902
JOURNAL OF APPLIED METEOROLOGY
VOLUME 43
Impact of Precipitation on Aerosol Spectral Optical Depth and Retrieved Size Distributions: A Case Study AUROMEET SAHA
AND
K. KRISHNA MOORTHY
Space Physics Laboratory, Vikram Sarabhai Space Centre, Thiruvananthapuram, India (Manuscript received 16 December 2002, in final form 20 December 2003) ABSTRACT A case study is presented on the impact of two isolated, strong thundershowers during a prevailing dry, sunny season on the spectral optical depths and inferred columnar size characteristics of atmospheric aerosols at a tropical station. Results show a remarkable decrease in the aerosol optical depth and change in the spectral slope after the rain. The scavenging was found to be dependent on the particle size distribution; the larger, supermicron particles were found to be removed faster during the first shower itself, even though it was of only moderate intensity, resulting in about a 64% decrease in the columnar mass loading. In the second shower, which was stronger and more widespread than the former, more of the submicron particles in the optically active submicron size range were removed, but the reduction in mass loading was very small. The effective radius decreased continuously and so too did the columnar mass loading (total aerosol volume). The data are used to estimate the apparent columnar scavenging efficiency. The inferred apparent scavenging efficiencies were ;57% and 68% for the aerosol columnar number density in the optically active submicron size range for the two events, whereas for the coarse aerosols (r . 0.5 mm), it was ;75% for the first event but insignificant for the subsequent event, in line with the pattern of mass loading. The prevailing synoptic conditions (continental air mass and dry weather) helped the atmosphere to ‘‘recharge’’ within about 1 week after the events (which removed more than 70% of the aerosol burden), unlike the case with extensive synoptic phenomena, like monsoons, for which the aerosol optical depths remain depleted over the entire season. This recharging is also dependent on the size distribution of aerosols; the fine and accumulation particles are replenished faster than the coarse particles.
1. Introduction Mesoscale weather phenomena, such as thundershowers or land–sea breezes, produce changes in the characteristics of atmospheric aerosols (to be precise, the aerosol particles) over short time scales either by removing them from the atmosphere or by spatially redistributing them. Aerosols (meant in this paper to represent the dispersed phase of the system) are removed from the atmosphere by the following two basic processes: 1) dry deposition or sedimentation and 2) wet removal. Of these, wet removal is more efficient, primarily because the falling speed of the precipitation greatly exceeds the dry deposition velocity of the particles. Wet deposition includes (i) nucleation scavenging or rainout (process taking place within the clouds), in which the aerosols either act as condensation nuclei in cloud formation or get attached to a cloud droplet by various processes (e.g., Brownian diffusion, phoretic processes) and then are removed as the droplet falls as precipitation; and (ii) impaction scavenging or washout Corresponding author address: Dr. K. Krishna Moorthy, Space Physics Laboratory, Vikram Sarabhai Space Centre, Thiruvananthapuram 695 022, India. E-mail:
[email protected]
q 2004 American Meteorological Society
(process taking place beneath the cloud base), in which aerosols are captured by the falling precipitation (Pruppacher and Klett 1978; Andronache 2003). Wet removal is the main mechanism that limits the residence time of surface and lower-tropospheric aerosols, particularly in the size range of 0.05–3 mm (Radke et al. 1980; Jaenicke 1984). There have been extensive simulation studies on the in-cloud and below-cloud scavenging of aerosols (e.g., Wang et al. 1978; Flossmann et al. 1985; Andronche 2003; Gonc¸alves et al. 2002, 2003), field measurements (e.g., Radke et al. 1980; Ebert et al. 1988, 1997; Dovgalyuk et al. 1995; Andronche et al. 2002), and laboratory studies (Beard 1974; Leong et al. 1982; Sinkevich et al. 1995; Pranesha and Kamra 1997). These have shown that the wet removal of aerosol particles depends on the size distribution and altitude profile of aerosols, rainfall rate, and raindrop size distribution. Several observational studies have also shown large depletion in aerosol concentration and extinction coefficient, both in the well-mixed layer as well as in the free troposphere; and a decrease in columnar optical depths and a change in aerosol size distribution associated with extensive and widespread (synoptic) rainfall, like those associated with the (Indian) monsoons (e.g., Subbaraya and Jayaraman 1982; Moorthy et al. 1991,
JUNE 2004
SAHA AND MOORTHY
903
1996; Pillai and Moorthy 2001). However, these impacts are long lasting (a few months) as against the short time scales involved in the mesoscale processes. Nevertheless, field studies are quite limited on the impact of strong, isolated precipitation events (mesoscale thundershowers) that occur during dry periods, particularly those that assess the column scavenging efficiency and the corresponding change in aerosol properties. Such studies are also useful in examining the ‘‘recharging’’ of the atmosphere (as it recovers from the impact) under favorable prevailing conditions. In this paper, we present the details of a case study of the impact of two isolated and strong thundershower events, occurring in short succession, during a long, dry season at a tropical station Trivandrum, India (8.558N, 778E; 2 m MSL). The columnar spectral optical depths and retrieved size distributions are examined for the distinctiveness of the impact in terms of aerosol size regimes and the extent of rainfall. The data are also used to understand the recovery from the impact under favorable prevailing conditions. 2. Experimental details and data a. Atmospheric aerosols The aerosol data used in this study consist of columnar aerosol optical depths (AODs), estimated at 10 wavelengths (380, 400, 450, 500, 600, 650, 750, 850, 935, and 1025 nm), using a 10-channel multiwavelength solar radiometer (MWR). The instrumental details, methods of analysis, and error budget are described in earlier papers (e.g., Moorthy et al. 1997, 1999, 2001; Satheesh and Moorthy 1997). Spectral AODs were estimated regularly using the MWR on clear or partly clear days (during periods of unobscured solar visibility) as part of a long-term program, and the data during the month of March 2002 are used for this study. The actual observation site, located at ;500 m due east of the Arabian Sea coast and ;6 km northwest of Trivandrum city, is a rural coastal location that is far removed from any major industrial activities (Pillai and Moorthy 2001). b. Prevailing (synoptic) meteorological conditions The prevailing meteorological conditions at Trivandrum during January–March comprise a synoptic northeasterly circulation (constituting a continental air mass), dry ambient with low relative humidity (RH; ;50%– ;70%), and scant rainfall, modulated regularly by the diurnal land–sea-breeze circulations (Prakash et al. 1992; Asnani 1993; Pillai and Moorthy 2001). This season is also characterized by generally clear and cloudfree sky conditions and an absence of any major weather phenomena. Figure 1 shows the distribution of daily rainfall at Trivandrum for the months of February and March 2002, which is typical to this location. There
FIG. 1. Distribution of daily total rainfall during Feb and Mar 2002 at Trivandrum.
were only four rainy days from 1 February to 31 March of 2002; the first occurred on 2 February, when there was a thundershower with ;14 mm of rainfall. Subsequently, a long, dry spell prevailed for over 40 days followed by two thundershowers in quick succession— the first on 15 March and the second, stronger one on 17 March. Both events were from ;1 to 2 h in duration. This was followed again by a dry spell until 27 March when there was a brief spell of weak rainfall. Because the MWR observations were available rather regularly, we have selected the period from 11 to 26 March 2002 for this study. c. Genesis and evolution of the thundershower events Meteorological measurements made at and around the observation site are used to characterize the thundershower events. Hourly rainfall data were available for the observation site from the nearby meteorological facility. More detailed data on surface and upper-air meteorological parameters, as well as the regional distribution of rainfall, were obtained from the Meteorological Centre in Trivandrum (MC-T), of the India Meteorological Department. These included hourly values of wind, air temperature, RH (at the surface), rainfall, the routine weather observations made by the observers, and regular upper-air radiosonde data at 0000 and 1200 UTC [UTC is 5.5 h behind local time (LT)]. Information on the regional distribution of rainfall was obtained from a chain of rain gauge stations maintained by MC-T, and provided the daily total (but not time resolved) rainfall
904
JOURNAL OF APPLIED METEOROLOGY
VOLUME 43
TABLE 1. Summary of weather observations at the observation site and MC-T. CBH: cloud-base height. Rainfall OS Date 15 Mar
16 Mar 17 Mar
Nature and evolution of clouds Isolated stratus clouds in the morning, cumulus activity around noon, evolving into cumulonimbus with medium CBH in the afternoon, followed by a thunderstorm, showers, and drizzle Similar to 15 Mar Cumulonimbus clouds with low CBH persisting, followed by severe lighting and thunder and heavy showers, culminating in drizzle and mist
MC-T
Duration
Amount (mm)
Duration
Amount (mm)
1555–1705 (70 min)
19.0
1530–1620 (50 min)
2.7
1500–1505 (5 min) 0245–0345 (60 min)
0.2
1550–1640 (50 min) 0315–0520 (125 min)
9.4
(during the previous 24 h), in accordance with the standard meteorological conventions. A summary of the day-to-day weather, cloud types, and rainfall for the event days is given in Table 1, based on the observations available at the site and at MC-T. The diurnal variation of air temperature and ambient RH are shown in Fig. 2, for the event days in comparison with the mean normal day values (shown by the continuous lines with vertical bars). The points are hourly values of the parameter plotted. The normal day pattern is obtained as an average of the hourly values on 4 days (2 days preceding the first event, and 2 days following the second event), and the vertical bars are the standard deviations. The duration of the thundershower events
FIG. 2. Diurnal variation of (top) air temperature and (bottom) ambient RH for the event days in comparison with the mean normal day pattern (continuous line with vertical bars). The points are hourly values of the parameter, and the vertical bars are the standard deviation (details are given in the text).
27.7
17.7
are marked by horizontal lines in the figure, with the dashed one representing the second event. The actual details and amount of rainfall at the observation site and at MC-T are given in Table 1, which indicates that the rainfall was not spatially homogeneous, even at the short spatial scales involved. The spatial distribution of rainfall is examined in Fig. 3 using data from the network of rain gauges. The locations of these are marked on the figures, along with the observation site (OS), MC-T, and the seven rain gauge stations around the site (RG-1– RG-7). Referring to the prevailing synoptic winds, the stations RG-1, RG-2, and RG-5 lie upwind of the observation site, while RG-6 and RG-7 are downwind. The continuous line shows the local coastline with the Arabian Sea to its left, over which no rainfall measurements are available. On a mesoscale, the terrain gradually slopes upward, inland reaching a height of ;1 km at the location of RG-1. Figure 3 clearly shows that the spatial distribution of the rainfall was distinct for the two events and spatially heterogeneous. During the first event (15 March), the thundershower was rather localized, with maximum rainfall occurring around the observation site, decreasing rapidly upwind. On 17 March, however, not only was the rainfall more intense at the observation site (as compared with the first) but severe thundershowers occurred upwind also, with the station RG-1 recording as much as ;64 mm of rainfall on that day. However, the duration and actual time of occurrence of these events at the rain gauge stations are not available. Nevertheless, the data also revealed that, other than the events shown in Fig. 1, no rainfall was recorded at any of these rain gauge stations during March 2002. Vertical profiles of RH, obtained from the regular radiosonde ascents from MC-T, are shown in Fig. 4. The left-hand side compares the RH profiles at 1200 UTC on 14 and 15 March, while the middle and right-hand side show the profiles at 0000 and 1200 UTC for 17 and 18 March. The profiles on 18 March are typical of clear days—devoid of any thunderstorm or extensive cloud buildup where the RH decreases rapidly with al-
JUNE 2004
SAHA AND MOORTHY
905
FIG. 3. The spatial distribution of daily total rainfall for (bottom) 15 Mar and (top) 16–17 Mar. The locations where actual measurements were made are marked on the figures, with OS being the observation site, MC-T being the Meteorological Center at Trivandrum, and RG-1–RG-7 being the rain gauge stations around the site. The continuous line shows the local coastline, with the Arabian Sea to its left.
titude. The genesis and evolution of the thundershowers are examined in the light of all of the above factors (Figs. 2–4 and Table 1). On the morning of 15 March, the sky was clear with a little cloud cover (stratus clouds, 1 octa) until ;1130 LT, after which cumulus clouds started developing, which were later followed by convective cumulonimbus clouds. The thunderstorm activity started around ;1330
LT, associated with a rapid decrease in air temperature by ;58–68C and a nearly simultaneous increase in RH by ;20%. The rainfall occurred at ;1600 LT (both at MC-T and the observation site), but there was a large difference in the amount of rainfall at these two stations (Table 1 and Fig. 3, bottom). The intensity was highest at and around the observation site, but at MC-T (;6 km away) the rainfall was meager. However, station RG-
906
JOURNAL OF APPLIED METEOROLOGY
VOLUME 43
FIG. 4. Vertical profiles of RH for the period of 14–18 Mar 2002. (left) The RH profiles at 1200 UTC for 15 Mar are compared with that of 14 Mar. (middle and right) The profiles at 0000 and 1200 UTC for the other days are shown (17 and 18 Mar). The profile for 16 Mar is not shown, because the balloon data were not available.
2 (farther away, but upwind) recorded rainfall of moderate intensity. The vertical profile of RH at 1200 UTC on that day showed significantly higher values of RH (;75%) in comparison with the previous day (or in comparison with 18 March). This RH profile showed little altitude variation up to ;5 km (600 hPa), indicating deep convection, before decreasing steeply at higher altitudes. The morning of the next day (16 March) had scattered stratocumulus, 1–2 octa, until ;1130 LT. The subsequent development of the thundershower was quite similar to the previous day (Table 1 and Fig. 2). On this day the upper-air data are not available because the balloon flight was aborted shortly after lift-off. However, the rapid decrease in the air temperature and increase in RH associated with the convection is seen in Fig. 2. On the late evening of 16 March, there was lightning and thunder activity, which then developed into a thunderstorm by midnight (Table 1). On the early hours of the following day (17 March), there was another spell that continued until ;0530 LT. The rainfall on this day was quite intense and widespread (Fig. 3, top), with the intensity increasing rapidly upwind. The high ambient RH (;90%) on the evening of 16 March continued throughout until the morning of 17 March. The vertical RH profile at 0000 UTC 17 March shows very high RH up to ;6 km (500 hPa). Subsequently, though the rain stopped, the sky continued to be cloudy (stratocumulus clouds, 3–4 octa). By 18 March, the conditions reverted
to normalcy, as shown by the diurnal variation of temperature and RH in Fig. 2 and the RH profiles (at 0000 and 1200 UTC) shown in Fig. 4. 3. Results Figure 5 shows the day-to-day variations in AODs at two representative wavelengths (500 and 750 nm) for the period of 11–26 March 2002, with the distribution of daily total rainfall at the observation site superposed using vertical bars. The AOD variations show several interesting features. Prior to the rain event, the AODs remained moderately high during 11 and 12 March. This was then followed by a significant increase on 13 March and a further increase to a peak on the morning of 15 March, when the AODs were quite high–1.02 at 500 nm and 0.6 at 750 nm—and the sky turned grayish and hazy. In the afternoon of 15 March, the station experienced the first thundershower lasting for ;1 h, during which time 19 mm of precipitation fell. This event was most pronounced around the observation site, and rapidly decreased away from it (Table 1 and the bottom of Fig. 3). Subsequently, the MWR observations could be made only in the morning of 16 March, when a remarkable decrease in the AOD is seen. The next thundershower occurred in two spells (Table 1) during which time the station received ;28 mm of rainfall. This event was more intense and widespread (Fig. 3, top). However,
JUNE 2004
907
SAHA AND MOORTHY
FIG. 5. Day-to-day variations of AOD at two wavelengths: (top) 500 and (bottom) 750 nm. The points are the daily mean AOD, and the vertical bars through them are the standard deviation. The two shaded long bars depict the thundershower events, the height of the bar being equal to the daily total rainfall (mm).
because the sky continued to be cloudy on this day, the MWR data are taken only on 18 March, when the sky was very clear and bright blue. There was a substantial reduction in the AODs. Interestingly, the AODs continued to decrease on the next day too, even though there was no further precipitation, before starting the recovery. Spectral dependence and Angstrom parameters In order to investigate how the impact affects the aerosol size distribution, the AOD spectra on 15, 16, and 18 March are examined in Fig. 6. The most striking feature is that the effect is not uniform at all the wavelengths. After the first event, there was a significant decrease in AOD at all the wavelengths, but relatively more at the longer ones, while after the subsequent event the decrease was more pronounced at the shorter wavelength side. Because the AOD spectra depend on the aerosol size spectrum, this suggests that the wet removal process on the two days has affected the different aerosol sizes differently. The simplest way to examine this is to compute the Angstrom parameters (Angstrom 1961) a and b, which connect AOD (represented by t p ) to the wavelength (l) through the relation
t p 5 bl 2a ,
FIG. 6. Spectral variation of AODs for 3 days: before the event (15 Mar), after the first event (16 Mar), and after the second event (18 Mar). The vertical bars are errors in the daily mean AOD.
particles. Angstrom parameters a and b were computed for the individual AOD spectra (shown in Fig. 6) by evolving a least squares fit to Eq. (1) in each case. The mean variance of the regression slope (Kleinbaum and Kupper 1978) is used to evaluate the standard deviation da of a. The values of a, da, and b are given in Table 2 for the three different cases, along with the correlation coefficient (g ) between ln(t p ) and ln(l). In all of the cases g is quite high and significant at P 5 0.05 or better (Fisher 1970), indicating a good fit to Eq. (1). Nevertheless, the higher value of da (and lower value of g ) for 18 March, in comparison with the other days, indicates that the AOD spectrum on that day deviated more from Eq. (1), in comparison with the previous days, because of the change in the size spectrum resulting from washout, as will be seen later. [The values of a and b prior to the start of the event were quite similar to the values generally seen during this period (in the range of 1–1.4) at this location. The mean values of a and b were, respectively, 1.24 and 0.18 for the period of 11–13 March and 1.25 and 0.26 for the period of 21–25 March 2002.] Both a and b are high on 15 March, the control day, with b as high as ;0.5. After the first thundershower (which occurred that evening at ;1555 LT and lasted TABLE 2. Angstrom parameters.
(1)
where a is the wavelength exponent and b is the Angstrom coefficient. Here, a is a measure of the relative dominance of small particles, while b is a measure of the aerosol loading and is more associated with the large
Date 2002
a 1 da
g
b
15 Mar 16 Mar 18 Mar
1.03 6 0.06 1.21 6 0.06 0.81 6 0.21
0.98 0.99 0.81
0.50 0.21 0.10
908
JOURNAL OF APPLIED METEOROLOGY
TABLE 3. Angstrom parameters for ‘‘washout AOD spectra.’’ Date 2002
a 1 da
g
b
15–16 Mar 16–18 Mar
0.88 6 0.11 1.62 6 0.23
0.94 0.93
0.29 0.10
for ;70 min, yielding a precipitation of 19 mm), the MWR data taken on the next morning revealed that the AOD spectrum has changed significantly with a higher value of a (increased from 1.03 to 1.21) and a lower value of b (decreased from 0.5 to ;0.2), besides an overall decrease in the AOD at all wavelengths (Fig. 6). This would mean an overall reduction in the aerosol burden, but particularly a larger reduction at the coarse particle end of the size spectrum (which contributes to b). This relatively larger decrease in b has resulted in a steeper AOD spectrum and the higher value of a on 16 March. The next thundershower occurred in parts on 16–17 March and produced 28 mm of rainfall. However, the clouds did not dissipate on 17 March and, thus, MWR could not be operated on that day. The next day (18 March) was clear and cloud free, and the AOD spectrum on this day is fairly flat with a 5 0.8, implying a substantial removal of fine, submicron aerosols. The decrease in b is comparatively much smaller. This can also be inferred by estimating the difference between the successive AOD spectra (i.e., the difference between AOD spectra before and after the shower, for the two events, treating them as independent) and estimating a and b of the ‘‘washed out’’ AOD spectra. These are given in Table 3, which shows a higher value of b for the washed-out component of the first event (as compared with the second) and a higher value for a for the second event, in line with the above discussion. The large removal of aerosol burden at the fine particle sizes resulted in the flatter AOD spectrum, deviating more from Eq. (1), with a higher value of da on 18 March. 4. Discussion The impact of the wet removal of atmospheric aerosols is very important not only because it is the main sink of aerosol particles, but also because it will help us to understand the effects of natural meteorological processes in modifying the aerosol properties and effects. Notwithstanding its significance, field experiments are quite limited (Radke et al. 1980; Pranesha and Kamra 1997; Devara et al. 1997; Andronache et al. 2002) and rather scarce on the impact on columnar AODs, particularly because of the difficulty in getting good data. Our study showed a considerable decrease in the AODs over a wide range of wavelengths associated with the events. Further, it showed a change in the AOD spectral shape, as shown by the a and b values; the spectra become steeper after the first rain event, because of the removal of comparatively large abun-
VOLUME 43
dance of supermicron (.1 mm) aerosols (resulting in a large decrease in b). The second event, stronger and extensive, brought about a much larger decrease in the small particle concentration as evidenced by the high value of a of the washed-out AOD spectrum of the second event in Table 3, but the decrease in b was small, because of the already reduced abundance of coarse particles. These result in a flatter AOD spectrum. Because the events followed in quick succession, the landbased production mechanisms (e.g., wind-blown dust and convective/turbulent mixing) would not have recovered from the impact of first event, which had left the land surface wet. The weak, continental prevailing winds during the season are not conducive for producing a substantial amount of coarse sea-spray aerosols or lift dust particles (from land) to replenish those already removed. Consequently, the coarse particle abundance remained low. However, the prevailing winds would be able to replenish, at least partly, the loss of submicron aerosols by advection from inland areas, not affected by the rainfall, during the period between the rain’s end and the next MWR measurements. To examine these aspects in more detail, we retrieved the columnar size distributions (SDs). a. Retrieved size distributions Columnar SDs [n c (r)] are obtained from the respective AOD spectra by numerically inverting the integral equation
t p (l) 5
E
r2
p r 2 Qext (r, l, m*)n c (r) dr,
(2)
r1
where Qext is the aerosol extinction efficiency factor that depends on the aerosol complex refractive index (m*), radius (r), and wavelength (l) of incident radiation; n c (r) represents the columnar number density of aerosols in an infinitesimal radius range dr centered around r; and r1 and r 2 , respectively, are the lower and upper radii limits that contribute significantly to Qext for the range of wavelengths used in the MWR. There are several methods available in the literature to retrieve the columnar size distribution from the AOD spectra. These include the constrained linear inversion (CLI) technique (e.g., King et al. 1978) and the randomized minimization search technique (RMST) (e.g., Anderson et al. 2000), both of which are being widely used in the field. Both of these methods essentially involve iterative inversion of the integral Eq. (2) within a finite radius range r1 and r 2 of the particles and physically meaningful constraints (such as the positivity criterion) to retrieve a columnar size distribution, which will yield an AOD spectrum that matches the spectrum obtained from the measurements well within the measurement errors. The application of these two techniques to retrieve the microphysical properties of aerosols from the simulated Stratospheric Aerosol and Gas Experiment
JUNE 2004
SAHA AND MOORTHY
(SAGE)-III measurements is demonstrated in detail by Anderson et al. (2000) who have shown that both the methods with their own size ranges are capable of retrieving the microphysical characteristics of aerosols. They also conclude that both the CLI and RMST techniques can obtain reliable aerosol information and can also be used to estimate extinction at other wavelengths with reasonable accuracy. We have considered this and used the constrained linear inversion technique in the present study. This technique provides fairly accurate information of the aerosol size distributions if the Lagrange multiplier, refractive index, and radius range are carefully selected (King, 1982), particularly when the AOD spectra do not show large oscillations. The application of this technique to the MWR data and the constraints used are described in earlier papers (Moorthy et al. 1991; 1997). The selection of inner and outer radii limits [r1 and r 2 in Eq. (2)] strongly depends on the shortest and longest wavelengths of the MWR [at which t p (l) are measured], and as such values of 0.05 and 3.0 mm were set, respectively, for r1 and r 2 . The wavelength-dependent complex refractive index m* of the aerosols depend on the aerosol chemical composition. However, because this was not available to us (because of the absence of a facility), we choose the values given by Lubin et al. (2002) based on the chemical analysis of aerosol samples collected at the Kaashidhoo Climate Observatory (KCO) in Maldives, lying ;500 km south (downwind) of Trivandrum, during the Indian Ocean Experiment (INDOEX) field phases as the best alternative. Because the KCO measurements correspond to the January–March period of 1998 and 1999, and the location is downwind of our site, these refractive indices are considered to be representative of the aerosol environment over Trivandrum also, particularly when there are no prominent aerosol sources downwind and the air mass is continental in nature. It is likely that different aerosol species would be having different size distributions and, hence, the refractive index might vary with size. However, here we considered the aerosol particles as an internal mixture, in line with Satheesh et al. (1999) and Lubin et al. (2002), and as such, these values are considered representative of the composite aerosol. Moreover, we restrict our attention to a narrow size range of the broad aerosol spectrum. Continuous measurements of aerosol black carbon (BC) at Trivandrum (Babu and Moorthy 2002) have shown that the mass fraction of BC to ambient aerosols (which effectively determines the imaginary part of the refractive index) is ;10%–12% during this season, a value very close to that reported during INDOEX (Lelieveld et al. 2001) and used by Lubin et al. (2002). Besides, King et al. (1978) and King (1982) have shown that the information content in the size distribution retrieved by the constrained linear inversion method is not highly dependent on the refractive index. The size distributions, thus retrieved (after a typical five iterations), are shown in Fig. 7 (bottom panel) in
909
FIG. 7. (bottom) Columnar size distributions for 15, 16, and 18 Mar, and (top) the corresponding AOD spectra in which the points with error bars correspond to the MWR-estimated values and the lines are the AODs reestimated from the inverted SDs.
a log–log scale for 3 days: the control day (15 March), after the first event (16 March), and after the second event (18 March). On the top panel of Fig. 7, the corresponding AOD spectra are shown; the points with error bars correspond to the MWR estimated values, and the lines are the AODs reestimated from the inverted SDs (bottom). All of the inverted SDs are bimodal, with a conspicuous coarse mode at r ø 0.9 mm, and an accumulation mode (which is only indicated) at r ø 0.1 mm. The coarse mode is mostly attributed to local processes, which would also include the sea-spray aerosols, wind-blown dust and aerosols from vegetation, while the accumulation aerosols would be mostly the secondary aerosols that have undergone coagulation and condensational growth, and also those advected from inland. The bimodal nature is also an artifact of the inversion procedure, where the radius range of the aerosols are constrained by the wavelength range over which AOD measurements are made (380–1025 nm in this case) with the MWR. It is quite likely that a third, smaller mode (the nucleation mode) would also be present, particularly at higher altitudes when RH also is high. Presence of such a nucleation mode has also been observed during INDOEX cruise measurements (Jayaraman et al. 2001). However, these particles are too small
910
JOURNAL OF APPLIED METEOROLOGY
VOLUME 43
to contribute significantly to Qext in Eq. (2), even for the shortest wavelength (380 nm) at which the AOD is measured and, hence, would not be present in the inverted distributions. Notwithstanding this, the change in the SD after each event is seen clearly in Fig. 7. After the first rain, there is a remarkable decrease in the coarse mode regime (r . 0.5 mm, diameter . 1 mm) and in the very fine particle regime (r , 0.1 mm), while in the optically active submicron regime (0.1 mm , r , 0.5 mm) the decrease is not that dramatic. The impact of the second event, however, is most discernible in this range, rather than in the fine/coarse ranges. To quantify these changes, the following physical parameters are deduced from the inverted size distribution: column concentration of coarse aerosols,
E
Nc 5
3.0
n c (r) dr,
(3)
0.5
and that in the optically active submicron range, Nop 5
E
0.5
n c (r) dr,
(4)
0.1
weighted mean radius,
E E
r2
R5
n c (r)r dr
r1
,
r2
(5)
n c (r) dr
r1
the effective radius,
Reff 5
E E
r2
n c (r)r 3 dr
r1
,
r2
(6)
n c (r)r 2 dr
r1
and the columnar mass loading, 4 M L 5 pr 3
FIG. 8. Day-to-day variation in aerosol parameters deduced from the retrieved SDs for the period of 15–19 Mar 2002. (top) The variation of columnar mass loading (M L ), (middle) the variation of effective radius (Reff ) and weighted mean radius (R), and (bottom) the variation of number concentration of optically active submicron aerosols (Nop ) and coarse aerosols (N c ) are shown.
E
r2
n c (r)r 3 dr,
(7)
r1
where a value of 2.25 g cm 23 was used for the particle density r, considering a wide range of values reported for aerosol density by different workers (e.g., Pruppacher and Klett 1978). However, the absolute value of r is not very significant in this study as long as r is considered to be constant before and after the events. [Strictly speaking, this is a debatable assumption, because the wet removal is known to be species specific (e.g., Flossmann et al. 1987; Andronache 2003) where hygroscopic aerosols, like sulfates, are more efficiently removed than the species like soot, and as such there could be small changes in r before and after the rain events. This is not considered here because of the ab-
sence of information on the composition of aerosols in the atmosphere, before and after the event.] The variations of the retrieved parameters are shown in Fig. 8. It shows that the large removal of the aerosol mass (M L ) in the first rain (Fig. 8 top) is mostly associated with the scavenging of coarse aerosols as can be seen from the decrease in N c (bottom of the same figure). Consequently, Reff (the ratio of volume to area) decreases (middle, Fig. 8) from ;0.36 to 0.30 mm. This shows that the aerosol size spectrum is now weighed more toward the smaller size. In the subsequent rain, the reduction in the mass loading is small mainly because the washout this time was confined primarily to the submicron (optically active) regime, as can be seen from the continued decrease of Nop in the bottom of Fig. 8. Consequently, Reff decreased further to 0.21 mm, but the decrease in M L was quite small. This is due to the decrease in concentration of the coarse particles due to the first rain and also due to a change in their altitude distribution, because most of the particles in the upper levels would have been removed by nucleation and impaction scavenging of the first rain. Gonc¸alves et al. (2003) have shown that wet scavenging is influenced by the altitude profile of aerosols and the vertical development of clouds. They also observed that the incloud scavenging dominates over below-cloud scavenging. The surface production, and subsequent vertical
JUNE 2004
911
SAHA AND MOORTHY
transport by convective mixing, would be less efficient on the next day because of the wet land conditions and a higher apportionment of the incident solar radiation to latent heat flux than to sensible heat (due to the evaporation of the increased surface soil moisture). Thus, most of those coarse particles that remained in the atmosphere after the first rain would be confined more to the lower atmospheric regions, while at cloud-level heights their concentration would be largely depleted. As such, nucleation/in-cloud scavenging will be insignificant for these particles during the second event. On the other hand, the fine particles do not easily get scavenged because they fall in the ‘‘Greenfield gap.’’ A substantial fraction of these would have remained as cloud interstitial particles too. This would be particularly true during the first event, for which the rate of rainfall was moderate (,20 mm h 21 ), and spatially limited. Based on analytical estimates of the below-cloud scavenging coefficient of atmospheric aerosols with different size distributions, Andronache (2003) has shown that impaction scavenging of accumulation aerosols becomes significant only during strong precipitation but is insignificant when the rate of rainfall is small. The impact of the second thundershower is to be examined in the light of the above. This event was quite intense and widespread, the rainfall exceeding 27 mm h 21 at the observation site, and was much more intense upwind (Fig. 3). This has brought in substantial wet removal by both in-cloud and below-cloud (impaction) scavenging processes of even the accumulation particles. This resulted in a considerable depletion in the aerosol loading at the site and even more at the upwind locations where the rainfall intensity was about 3 times as much. Over the land, this would result in vast areas that are wet and nonconducive for aerosol (mainly coarse) generation for a few days to follow. Thus, even though there was no further precipitation on 18 March, the concentrations and mass loading (Nop and M L ) continued to decrease, because the local meteorological conditions were not conducive for extensive aerosol production over the land on the one hand, while the upper-air synoptic winds continued to advect the aerosol-depleted (due to high scavenging by the previous day’s intense rainfall) air mass from upwind on the other hand. This resulted in a change in the aerosol size distribution with a decrease in the relative dominance of submicron aerosol concentration, decrease in a, and a consequent increase (though small) in Reff and R, even while the overall loading decreases. This aspect is examined by obtaining an ‘‘apparent scavenging efficiency’’ S c as Sc 5
Ps , Po
(8)
where P s is the particular parameter of aerosol (mass or number) that has been scavenged (removed) and P o is its value before the rain (scavenging process). Using Eq. (8), the scavenging efficiency for columnar mass
TABLE 4. Apparent scavenging efficiency for columnar mass loading and number concentrations.
Parameter (P)
Event 1 (15–16 Mar 2002)
Event 2 (16–18 Mar 2002)
Po
Ps
Sc
Po
Ps
Sc
ML N c (1010 )* Nop (1012 )
565 3.29 3.91
361 2.48 2.27
0.64 0.75 0.57
204 0.81 1.64
35 — 1.12
0.17 — 0.68
* For Nc there was a small increase after the second rain; hence S c was not computed.
loading (M L ), columnar number concentration of aerosols in the optically active submicron (Nop ), and coarse (N c ) size regimes are estimated following the criteria used in Eqs. (3) and (4) earlier. In Table 4, these coefficients are given for the two events separately. These apparent scavenging efficiencies will be lower than the true values, because the local and synoptic meteorological processes would have partly replenished the loss during the time between the rain’s end and the next observations. Nevertheless, the findings in Table 4, in general, support the earlier arguments. The apparent mass scavenging efficiency was quite large (0.64) for the first event, while for the second event it was only 0.17, irrespective of it being stronger. This is closely associated with the very high value of S c (0.75) for the concentration N c of coarse particles during the first event, whereas it was negligible during the second. In contrast, in the submicron aerosol regime, the scavenging efficiency remained rather steady and high (0.57 for the first and 0.68 for the second). In fact, the efficiency was higher for the second event, which was stronger and extensive, because of the more efficient scavenging of accumulation aerosols by the intense precipitation both locally and upwind. Based on analytical model computations, Andronache (2003) has reported that the below-cloud scavenging (BCS) rate increases with the rate of rainfall and is strongly dependent on the aerosol size distribution. For low to moderate rainfall (0.1–10 mm h 21 ), the BCS is important only for very small particles (r , 0.005 mm) and for coarse particles with r . 1 mm, whereas for aerosols in the inbetween size range it is negligible. Consequently, accumulation aerosols dominate the below-cloud aerosol size distribution. However, for extensive/intense rainfall BCS becomes important for these particles also and becomes comparable to in-cloud scavenging. Nevertheless, its effect on the mass scavenging efficiency is not very high because of the very small contribution of these particles to the total aerosol mass. Field observations (Hobbs 1993) and model calculations (Flossmann et al. 1985) showed that nucleation scavenging (in-cloud scavenging) results in ;48%–94% removal of particle concentration (depending on the particle composition) and that this reduction occurs mostly at particle sizes .0.1 mm. Flossmann et al. (1985) have also shown that in-cloud scavenging is the major process of removal of aerosol mass,
912
JOURNAL OF APPLIED METEOROLOGY
even though below-cloud scavenging might contribute to a much-reduced extent. Based on model simulations of in-cloud and below-cloud scavenging of hygroscopic particles, supported with measurements from Brazil, Gonc¸alves et al. (2002) reported that the in-cloud scavenging accounts for 56%–97% of the total aerosol scavenged. From aircraft measurements of aerosol size distribution in aged air masses at cloud levels, Radke et al. (1980) reported higher precipitation-scavenging efficiencies for large (coarse) as well as for fine (nucleation mode) particles, the efficiency of which is raindrop size distribution dependent. Based on laboratory simulations, Pranesha and Kamra (1996, 1997) have reported that the scavenging efficiency increases with the increase in the size of aerosol particles and decrease in the size of raindrops. They also showed that the collection efficiency increases in the presence of electric fields, which generally occurs in thunderclouds. Our observations are consistent with all the above. In the coarse regime ;75% of the column concentration was removed during the first event. Most of this would have resulted from in-cloud scavenging and, hence, there will be a large reduction in the concentration of these particles at the cloud-level altitudes. Because there was not enough time to supplement this loss before the second event (for reasons stated earlier), further removal by nucleation scavenging would be insignificant. Because these aerosols are chief contributors to aerosol mass (volume), the reduction in M L in the second event is only ;17% (which was mainly due to the removal of accumulation aerosols) as can be seen from Table 4. These smaller aerosols were removed comparatively less efficiently in the first rain, which was only moderate. Within the cloud region, a significant amount of these particles can remain as cloud–interstitial particles (Pruppacher and Klett 1978) also. It is also known that particles in the size range from 0.05 to 1 mm are scavenged less effectively in the absence of external forces (such as electric forces). In the discussion that had forgone, it is to be borne in mind that the number concentration and mass loading parameters retrieved from the inverted size distribution function serve only as indicators and are not based on absolute measurements. Nevertheless, they serve as useful tools for an assessment of the impact of scavenging, as long as the constraints used in the inversion procedure are unaltered (as in the present study) between the cases. It is also possible that the scavenging is species dependent. However, because there are no major aerosol sources around the observation site, the aerosols over this region can be considered as old, well-mixed aerosols so that this effect can be considered insignificant. Using model simulations and field data during INDOEX, Andronache et al. (2002) have studied the effect of deep convection on aerosol under typical intertropical convergence zone (ITCZ) conditions over the Indian Ocean. They reported that the scavenging efficiency of fresh carbonaceous particles is lower than that of sul-
VOLUME 43
FIG. 9. Evolution of the AOD spectra for the period of 19–25 Mar 2002 as the atmosphere recharges. The vertical bars are errors in the daily mean values. Initially AODs increase rapidly at the shorter wavelengths (l , 650 nm), while at longer wavelengths the increase occurs rather gradually.
fates but their removal becomes efficient once they are sufficiently aged and mixed with other hydrophilic aerosols. b. ‘‘Recharging’’ of the atmosphere Subsequent to the thundershowers, the dry synoptic conditions continued, and there was no further rain for several days in and around the study area to the extent covered by the rain gauge stations in Fig. 3. This allowed the atmosphere to recover from the effect of perturbations because the aerosol generation/transport mechanisms continue as they were before the events. Thus, the atmosphere gets ‘‘recharged’’ with aerosols and eventually returns to the ‘‘normal’’ state that prevailed prior to the event that is typical to that season. An estimate of the recharging time can be had from Fig. 5 itself, which shows the subsequent evolution of the AODs. After the rain on 17 March, the AODs continued to decrease (though very slightly) for one more day, for reasons already discussed. Then they started recovering and reached fairly high levels, comparable to the climatological values as well as those that prevailed before the rain, by 23/24 March 2002, that is about 1 week after the final and strong perturbation was over. The progress with time of this ‘‘recharging’’ can be seen in Fig. 9, where the time evolution of the AOD spectra is
JUNE 2004
TABLE 5. Evolution of a and b as the atmosphere recharges. Date 2002 19 21 23 24 25
Mar Mar Mar Mar Mar
913
SAHA AND MOORTHY
a 1 da
g
b
6 6 6 6 6
0.36 0.99 0.99 0.99 0.89
0.09 0.14 0.24 0.33 0.32
0.28 1.33 1.30 1.27 1.08
0.19 0.12 0.05 0.14 0.08
shown for the period from 19 to 25 March 2002. Further, the values of a 6 da, b, and g of the AOD spectra are also estimated as described earlier and the results are given in Table 5. It is interesting to note that the rapid recovery of AOD occurs initially at the shorter wavelengths (l , 650 nm). After the extensive rainfall and the consequent washout of accumulation aerosols, the AOD spectrum continued to remain flat on 18 and 19 March, for reasons discussed above (weak advection from upwind regions where the depletion would have been much larger) before the recovery started. The longer wavelengths recovered much later. This is primarily because much larger buoyant forces are required to lift heavier aerosols into the atmosphere, which become available as the dry, sunny periods continue to be longer. Moreover, for the heavier particles, local production mechanisms are of more significance, while smaller particles can be quickly advected from further inland regions, which are not severely affected by the rain, by the prevailing winds constituting the continental air mass, which is known to be submicron aerosol dominant (Pillai and Moorthy 2001). They can also form in situ by secondary production processes at higher regions of the atmosphere. Thus, during this season, when dry conditions prevail synoptically and a continental air mass is persistent, the columnar optical depths recover quite fast, within 1 week, even after getting depleted by ;70%. This is quite different from the behavior seen at the same location during the monsoon season when the rainfall is extensive (spatially widespread) and the air mass is marine. Under such conditions even 7–10 sunny days during the monsoon period break does not lead to any significant recharging of the atmosphere, and the AODs continue to be low almost the entire season (Subbaraya and Jayaraman 1982; Jayaraman et al. 1987; Moorthy et al. 1991, 1996). This is also due to the fact that the depletion during this season is over a very large spatial extent, unlike the mesoscale thundershowers (which produce only local effects). When such depletion occurs synoptically, long-range transport also becomes ineffective to supplement the loss due to scavenging. 5. Conclusions 1) A case study is presented of the impact of two mesoscale thundershowers on columnar aerosol optical depth spectra and the retrieved size parameters.
2) The two thundershowers had distinct impacts, even though both led to a substantial scavenging of aerosol particles from the atmosphere. During the first shower, which was only of moderate intensity and was rather localized, ;64% of the columnar mass was scavenged, primarily through the removal of supermicron coarse aerosols. This resulted in not only a significant reduction in the AOD over a wide wavelength range, but also in a change in the aerosol size distribution with an increase in the dominance of submicron aerosols and, consequently, the AOD spectra became steeper. 3) The second thundershower, which was much more intense and widespread, with heavy precipitation upwind, removed ;68% of the accumulation size aerosols by nucleation and impaction scavenging, but its effect was insignificant at the coarse size range. Thus, the resulting decrease in the mass was only ;17%, but the large reduction in the submicron aerosol concentration resulted in flattening of the AOD spectrum. The effective radius decreased continuously through both the events. The significant loss of both the coarse and accumulation aerosols and the strong rainfall upwind resulted in the AODs decreasing for one more day after the events and the resulting changes in the size distribution leading to an increase, though small, in the effective radius. 4) The apparent scavenging efficiency of supermicron aerosols was extremely high for the first rain (0.75) but insignificant for the second event, whereas for aerosols in the optically active submicron size range the efficiency was 0.57 for the first rain and 0.68 for the second, in line with the patterns reported from model estimates of the impact of scavenging on aerosol particles of different sizes and rate of rainfall. 5) The prevailing synoptic conditions, consisting of a dry, continental air mass and an absence of further precipitation, helped the atmosphere to recover from the impact within about 1 week. The recovery was also size dependent; the submicron aerosols were replenished faster, probably through advection by the synoptic winds, than the coarse aerosols, which depend more on the land conditions and local effects. Acknowledgments. This work was carried out under the Aerosol Climatology and Effects project of ISROGeosphere Biosphere Program. We are thankful to the meteorological center at Trivandrum for providing the complementary surface and upper-air data during the events. We thank the reviewers for several useful suggestions. REFERENCES Anderson, J., C. Brogniez, L. Cazier, V. K. Saxena, J. Lenoble, and M. P. McCormick, 2000: Characterization of aerosols from simulated SAGE III measurements applying two retrieval techniques. J. Geophys. Res., 105 (D2), 2013–2027.
914
JOURNAL OF APPLIED METEOROLOGY
Andronache, C., 2003: Estimated variability of below-cloud aerosol removal by rainfall for observed aerosol size distributions. Atmos. Chem. Phys., 3, 131–143. ——, L. J. Donner, C. J. Seman, and R. S. Hemler, 2002: A study of the impact of the Intertropical Convergence Zone on aerosols during INDOEX. J. Geophys. Res., 107, 8027, doi:10.1029/ 2001JD900248. Angstrom, A., 1961: Techniques of determining the turbidity of the atmosphere. Tellus, 13, 214–223. Asnani, G. C., 1993: Tropical Meteorology. Vol. 1. Indian Institute of Tropical Meteorology, 602 pp. Babu, S. S., and K. K. Moorthy, 2002: Aerosol black carbon over a tropical coastal station in India. Geophys. Res. Lett., 29, 2098, doi:10.1029/2002GL015662. Beard, K. V., 1974: Experimental and numerical collision efficiencies for submicron particles scavenged by small raindrops. J. Atmos. Sci., 31, 1595–1603. Devara, P. C. S., P. V. Subba Raju, R. S. Maheskumar, G. Pandithurai, P. E. Raj, and K. K. Dani, 1997: Atmospheric aerosol scavenging: A study with concurrent mie lidar and meteorological observations. J. Aerosol Sci., 28, S583–S584. Dovgalyuk, Yu. A., I. A. Molotkova, E. V. Orenburgskaya, Yu. F. Ponomarev, A. A. Sinkevich, D. D. Stalevich, E. N. Stankova, and V. D. Stepanenko, 1995: Field investigations of aerosol scavenging by precipitation. J. Aerosol Sci., 26, S369–S370. Ebert, P., M. Kibler, A. Mainka, B. Tenberken, K. Baechmann, G. Frank, and J. Tschiersch, 1988: A field study of particle scavenging by raindrops of different sizes using monodisperse trace aerosol. J. Aerosol Sci., 29, 173–186. ——, K. Ba¨chmann, G. Frank, M. Kibler, A. Mainka, B. Tenberken, and J. Tschiersch, 1997: Field experimental investigation of below-cloud scavenging processes. J. Aerosol Sci., 28, S575–S576. Fisher, R. A., 1970: Statistical Methods for Research Workers. Oliver and Boyd, 362 pp. Flossmann, A. I., W. D. Hall, and H. R. Pruppacher, 1985: A theoretical study of the wet removal of atmospheric pollutants. Part I: The redistribution of aerosol particles captured through nucleation and impaction scavenging by growing cloud drops. J. Atmos. Sci., 42, 583–606. ——, H. R. Pruppacher, and J. H. Topolian, 1987: A theoretical study of the wet removal of atmospheric pollutants. Part II: The uptake and redistribution of (NH4)2SO4 particles and SO2 gas simultaneously scavenged by growing cloud drops. J. Atmos. Sci., 44, 2912–2923. Gonc¸alves, F. L. T., A. M. Ramos, S. Freitas, M. A. Silva Dias, and O. Massambani, 2002: In-cloud and below-cloud numerical simulation of scavenging processes at Serra Do Mar region, SE Brazil. Atmos. Environ., 36, 5245–5255. ——, M. F. Andrade, M. C. Forti, R. Astolfo, M. A. Ramos, O. Massambani, and A. J. Melfi, 2003: Preliminary estimation of the rainfall chemical composition evaluated through the scavenging modeling for north-eastern Amazonian region (Amapa State, Brazil). Environ. Pollut., 121, 63–73. Hobbs, P. V., 1993: Aerosol cloud interactions. Aerosol Cloud Climate Interactions, P. V. Hobbs, Ed., Academic Press, 33–95. Jaenicke, R., 1984: Physical aspects of atmospheric aerosol. Aerosols and Their Climatic Effects, H. E. Gerbard and A. Deepak, Eds., A. Deepak Publishing, 7–34. Jayaraman, A., B. H. Subbaraya, and Y. B. Acharya, 1987: The vertical distribution of aerosol concentration and their size distribution function over the tropics and their role in radiation transfer. Phys. Scr., 36, 358–361. ——, S. K. Satheesh, A. P. Mitra, and V. Ramanathan, 2001: Latitudinal gradients in aerosol properties across the inter-tropical convergence zone: Results from the joint INDO-US study onboard Sagar Kanya. Curr. Sci., 80, 128–137.
VOLUME 43
King, M. D., 1982: Sensitivity of constrained linear inversion to the selection of Lagrange multiplier. J. Atmos. Sci., 39, 1356–1369. ——, D. M. Byrne, B. M. Herman, and J. A. Reagan, 1978: Aerosol size distributions obtained by inversion of spectral optical depth measurements. J. Atmos. Sci., 35, 2153–2167. Kleinbaum, D. G., and L. L. Kupper, 1978: Applied Regression Analysis and Other Multivariable Methods. Duxburg, 49 pp. Lelieveld, J., and Coauthors, 2001: The Indian Ocean Experiment: Widespread air pollution from South and Southeast Asia. Science, 291, 1031–1036. Leong, K. H., K. V. Beard, and H. T. Ochs III, 1982: Laboratory measurements of particle capture by evaporating cloud drops. J. Atmos. Sci., 39, 1130–1140. Lubin, D., S. K. Satheesh, G. McFarquar, and A. J. Heymsfield, 2002: Longwave radiative forcing of Indian Ocean tropospheric aerosol. J. Geophys. Res., 107, 8004, doi:10.1029/2001JD001183. Moorthy, K. K., P. R. Nair, and B. V. Krishna Murthy, 1991: Size distribution of coastal aerosols: Effects of local sources and sinks. J. Appl. Meteor., 30, 844–852. ——, ——, ——, and S. K. Satheesh, 1996: Time evolution of the optical effects and aerosol characteristics of Mt. Pinatubo origin from ground-based observation. J. Atmos. Terr. Phys., 58, 1101– 1106. ——, S. K. Satheesh, and B. V. K. Murthy, 1997: Investigations of marine aerosols over the tropical Indian Ocean. J. Geophys. Res., 102, 18 827–18 842. ——, K. Niranjan, B. Narasimhamurthy, V. V. Agashe, and B. V. K. Murthy, 1999: Aerosol climatology over India, 1. ISRO GBP MWR network and database, SR 03 99, ISRO GBP Program Office Rep. ISRO GBP SR 03 99, 1–30. ——, A. Saha, B. S. N. Prasad, K. Niranjan, D. Jhurry, and P. S. Pillai, 2001: Aerosol optical depths over peninsular India and adjoining oceans during the INDOEX campaigns: Spatial, temporal and spectral characteristics. J. Geophys. Res., 106, 28 539– 28 554. Pillai, P. S., and K. K. Moorthy, 2001: Aerosol mass-size distributions at a tropical coastal environment: Response to mesoscale and synoptic processes. Atmos. Environ., 35, 4099–4112. Prakash, W. J. J., R. Ramachandran, K. N. Nair, K. S. Gupta, and P. K. Kunhikrishnan, 1992: On the structure of sea breeze front effects observed near the coast line of Thumba, India. Bound.Layer Meteor., 59, 111–124. Pranesha, T. S., and A. K. Kamra, 1996: Scavenging of aerosol particles by large water drops, 1, Neutral case. J. Geophys. Res., 101, 23 373–23 380. ——, and ——, 1997: Scavenging of aerosol particles by large water drops, 3, Washout coefficients, half-lives, and rainfall depths. J. Geophys. Res., 102, 23 947–23 953. Pruppacher, H. R., and J. D. Klett, 1978: Microphysics of Clouds and Precipitation. D. Reidel, 714 pp. Radke, L. M., M. Ellgrowth, and P. Hobbs, 1980: Scavenging of aerosol particles by precipitation. J. Appl. Meteor., 19, 715–722. Satheesh, S. K., and K. K. Moorthy, 1997: Aerosol characteristics over coastal regions of the Arabian Sea. Tellus, 49B, 417–428. ——, V. Ramanathan, X. L. Jones, J. M. Lobert, I. A. Podgorny, J. M. Prospero, B. N. Holben, and N. G. Loeb, 1999: A model for the natural and anthropogenic aerosols over the tropical Indian Ocean derived from Indian Ocean Experiment data. J. Geophys. Res., 104, 27 421–27 440. Sinkevich, A. A., D. Yu. Afanasjev, and T. A. Pershina, 1995: Laboratory investigations of aerosol scavenging. J. Aerosol Sci., 26, S433. Subbaraya, B. H., and A. Jayaraman, 1982: Aerosol concentration and size distribution in the troposphere and lower stratosphere over Thumba. Pure Appl. Geophys., 120, 407–421. Wang, P. K., S. N. Grover, and H. R. Pruppacher, 1978: On the effect of electric charges on the scavenging of aerosol particles by cloud and small raindrops. J. Atmos. Sci., 35, 1735–1743.