1200 UTC of 24 May and into a severe cyclonic storm at 0600 UTC of 25 May. The system moved in a northerly direction and crossed the West Bengal coast close to Sagar Island, India, between 0800 and 0900 UTC of 25 May. A characteristic of the storm was the system's movement in a constant northward direction without showing much curvature. The devastation by this TC was particularly enormous as it maintained its intensity even 15 h after land fall.

3.2 Simulation of TC Aila

[12] The 12 hourly track, maximum surface wind, and track errors for the 60 h forecast of TC Aila with three categories of aerosol perturbations, namely, clean, average, and urban, are shown in Figures 1a–1c. It is found that the influences of the changing aerosol types on the tracks are marginal, but the intensity in terms of the maximum surface wind is significantly influenced by aerosols. The urban aerosol results in a weaker simulated cyclone (wind speeds are less). In contrast, the clean aerosol perturbation produced the most intense TC with higher wind speeds. The average aerosol concentration is found to simulate a TC of intermediate intensity and with lower instability which lies in between clean and urban simulation. Consistent with the latter findings, the clean (urban) aerosol is found (Figure 2a) to produce a maximum (minimum) surface-accumulated rainfall associated with a higher (lower) rain rate. If we add another aerosol concentration, i.e., maritime-clean type (much less CCN compared to a clean environment; see Table 1), an interesting finding is noticed where the surface-accumulated precipitation is not higher but is rather found to be less compared to that in a clean environment (Figure 2a). Thus, a “tipping point” is found for CCN concentrations where, either for lower or higher CCN concentrations, precipitation will be decreased (Figure 2a).

Figure 1.

The tropical cyclone (TC) tracks, maximum surface winds, and track errors for the 60 h forecast of (a–c) TC Aila and (d–f) TC Bijli, respectively, with three categories of aerosol perturbations, namely, clean, average, and urban.

Figure 2.

Total surface rainfall accumulation for two tropical cyclone (TC) cases (a) Aila and (b) Bijli under four types of background aerosol environments (maritime clean, clean, average, and urban). (c) Rain rates are compared with the TRMM3B42 observation for TC Aila. (d) Time evaluation minimum surface pressure as compared with the IMD observation.

[13] The inner domain-averaged precipitations simulated by the model are compared with observation (Figure 2c) on the response to CCN number changes. Figure 2c shows the 24 h time evolution of observed precipitation rates from TRMM3B42 data sets along with the simulated rainfall from 0000Z of 24 May to 0000Z of 25 May. According to the observations, the rain events are well simulated, except at 1200Z, where it is underestimated. This failure may be attributed to the error arising from the uncertainties of other physical processes, microphysics, or the lateral boundary conditions. The time evolution of the central minimum surface pressure as simulated by the model for clean and urban aerosol perturbations and the comparison with the observation (IMD) are also presented (Figure 2d). The results clearly show that the minimum surface pressure becomes less (more) for the clean (urban) type of background aerosol conditions. Figure 3 shows the vertical profiles of time- and domain-averaged cloud fields (cloud drop size and cloud drop number), simulated by the CLR scheme, with the change in CCN number (maritime clean, clean, average, and urban). The cloud water mixing ratio is also shown in Figure 4 for both TCs. The features of the first indirect effect (smaller droplets) and the second indirect effect (more cloud water) are well simulated. The number of cloud drops increases due to the increase in CCN number concentration (from clean to urban); however, cloud drop sizes decrease (Figure 3). Recent studies by several scientists [e.g., Rosenfeld et al., 2012; Khain et al., 2010; Cheng et al., 2010] have also shown that the cloud water content is substantially larger in the polluted (urban) clouds and produce smaller cloud droplets. The greater droplet condensation on small droplets is responsible for releasing latent heat [Rosenfeld et al., 2012]. In addition, the greater amounts of supercooled water result in greater latent heat of freezing as ice particles rime the supercooled water and drops freeze [van den Heever et al., 2006; van den Heever and Cotton, 2007; Carrió and Cotton, 2010].

Figure 3.

Vertical profile of domain-averaged (longitude: 87.5°E–89.5°E and latitude: 20°N–22°N for Aila; longitude: 87.5°E–89.5°E and latitude: 15°N–18°N for Bijli) and time-averaged (UTC of 24 May to 0600 UTC of 25 May for Aila and 1200 UTC of 15 April to 0000 UTC of 17 April for Bijli) cloud drop sizes (µm) and cloud drop number concentrations (cm−3) for (a–c) TC Aila and (d–f) TC Bijli.

Figure 4.

Vertical profile of domain-averaged (longitude: 87.5°E–89.5°E and latitude: 20°N–22°N for Aila; longitude: 87.5°E–89.5°E and latitude: 15°N–18°N for Bijli) and time-averaged (UTC of 24 May to 0600 UTC of 25 May for Aila and 1200 UTC of 15 April to 0000 UTC of 17 April for Bijli) cloud water mixing ratios (mg/kg) for (a) Aila and (c) Bijli and rainwater mixing ratios (mg/kg) for (b) TC Aila and (d) TC Bijli.

[14] It is found that the large numbers of cloud drops with lesser size are available in the atmosphere due to high (urban) background aerosol conditions. The vertical velocity becomes more for clean in comparison to maritime-clean and urban-type background aerosol environments (Figures 5a and 5c) for both TCs. The vertical velocity differences (maritime clean minus clean and urban minus clean) (Figures 5b and 5d) indicate that for lower (maritime clean) and higher (urban) CCN concentrations, the vertical velocity will be decreased compared to clean for both TCs.

Figure 5.

Vertical distribution of domain-averaged (longitude: 87.5°E–89.5°E and latitude: 20°N–22°N for Aila; longitude: 87.5°E–89.5°E and latitude: 15°N–18°N for Bijli) and time-averaged vertical velocity (w) profiles for (a) Aila and (c) Bijli using the mean values for the three aerosol types. Differences in vertical velocity (w_maritime-clean minus w_clean; w_urban minus w_clean) are also shown here for (b) Aila and (d) Bijli.

[15] To understand the microphysical reasons behind such differences in simulated precipitation, the domain- and time-averaged vertical profiles of rainwater are also analyzed here (Figures 4b and 4d). These results explain the presence of more cloud liquid water mass at the upper level and less rainwater mass at the lower level due to higher CCN concentrations (Figure 4). The mean bulk water loading of cloud water and rainwater is also shown in Figures 6a and 6b for both TC cases (Aila and Bijli). As the higher CCN concentrations generate greater cloud water (Figures 4a and 6a), it reduces the collision-coalescence efficiency to produce less rainwater (Figures 4b and 6a). Therefore, the rainwater mass (Figure 6a) shows a decreasing trend toward a higher CCN concentration. This essentially indicates that the addition of higher pollution reduces the precipitation formation processes as compared to the average or clean condition. The present study for TC Aila also suggests that graupel production first increases as CCN increases, and then, a “tipping point” occurs when graupel riming efficiencies get smaller as droplets get smaller with still higher CCN concentrations (Figure 6a). The rainwater production also increases first and then decreases with CCN concentrations (Figure 4a), which is also reflected in the accumulated surface precipitation (Figure 2a) of the TC “Aila” case.

Figure 6.

The mean bulk water hydrometeors (vertically integrated for all levels) for different aerosol types for the two TC cases: (a) Aila and (b) Bijli. Averaging is carried out over the domain (longitude: 87.5°E–89.5°E and latitude: 20°N–22°N for Aila; longitude: 87.5°E–89.5°E and latitude: 15°N–18°N for Bijli) and time (UTC of 24 May to 0600 UTC of 25 May for Aila and 1200 UTC of 15 April to 0000 UTC of 17 April for Bijli).

[16] To understand further the microphysics of the hydrometeors in different aerosol regimes, the cloud ice, snow, and graupel/hail (mixed phase) are analyzed. First, we will discuss the details of microphysical production terms for the case of “Aila.” For simplicity, we classified their production terms into four categories, namely, initiation, deposition growth, accretion (riming) growth, and melting for the three ice-phase hydrometeor species (i.e., cloud ice, snow, and graupel/hail) considered in the model.

[17] Cloud ice initiation includes homogeneous and heterogeneous nucleation from the vapor and liquid phases, as well as secondary ice multiplication. Snow is initiated primarily from the growth of cloud ice mainly by deposition and also by riming. Graupel/hail comes from snow riming or freezing of rain by contacting ice particles or by nucleation. Rain initiation includes two warm-rain processes, namely, rain embryo formation and autoconversion, and two cold-rain processes, such as snowmelting and graupel/hail melting. If we start analyzing the cloud ice processes, we find that the initiation of cloud ice (Figure 7a) decreases with the increase in CCN concentration. Actually, several controlling factors are responsible for cloud ice initiation. First, the cloud drop freezing rate (in terms of number) for either homogeneous or heterogeneous nucleation is proportional to the liquid water, which tends to increase with CCN concentration. Such effect is evident if we see the results of maritime-clean and clean background aerosols where cloud ice initiation increases first. Note that this effect is self-limiting as liquid water tends to decrease in glaciating clouds.

Figure 7.

Grouped microphysical conversion rates (a–d) for the TC case Aila and (e–h) for the TC case Bijli at the three aerosol types. The values of cloud ice initiation have been exaggerated by 10 times, and autoconversion has been exaggerated by 100 times for better visibility.

[18] A stronger influence of aerosol concentrations on freezing nucleation comes from the size effect as the mean cloud drop size is inversely proportional to its number. The cloud ice is found to be higher for urban (higher) aerosol conditions (Figure 6a), which indicates that the initial cloud ice size would be smaller for higher CCN concentrations. This initial particle size predominantly influences the subsequent growth of cloud ice by either vapor deposition or riming. The initiation processes directly influence the subsequent growth of cloud ice by vapor deposition (Figure 7a), suggesting that deposition is enhanced with increasing aerosol concentration, although ice initiation is the opposite [Tao et al., 2012].

[19] The vapor deposition is essentially enhanced by the Wagner-Bergeron-Findeisen (WBF) process [Tao et al., 2012]. This is due to smaller and numerous cloud droplets which evaporate faster, leading to enhanced water vapor for subsequent ice particle growth. This process was noted by Saleeby and Cotton [2008] for orographic clouds. This is seen in the vertical profile of cloud ice mixing ratio (Figure 8a), which reveals a similar aspect of increasing trends of cloud ice mixing ratio toward higher CCN concentrations. This also supports the earlier conclusion about cloud ice deposition.

Figure 8.

Vertical profile of domain-averaged (longitude: 87.5°E–89.5°E and latitude: 20°N–22°N for Aila; longitude: 87.5°E–89.5°E and latitude: 15°N–18°N for Bijli) and time-averaged (UTC of 24 May to 0600 UTC of 25 May for Aila and 1200 UTC of 15 April to 0000 UTC of 17 April for Bijli) cloud ice mixing ratios (mg/kg) for (a) Aila and (d) Bijli; snow mixing ratios (mg/kg) for (b) TC Aila and (e) TC Bijli; and graupel/hail mixing ratios (mg/kg) for (c) TC Aila and (f) TC Bijli.

[20] In the urban aerosol environment, more nonprecipitating cloud drops (Figures 3b, 3c, and 4a) are available, and this has produced a significantly higher amount of cloud ice (Figures 6a and 8a) compared to the average and clean aerosol environments. Snow is initialized mainly through autoconversion from cloud ice that is maintained under the growth by vapor deposition or accretion. So its rates show strong similarity with deposition growth of cloud ice, indicating a significant increase in the high CCN case (Figure 7b). The vapor deposition of snow is controlled by the amount of snow and available water vapor (i.e., ice supersaturation). The latter increases with aerosol concentration through the WBF mechanism [Tao et al., 2012], as mentioned above. Therefore, snow deposition slightly decreases at medium aerosol concentrations and increases at high aerosol concentrations (Figure 7b). The area- and time-averaged vertical profile of the snow mixing ratio shown in Figure 8b also indicates the same feature. We have also computed vertically integrated snow (Figure 6a) for TC Aila. The snow production decreases first then increases with increasing aerosol concentration. Snow growth by riming (Figure 7b) reduces with increasing CCN concentration due to two main factors: (1) large numbers of snow particles lead to a smaller snow size; and (2) large numbers of cloud ice (and snow) cause reduced cloud water content and cloud drop size due to the enhanced WBF mechanism [Tao et al., 2012]. Both these factors lead to lower collision efficiency. In addition, smaller snow falls slowly, sweeping a lesser cloud volume per unit time. Smaller sizes lead to lower riming efficiencies, as manifested by the snow riming curve in Figure 7b.

[21] Graupel may be initiated from cloud ice or snow during the liquid drop collection process. Although this graupel production rate is generally higher when there is more cloud ice or snow or cloud water, initiation is dominated by an additional factor, namely, the sizes of evolving particles. The graupel initiation rate (Figure 7c) reveals a decreasing trend with CCN concentrations. The domain- and time-averaged vertical profile of the graupel/hail mixing ratio has also been presented (Figure 8c), which also shows similar features.

[22] These results are in agreement with Carrió and Cotton [2010] and Cotton et al. [2012], wherein a “tipping point” was found when graupel formation is increased with CCN concentrations. But after the “tipping point,” a further increase in CCN concentrations reduced the droplet sizes so much that the efficiencies of riming and graupel formation are suppressed. Thus, the present study, even with a bulk microphysical scheme, shows similar results with that from the bin-emulating approach study of Carrió and Cotton [2010] and Cotton et al. [2012].

[23] Additional information for vertically integrated graupel/hail is shown in Figure 6a, which supports the statement mentioned above for both TC cases. As we have discussed earlier that more cloud ice and snow basically produce smaller particles at higher CCN concentrations, these smaller sizes help to lower riming efficiency (as shown in the snow riming curve in Figure 7b) that manifests in the graupel/hail initiation. Graupel/hail deposition is similar to vapor deposition of snow showing a slight decrease at medium aerosol concentrations and an increase at high aerosol concentrations (Figure 7c). But the riming process (Figure 7c) shows first increases as CCN increases and reaches a “tipping point” and then decreases. Similarly, the melting of graupel/hail also increases first and then decreases as droplets become small; a “tipping point” occurs when graupel riming efficiencies get smaller as droplets get smaller with still higher CCN concentrations. Thus, cold-rain production through melting of graupel also shows first increases reaching a tipping point and then a monotonic decrease with increasing CCN concentrations (Figure 7c). The mixed-phase hydrometeor (i.e., graupel) is found to be higher in a clean environment (Figures 6a and 8c), followed by average and urban aerosol conditions. The graupel/hail mixing ratio becomes the minimum for the maritime-clean environment (Figure 8c). These analyses bring out the fact that a clean environment supports processes such as melting, accretion, etc., with greater efficiency for the formation of graupel which further adds to the raindrop number distribution. More cloud ice with the urban aerosol environment is consistent with higher ice deposition with urban aerosol.

[24] Now, we evaluate whether the total melting (snow and graupel/hail melting) and warm-rain processes contribute to rainfall on the ground. The warm-rain initiation by autoconversion and the subsequent growth of rain by accretion are depicted in Figure 7d. The warm-rain initiation rate decreases monotonically with the CCN concentration as autoconversion decreases (Figure 7d), and at the same time, the rain growth by accretion exhibits a decreasing tendency with increasing CCN concentration (Figure 7d). Snowmelting shows first a decrease and then an increasing trend toward the high CCN concentrations (Figure 7b). This trend is very similar to that of the overall snow conversion rates, dominated by riming at low CCN and by initiation and deposition at high CCN (Figure 7b). On the other hand, the curve for graupel/hail melting indicates an opposite behavior with respect to snow melting: first an increase followed by a decreasing tendency is noticed towards high aerosol concentration for graupel/hail melting (Figure 7c) process. Actually, this pattern reflects the conversion rate of the riming process (Figure 7c). Essentially, graupel/hail melting has a substantial contribution to the cold-rain production compared to snow.

[25] The cyclone with clean aerosol has a higher graupel mixing ratio and more rain formation. In a recent study [Mukhopadhyay et al., 2011], we have found that the cyclone intensity is strongly influenced by the vertical distribution of the graupel (mixed-phase) particle which is generally formed near the freezing level (around the middle of the troposphere). Middle-level heating is found to influence the cyclone core instability significantly. The higher heating due to latent heat release as a consequence of microphysical processes such as more rain formation and higher graupel formation [Lord et al., 1984], the clean aerosol regime produces more instability within the cyclone core, and the model simulates an intense cyclone as compared to the maritime-clean condition. On the other hand, the urban aerosol produces weaker instability mainly due to weaker heating resulting from lesser latent heat due to reduced precipitation and graupel formation mechanism. It is interesting to note that during the ingestion of enhanced CCN, precipitation was reduced due to suppressed collision and coalescence, as depicted in Figures 3, 7d, and 7h. Subsequently, greater amounts of condensate were thrust into supercooled levels, where the freezing of droplets results in greater latent heating, as shown in Figures 4a and 4c. Convection thereby intensified, resulting in enhanced rainfall from the maritime-clean environment to the clean environment (Figures 2a and 2b) with more vigorous updrafts. This as such demonstrates that the monotonically increasing level of aerosol concentration cannot uniformly increase or decrease the convective intensity, but rather, it depends on the response of the microphysical processes. This result is consistent with earlier works of Cotton et al. [2012], Carrió and Cotton [2010], and Zhang et al. [2009]. Therefore, one may argue that invigoration in both sides of the center appears first and then gets reduced, which again supports the concept of “tipping point” for the enhancement/suppression of TC intensity.

3.3 Tropical Cyclone Bijli

[26] A low-pressure area developed over the southeast Bay of Bengal on 13 April 2009. It concentrated into a depression at 0900 UTC of 14 April over the southeast and the adjoining east central Bay of Bengal. It moved in a north northeasterly direction and intensified into cyclonic storm Bijli at 1200 UTC of 15 April over the east central Bay of Bengal. It thereafter moved in a northerly direction and finally recurved northeastward toward the Bangladesh coast. It gradually weakened prior to landfall and crossed the Bangladesh coast close to the south Chittagaon around 1600 UTC of April 2009. It caused moderate damages over Bangladesh.

3.4 Simulation of TC Bijli

[27] Being a weaker system, we started the integration of tropical cyclone Bijli from 15 April, 0000 UTC, and ran the model for 60 h with different aerosol perturbations similar to the TC Aila simulation. The track, maximum surface wind speed, and track errors are shown in Figures 1d–1f, respectively. It can be seen that the change in aerosol concentration does not influence the tracks very much in the three different (Figure 1d) sets of experiments, as found in the other TC as well. There is a difference between the clean case and the other two cases. Although the clean aerosol produces a marginally better track, as seen earlier in the case of Aila, the simulated winds show (Figure 1e) a significant impact with the “clean” environment, producing higher (intense) wind speeds closer to observations. The track error is the least for the clean environment and higher for the other two categories. The error shows a sharp increase, particularly after 36 h of integration.

[28] Consistent with the higher instability shown by the “clean” aerosol perturbation experiment, the total surface rainfall (Figure 2b) is found to be higher in the “clean” aerosol as compared to the “average” or “urban” category. The findings of a “tipping point” are again found for CCN concentrations; for lower and higher CCN concentrations, precipitation will be decreased (Figure 2b), which is consistent with the previous TC Aila.

[29] Similar to the previous TC case, the high concentration of aerosol (urban) perturbation shows larger amounts of cloud water in terms of mass (Figure 4e) and number (Figure 3g), resulting in lower rainwater mass (Figure 4f). This indicates lower efficiency of precipitation formation for the TC Bijli. The domain- and time-averaged vertical profiles of cloud water and rainwater are also presented for the TC Bijli (Figures 4c and 4d). These results also explain the presence of more cloud liquid water mass at the upper level and less rainwater mass at the lower level due to higher CCN concentrations (Figures 4c and 4d), as shown in the earlier section. The vertically integrated mass of cloud water and rainwater is shown in Figure 6b for the TC Bijli. This shows that higher pollution reduces the precipitation formation processes as compared to the average or clean condition also for the TC Bijli.

[30] The cloud ice deposition (Figure 7e) increases at higher CCN concentrations by strong WBF processes as there are numerous smaller cloud drops (Figure 3) which evaporate faster and thus supply additional water vapor for ice particle growth. The greater cloud ice with the urban aerosol environment is consistent with the higher ice deposition in that aerosol regime. As there is a significantly high amount of cloud ice in higher CCN concentrations, their autoconversion produces more snow initiation and deposition (Figure 7f).

[31] Although riming of snow (Figure 7f) reduces at higher CCN concentrations (urban-type aerosols) due to lesser size, snowmelting (Figure 7f) increases with higher CCN concentrations because of greater snow production. The mixed-phase hydrometeor (graupel) is greater in the clean environment (Figure 6b), followed by average and urban aerosol environments. This analysis brings out the fact that a clean environment supports processes such as melting (Figure 7g), accretion/riming (Figure 7g), etc. The domain- and time-averaged vertical profiles of cloud ice, snow, and graupel/hail mixing ratio have also been shown (Figures 8d–8f), which show a similar feature as explained before. Therefore, the importance of the mixed-phase hydrometeors seems to be the key for modulating the cyclone intensity.

4 Conclusions

[32] A set of simulations for two recent tropical cyclones Aila and Bijli of differing strengths (strong and weak, respectively) are represented for a variety of aerosol perturbations ranging from very low (maritime clean), low (clean), through medium (average) to high (urban). One was of tropical storm strength, and the other was only a tropical depression. Our objective is to show the microphysical changes occurring within the cyclone environment due to aerosol perturbations and, subsequently, its impact on the simulated track and intensity of the cyclone. This idealized study is particularly aimed to examine the cloud process responses to a wide range of aerosol concentrations in a cyclone environment.

[33] Similar to the findings of Cotton et al. [2012], it is demonstrated that the aerosol perturbations hardly affect the simulated cyclone tracks. However, the simulated cyclone intensity varies with changing aerosol conditions. For both simulated cyclones, it is found that the precipitation efficiency decreases with the increase in aerosol number concentration. This decrease in efficiency of rain formation is attributed to the lowering in efficiency of the collision-coalescence process. The reduced rainwater in higher pollution gives rise to more cloud drops and more cloud ice deposition in the periphery of the cyclone.

[34] It is interesting to put the results of these research results in perspective with findings summarized by Rosenfeld et al. [2012]. Altogether, all these studies show the robustness of the concept of TC weakening by storm ingestion of high concentrations of CCN. These simulations included actual TC case studies in the Atlantic basin [Khain et al., 2009], idealized simulations of Atlantic basin TCs [Zhang et al., 2007, 2009; Carrió and Cotton, 2011], a case study simulation of a western Pacific basin TC [Krall and Cotton, 2012], and these simulations of Indian Ocean basin TCs. Moreover, these simulations varied markedly in the representation of microphysics, ranging from full-bin-resolved microphysics [Khain et al., 2009], to bin-emulating microphysics [Zhang et al., 2007, 2009; Carrió and Cotton, 2011; Cotton et al., 2012], to the simulations with bulk microphysics with a limited number of ice categories. Only the simulations in Cotton et al. [2012] exhibited a temporary surge in storm intensity during the early stage of aerosol ingestion into the storm when the enhanced CCN population entered the storm interior (eyewall and inner rainband). The robustness of these model results gives us confidence in stating there is a high probability that ingestion of high CCN concentrations in a TC will lead to weakening of the storm but has little impact on storm direction.


[35] The Indian Institute of Tropical Meteorology (IITM), Pune, is fully funded by the Ministry of Earth Sciences, Government of India, New Delhi. The authors express their gratitude to C. T. Cheng for scientific discussions, particularly about the model. The authors also thank an anonymous reviewer for his valuable suggestion. S.T. acknowledges support from the U.S. Department of Energy Regional and Global Climate Modeling Program. The Pacific Northwest National Laboratory is operated for the U.S. DOE by Battelle Memorial Institute under contract DE-AC06-76RLO1830. W. R. Cotton acknowledges support from the DoD Center for Geosciences/Atmospheric Research, Colorado State University, under cooperative agreement W911NF-06-2-0015 with the Army Research Laboratory. The authors thank two anonymous reviewers and the Editor for their suggestions and comments which have helped improve the manuscript.