1. No category

Aerosol-Induced Large-Scale Variability in
Precipitation over the Tropical Atlantic
Jingfeng Huang*, Chidong Zhang, and Joseph M. Prospero
Rosenstiel School of Marine and Atmospheric Science, University of
Miami, Miami, Florida
____________________
*Corresponding author address: Dr. Jingfeng Huang, Rosenstiel School of Marine and Atmospheric
Science, University of Miami, 4600 Rickenbacker Causeway, Miami, FL 33149.
E-mail: [email protected]
1
ABSTRACT
We used multiyear satellite observations to document a relationship between the
large-scale variability in precipitation over the tropical Atlantic and aerosol traced to
African sources. During boreal winter and spring, there is a significant reduction in
precipitation south of the Atlantic marine intertropical convergence zone during months
when aerosol concentrations are anomalously high over a large domain of the tropical
Atlantic Ocean. This reduction cannot be attributed to known climate factors such as El
Niño-Southern Oscillation, North Atlantic Oscillation, and zonal and meridional modes
of tropical Atlantic sea surface temperature, or to meteorological factors such as water
vapor. The fractional variance in precipitation related to aerosol is about 12% of the total
interannual variance, which is of the same order of magnitude as that related to each of
the known climate and weather factors. A backward trajectory analysis confirms the
African origin of aerosols that directly affect the changes in precipitation. The reduction
in mean precipitation mainly comes from decreases in moderate rain rates (10 – 20
mm/day), while light rain (<10 mm/day) can actually be enhanced by aerosol. Our results
suggest aerosols have a clearly identifiable effect on climate variability in precipitation in
the Pan-Atlantic region.
2
1. Introduction
The possibility that aerosol may influence precipitation and thereby modulate the
global hydrological cycle has been the focus of much research (Rosenfeld, 2000;
Haywood and Boucher 2000; Ramanathan et al. 2001; Rosenfeld et al. 2001; Lohmann
and Feichter 2005). Aerosol influences on precipitation may derive from radiative effects
(Carlson and Benjamin 1980; Miller and Tegen 1998; Diaz et al. 2001; Ramanathan et al.
2001) and/or its role as cloud condensation nuclei (CCN) or ice nuclei (IN) (Twomey et
al. 1984; Albrecht 1989; Wurzler et al. 2000; Sassen et al. 2003; Lohmann and Feichter
2005). While the radiative effects of aerosol (absorption and scattering) modify the
atmospheric and surface conditions for precipitation, precipitation efficiency and amount
sensitively depend on aerosols serving as CCN and IN. Intuitively, aerosol modulation of
large-scale variability in precipitation should be identifiable. In reality, however,
convincing evidence of such modulation has not been easy to obtain because of the
complexity of the processes. Precipitation variability depends on many factors, ranging
from the variability of the large-scale circulation, cloud dynamics, and microphysics and
the complex interrelationships of these factors which are poorly understood. Many
observational and modeling studies of aerosol effects on precipitation have been
conducted but they yielded inconsistent results (Menon, 2004; Takemura et al. 2005;
IPCC, 2007; Menon and Genio, 2007; Tao et al. 2007). The inconsistency mostly likely
comes from the fact that aerosol effects on precipitation sensitively depend on cloud
types and environmental conditions such as water vapor supply and atmospheric
instability (Lohmann, and Lesins 2002; Matsui et al. 2004; Menon 2004; Segal et al.
2004; Koren et al. 2005; Tao et al. 2007), as well as on the characteristics (e.g., size
3
distribution, chemical and physical properties, concentration) of aerosols such as smoke
(Kaufman et al. 2005a; Rosenfeld 1999), urban and industrial air pollution (Rosenfeld
2000; Menon et al. 2002; Kaufman et al. 2005a; Huang et al. 2007), and mineral dust
(Kaufman et al. 2005a; Rosenfeld et al. 2001).
Investigations on aerosol-precipitation interaction on large or climatic scales face
difficult challenges (IPCC, 2007). Observational and modeling studies suffer from similar
problems. There are uncertainties, sometimes large, in satellite observations of aerosol
and precipitation because of limitations in their retrievals. Numerical models suffer from
limitations in their parameterizations of aerosol effects and precipitation processes. Insitu observations from field campaigns and high-resolution model simulations are often
limited to certain specific environments and specific type of precipitating clouds;
consequently it is impossible to generalize their results to climatic scales.
Given the complications of precipitation mechanisms and aerosol effects on
precipitation, it is intriguing whether coherent large-scale variability in precipitation can
be induced by aerosol, and if it can, whether it can be detected by observations. We
address this question by proposing a null hypothesis H0: There is no coherent large-scale
variability between aerosol and precipitation in the tropics.
In this study, we attempted to test null hypothesis H0 using satellite observations
of aerosol and precipitation, augmented by other climate and weather data. We have two
motivations. First, disapproving null hypothesis H0 should be a necessary step to justify
other studies, observational and modeling alike, on climatic effects of aerosol on
precipitation. Second, observations of coherence large-scale variability between aerosol
4
and precipitation are needed to validate numerical models in the study of climatic effects
of aerosol on precipitation.
We chose the tropical Atlantic as our focus. In this region, the concentration and
variability of aerosol from Africa are among the largest, if not the largest, in the world
(Anderson et al. 1996; Prospero 1996, 1999; Prospero and Lamb, 2003; Torres et al. 2002;
Kaufman et al. 2002). Figures 1, 2 and 3 show climatological means, seasonal cycles, and
interannual variability of aerosol and precipitation in the tropical Atlantic, both derived
from long-term (1979-2000) satellite observations (see section 2 for descriptions). The
seasonal cycles in aerosol and precipitation are salient in both latitudinal positions and
amplitudes (Fig. 2a). They also undergo obvious interannual and decadal variability (Fig.
3, Cakmur et al. 2001; Prospero and Lamb 2003; Gu and Adler, 2006). A prominent
feature of the climatology is that the intertropical convergence zone (ITCZ), represented
by the narrow band of concentrated precipitation in Figs. 1 and 2, is flanked most of the
time by high concentration of aerosol on both the North and the South sides. There is
little doubt that precipitation in the ITCZ helps remove aerosol from the atmosphere
through wet deposition. The question is whether the close contact between the two would
result in any aerosol-induced large-scale variability in precipitation and, if it does,
whether such variability can be creditably detected using satellite observations.
It is well known that mineral dust from West Africa is often transported westward
into the tropical Atlantic region north of the ITCZ within a layer of the lower troposphere
known as the Saharan Air Layer (SAL, Carlson 1979; Carlson and Prospero 1972;
Karyampudi et al. 1999; Kaufman et al. 2005b). African dust can reach the Caribbean
(Prospero and Lamb 2003), southeastern United States (Prospero 1999; DeMott et al.
5
2003; Sassen et al. 2003), especially in boreal summer. In boreal winter, mineral dust
from Africa can be transported southward across the Guinean coast and westward into the
tropical Atlantic south of the ITCZ, even into South America (Prospero et al., 1981;
Chiapello and Moulin 2002; Prospero and Lamb, 2003; Chiapello et al. 2005). The
interannual variability of aerosol in the tropical Atlantic is controlled by the interannual
variability of Sahel precipitation (Prospero and Lamb, 2003) and transport mechanisms
(Swap et al. 1996; Moulin et al. 1997; Chiapello et al. 2005; Washington and Todd
2005).
Climate, in the larger sense, also plays a role. The El Niño-Southern Oscillation
(ENSO) effect on aerosol was also demonstrated as the enhanced Barbados dust
concentration in extreme ENSO warm events which are known to be linked to drought in
West Africa (Prospero and Lamb 2003). In boreal winter dust transport is modulated by
the North Atlantic Oscillation (NAO) (Ginoux et al. 2004).
At the microphysical level, mineral dust may serve both as giant CCN to initiate
warm rain (Johnson 1982; Rosenfeld and Farbstein 1992) and as IN to initiate cold rain
(DeMott et al. 2003). However, precipitation suppression due to desert dust is also
possible (Rosenfeld et al. 2001). Model simulations showed that the dust effect on
precipitation may sensitively depends on dust layer altitude and ambient temperature
(Yin and Chen 2007), sea surface temperature (Yoshioka et al. 2007), and ocean
dynamical heat transport (Miller and Tegen 1998). Meanwhile, carbonaceous (or black
carbon) aerosol from biomass burning over Africa are also transported westward into the
tropical Atlantic, especially in boreal winter and south of the ITCZ (Hao and Liu, 1994;
Kaufman et al. 1997; Torres et al 2002; Kaufman et al. 2005b). It has been reported that
6
black carbon aerosol tend to suppress warm rain in the tropics (Kaufman and Fraser,
1997; Rosenfeld, 1999; Wang 2007).
Given the intensity and scale of aerosol transport that takes place over the tropical
Atlantic for so much of the year, we might expect that if aerosols have an impact on
precipitation, the impact should b most readily detectable over this region. There are both
pros and cons in choosing the tropical Atlantic as a test ground for detecting climatic
effects of aerosol on precipitation. Dynamic forcing to precipitation in the tropical
Atlantic includes local sea surface temperature (SST, Zebiak, 1993) and synoptic-scale
waves (e.g. Thompson et al. 1979; Gu and Zhang 2002), and remote influences by ENSO
(e.g. Alexander and Scott 2002, Xie and Carton 2004), NAO (e.g. Wu and Liu 2002;
Kyte et al. 2006), and perhaps West African monsoon (Hagos and Cook, 2005). But
arguably, these forcing mechanisms are relatively weak in comparison to those in other
tropical regions, such as the eastern Pacific where ENSO signals dominate, and the
western Pacific and Indian Oceans where the Madden-Julian Oscillation reins (Madden
and Julian 1971). The large aerosol concentration and variability combined with a
relatively weak dynamic forcing to precipitation make the tropical Atlantic an ideal
region to detect climatic effects of aerosol on precipitation. If we cannot find any
convincing evidence for large-scale effects in this region, then there is little hope to find
it anywhere else. On the other hand, the mixture of mineral dust and black carbon aerosol
in the tropical Atlantic region complicates the interpretation of any observed large-scale
variability in precipitation associated with aerosol. Satellite aerosol data especially those
with long-term records that we use in this study (see section 2.a), cannot readily
distinguish between the two. We are therefore forced to settle with a low expectation: To
7
find out whether there is coherent large-scale variability in precipitation associated with
anomalous aerosol regardless of their types.
Our analysis strategy is the following. Within the limitation of linear data analysis,
we remove influences of known climate and weather factors, such as ENSO, NAO,
tropical Atlantic SST, and water vapor from long-term satellite data of aerosol and
precipitation. If there is large-scale co-variability between the resulting anomalies in
aerosol and precipitation data that passes statistical significance tests and is confirmed by
more recent high-resolution and high-accuracy satellite data, we interpret it as a signal of
aerosol effects on precipitation. The complication of wet deposition can be eased by
carefully selecting analysis domains (see section 2.b). The physical significance of such a
signal is assessed by comparing the magnitudes of precipitation variability due to aerosol
and the other climate and weather factors. Finally, we use backward trajectory analysis to
identify the source regions of aerosol that may directly responsible for the observed
variability in precipitation. Null hypothesis H0 will be rejected only if all these steps are
successfully accomplished.
The data used in this study and our analysis method are described in section 2.
Results are presented in section 3. A summary and discussions are provided in section 4.
(FIG. 1, FIG. 2, and FIG. 3 approximately go here)
2. Data
a. Main Datasets
Our analyses cover three periods (Fig. 3). The longest is from 1979 through 2000
with a data gap over 1993-1996. In this period, precipitation data from the Global
8
Precipitation Climatology Project (GPCP) and aerosol data from the Total Ozone
Mapping Spectrometer (TOMS) are available. Our main statistics about aerosol and
precipitation are from this period. The second period is from 1988 through 2000, during
which the Special Sensor Microwave Imager (SSM/I) precipitable water data are
available. The objective of the analysis for this period is to isolate water vapor effects on
precipitation from aerosol effects. The third period covers the most recent years, 2000 –
2006, when the Tropical Rainfall Measuring Mission (TRMM) precipitation and the
Moderate Resolution Imaging Spectroradiometer (MODIS) aerosol data are available.
These two sensors yielded data sets with higher resolution and higher quality; they were
used to confirm the results based on the GPCP and TOMS data.
The GPCP version 2 monthly and pentad (five-day) precipitation data (mm/day,
2.5˚×2.5˚) incorporate precipitation estimates from low-orbit satellite microwave data,
geosynchronous-orbit satellite infrared data, and surface rain gauge observations but
there were no microwave data before mid-1987 (Huffman et al., 1997; Adler et al., 2003;
Yin et al. 2004). The GPCP pentad precipitation data is used to examine variability in
precipitation at different rain rates.
The TOMS aerosol index (AI) version 8 (1˚ ×1.25˚) monthly data are available
from two platforms: Nimbus-7 (November 1978 – April 1993), and Earth Probe (August
1996 – December 2000). Earth Probe TOMS AI data after December 2000 were not used
because of their known calibration drift and biases1. Also there is a data gap from 1993
to 1996. TOMS AI can be taken as a qualitative indicator of UV absorbing aerosols of
which dust and carbonaceous aerosol are the dominant types (Herman et al. 1997; Torres
1
http://toms.gsfc.nasa.gov/news/news.html.
9
et al. 1998; Torres et al. 2002). We used 0.5 as threshold values for TOMS AI because AI
approaches zero may contain ground noise or cloud contamination (Herman et al. 1997).
TOMS AI is more sensitive to aerosols at high altitudes than those in low lying layer but
Barbados surface dust concentration and TOMS AI are closely corrected (Chiapello et al.
1999).
Monthly Special Sensor Microwave/Imager (SSM/I) atmospheric water vapor
data (mm, 0.25° × 0.25°) (Jackson and Stephens 1995; Wentz 1997) were used to
examine water vapor effects on precipitation in the second analysis period (1988-2000).
The monthly TRMM precipitation 3B43 version 6 product (mm/day, 0.25° × 0.25°)
(Fisher 2004) and the monthly Terra MODIS aerosol optical depth (AOD) (1° × 1°,
Kaufman et al. 1997), both available for the third analysis period (2000-2006), were used
to validate GPCP-TOMS results.
Major climate factors influencing Atlantic precipitation include ENSO (Philander
1990), NAO (Hurrell 1995), and the zonal and meridional modes of Atlantic SST (Zebiak
1993). Indices used for these climate factors are the Niño 3.0 SST index (5°S–5°N,
90°W–150°W) index for ENSO, the NAO index for the North Atlantic Oscillation, and
the Atl.3.0 SST index (3°S–3°N, 20°W–0°) and TNA-TSA SST index (SST anomaly
difference between the TNA region (5°–25°N, 55°–15°W) and the TSA region (0°–20°S,
30°W–10°E), Servain 1991; Wang 2002) for the zonal and meridional modes of tropical
Atlantic SST respectively.
b. Analysis Domains
10
We focus mainly on precipitation variability in the ITCZ which has a strong
influence on the climate of adjacent continents especially northeast Brazil and West
Africa (Hastenrath and Heller 1977; Lamb 1978; Janicot et al. 1998; Robertson et al.
2004). The analysis domains for precipitation and aerosol are selected to reflect both the
spatial and the temporal (monthly and interannual) variability of precipitation and
aerosol. Land effects are minimized by excluding continental regions from the analysis
domain. A domain for precipitation extends over the range that encompasses the seasonal
shifting of the AMI, from its southernmost position in boreal spring to its northernmost
position in boreal summer (Fig. 2). This domain will hereafter be referred to as the AMI
domain (Fig. 1). The analysis domain for aerosol, denoted as Atlantic aerosol (AA)
domain in Fig. 1, covers a much larger region than the AMI domain. We assume that the
variability of aerosol directly affecting AMI precipitation is part of the basin-scale
coherent variability that can be measured by aerosol data averaged over the larger AA
domain. The assumption is supported by daily satellite observations of aerosol and largescale synoptic patterns responsible for transporting aerosol from Africa westward to the
tropical Atlantic region (Chiapello and Moulin 2002; Prospero and Lamb 2003; Chiapello
et al. 2005). This two-domain approach is to ensure that the detected variability in aerosol
averaged over the AA domain is not contaminated by the complications of wet deposition
and uncertainties in satellite aerosol retrieval in cloudy areas but meanwhile is
representative of aerosol fluctuations in the neighborhood of the ITCZ that may directly
affect its precipitation.
c. Data processing
11
Aerosol and precipitation anomalies are calculated by removing their seasonal
cycles. Effects on aerosol and precipitation by known climate factors, such as the ENSO,
NAO, and the zonal and meridional modes of Atlantic SST, were removed through multivariable linear regression. This is to minimize the ambiguity that co-variability between
aerosol and precipitation that might be caused simultaneously by these climate factors
instead of serving as a signal of aerosol effect on precipitation. The resulting aerosol and
precipitation anomalies could still exhibit residual seasonal cycles in their variance. To
treat all months equally so that the total sample size is increased, the anomalies are
normalized by their interannual standard deviation in each month averaged over the
respective domains. We will specifically mention “normalized anomalies” when these
latter time series are used.
The data are not sufficiently long to explicitly address issues related to any longterm trend. To avoid uncertainties in this and keep the focus of the current study within
its scope, the regional means of normalized anomalies were detrended. Using the data not
detrended yields similar results.
3. Large-Scale Co-Variability
We begin by examining the domain mean time series of precipitation (in the AMI
domain) and aerosol (in the AA domain). Both precipitation and aerosol undergo
prominent seasonal cycles (Fig. 2), interannual and decadal variability (Fig. 3). In
general, the domain averaged Nimbus-7 and Earth Probe TOMS AI are comparable in
their seasonal cycles. The TOMS AI values are relatively higher in boreal spring and
summer and lower in autumn and winter. There is an increasing trend in the early years
12
of the Nimbus-7 TOMS AI; the increase is consistent with long-term dust measurements
made on Barbados (Prospero and Lamb, 2003) and is linked to the onset of severe
drought in the Soudano-Sahel region of North Africa. Beyond that period, the TOMS AI
varies around a mean of about 0.75 from year to year. The interannual variability of
precipitation is steady with an averaged annual mean of ~3 mm/day. Its seasonal cycle
varies from 2 mm/day in boreal winter to as high as 5 mm/day in boreal summer.
a. Effects from known climate factors
We apply a multi-variable regression to the anomalous time series, and we
compare the aerosol-related precipitation variance (averaged over the AMI domain) to
that related to each of the known climate factors, namely, ENSO, NAO, and Atlantic SST
zonal and meridional modes (Fig. 4a). The highest aerosol-related precipitation variance
is about 30% of the total in January and about 3-15% in other months. The largest
fractional variance in precipitation anomalies related to ENSO occurs in January (~43%).
The Atlantic zonal SST mode dominates the precipitation variability in boreal summer
(July and August) and the meridional mode dominates in boreal spring (March and
April), each of them accounts for 25 – 40% of the variability. The largest fractional
variance linked to the NAO is in February (~15%); the NAO is not a dominant factor in
the ITCZ precipitation variability in any month. Although aerosol is not a dominant
factor in precipitation variability in any one month, its influence on precipitation is
comparable to those by ENSO, NAO, and tropical Atlantic SST, especially in boreal
winter and spring. On average, the fractional interannual variance in precipitation related
13
to aerosol is about 12%, the same order of magnitude as those related to the other known
climate factors.
These climate factors also affect the aerosol variability. Similar multi-variable
regression was conducted on aerosol variance and results are shown in Fig. 4b. ENSO
explains as high as 38% of aerosol variance in January. Effects from the tropical Atlantic
SST on aerosol are at a maximum in April for the zonal mode (~35%) and in June for the
meridional mode (~40%)., The NAO, with an average less than 10% of the aerosol
interannual variance, does not seem to be a dominant factor for the aerosol variability in
any month.
These effects from the known climate factors on both precipitation and aerosol
may complicate the interpretation of empirical aerosol-precipitation relationships. To
avoid such possible complication, their effects were removed from the anomalous time
series of both precipitation and aerosol. In other words, the analyses described hereafter
were made using data that are linearly independent of ENSO, NAO, and tropical Atlantic
SST.
(FIG. 4 approximately goes here)
b. Correlation
Fig. 5 compares time series of monthly normalized anomalies in GPCP
precipitation averaged over the AMI domain and TOMS AI averaged over the AA
domain. A visual inspection gives an impression that anomalously low (high)
precipitation tends to occur in months of anomalously high (low) aerosol concentration.
A scatter diagram for the two variables (Fig. 6) confirms this casual observation. The
14
aerosol and precipitation normalized anomalies are negatively correlated with a
correlation coefficient (R) of -0.324 which is statistically significant at the 99%
confidence level2. The negative correlation implies that precipitation averaged in the AMI
domain is reduced when aerosol concentration averaged in the AA domain is
anomalously high.
To test the sensitivity of the correlation to the domain selection, we list in Table 1
six cases in which correlation coefficients were calculated for the two variables averaged
in different domains. Because there is not much precipitation outside the AMI domain,
the selection of precipitation domain between AA and AMI (Cases A and B) does not
make much difference. The only insignificant correlation for precipitation and aerosol is
found for both in the NAA domain (Case E). This is the first indication that precipitation
at the northern edge of the ITCZ is less susceptible to aerosol effects, or its susceptibility
is less observable than at the southern edge. But Cases C and D suggest that aerosol north
and south of the ITCZ may both affect its precipitation.
The correlation between the two variables, aerosol and precipitation, varies
through the year (Fig. 6). It is largest in January (R=-0.626) and February (R=-0.649),
both being significant at the 99% confidence level. The correlations are significant in
boreal winter (Dec-Jan-Feb, R=-0.51, 99% confidence level), summer (Jun-Jul-Aug, R=0.31, 95% confidence level) and spring (Mar-Apr-May, R=-0.27, 95% confidence level).
This seasonality in the aerosol-related precipitation change is the second signal that
precipitation is more susceptible to aerosol effect or its susceptibility is more observable
south of the ITCZ than north of it. The most significant correlation between the two
2
Two-tailed (non-directional) T-test is used to determine confidence levels of the correlation significance.
15
occurs in months of the highest aerosol concentration south of the ITCZ (e.g. Jan and
Feb), while north of the ITCZ the months with the highest aerosol concentrations are
June and July.
To demonstrate the spatial distribution of the precipitation variability associated
with aerosol, we calculated correlation between the normalized anomalous aerosols
averaged over the AA domain and the normalized anomalous precipitation at each grid
point. Fig. 7 shows the correlation coefficients. Areas with correlation statistically
significant at the 95% and 99% confidence levels are outlined. The negative correlation is
highest (at the 99% confidence level) south of the ITCZ in the western tropical Atlantic.
The negative correlation suggests that precipitation is reduced along the southern edge of
the ITCZ because of high aerosol concentrations. Evident in Fig. 2, December – May is
the season when absorbing aerosol south of the equator is relatively high and the ITCZ
approaches its most southern position. The high value of TOMS AI are possibly the
combined effect of carbonaceous particulate produced by the intense fire season in
central Africa and the southward flow of mineral dust from Lake Chad, the SoudanoSahel region, and other significant sources of mineral dust in West Africa (Torres, 2002;
Prospero et al. 2002). The implication of the seasonality and locality in the correlation
will be discussed in section 4.
(FIG. 5, FIG. 6, FIG. 7, and TABLE 1 approximately go here)
c. Composites
To further explore the co-variability between aerosol and precipitation, we
compare precipitation and aerosol between months of extreme high and low aerosol
16
concentrations. The top and bottom one-thirds (terciles) of all months ranked by
normalized anomalies in aerosol averaged over the AA domain were selected to represent
such extreme months of aerosol concentrations. Composites of the normalized anomalies
in precipitation and aerosol for the two terciles were made at each grid point. The
differences between the composites of the two terciles highlight the variability between
months of high and low anomalous aerosol concentrations. They are shown in Fig. 8a-b
for precipitation and in Fig. 8c-d for aerosol. The spatial distribution of the difference in
precipitation shows a coherent pattern over the western tropical Atlantic (Fig. 8a),
consistent to the correlation pattern in Fig. 7. The largest difference south of the ITCZ is
about -1.5, corresponding to a reduction in monthly precipitation of 55 mm, roughly 40%
of the mean monthly precipitation in the region (Fig. 1). When the composite procedure
is repeated for each calendar month, a distinct seasonal cycle in the difference emerges
(Fig. 8b). The reduction in precipitation remains south of the ITCZ center and migrates in
latitude along with the ITCZ. Its largest amplitudes are in boreal winter and spring.
The same composites and their differences were made for aerosol (Figs. 8c and
d). The most striking feature is that largest differences in aerosol between the top and
bottom terciles are not associated with the regions where the largest differences in
precipitation are found. This is partially understandable. The largest aerosol variability
should be near the aerosol sources, over Africa in this case, where there might not be
much precipitation at all. What directly matters to ITCZ precipitation processes is not so
much the amount of aerosol present but the availability of aerosols to interact with
precipitating clouds in the ITCZ. Meanwhile, Fig. 8 also justifies our approach of using
the AA domain to represent the large-scale variability of aerosols that may affect ITCZ
17
precipitation. The variability in AA domain averaged aerosol is not controlled by wet
deposition in the ITCZ.
The significance of the difference shown in Fig. 8a was tested from two
standpoints. First, we calculated the chance at which the difference pattern could be
produced by randomly selecting the same number of months for the composites. This was
done by comparing the probability distribution function (PDF) of the composite
difference within the AMI domain (bars in Fig. 9) to a Gaussian distribution with the
same standard deviation but zero mean, which represent the PDF from random sampling
(solid curve in Fig. 9). A Kolmogorov-Smirnov (K-S) test indicates that the two
distributions are different at the 99% confidence level. In other words, the chance for the
difference pattern in Fig. 9a to be reproduced by random sampling is only 1%. Second,
we calculated the chance at which this difference pattern is the same as the interannual
variability of precipitation without considering any influences from aerosol. The PDF of
interannual variability in precipitation was calculated using the entire normalized
anomalies in the same domain (dotted curve in Fig. 9). Again, a K-S test indicates that
this PDF is different from that of composite difference at the 99% confidence level. We
thus conclude that the difference pattern in Fig. 8a indeed represents a significant change
in precipitation related to anomalous aerosol concentrations.
(FIG. 8 and FIG. 9 approximately go here)
d. Variability at different rain rates
The previous analyses focus on changes in mean precipitation. It is possible that a
change in the mean is not uniformly contributed through the entire range of the rain rate.
In other words, it is reasonable to expect that the observed reduction in mean
18
precipitation associated with aerosol might be strongest in a certain range of the rain rates
than in others. For example, when small, statistically insignificant changes in mean
precipitation are found, there could be little aerosol effect at all; or aerosol effects could
be opposite for different precipitating cloud systems with different rain rates so that in the
mean they cancel each other. To address this issue, we used the pentad GPCP
precipitation data to compare monthly precipitation as a function of the rain rate for the
top and bottom tercile months of aerosol concentrations. Their differences are shown in
Fig. 11 for several subdomains in the analysis region (Fig. 10f).
For subdomain A, the precipitation reduction related to aerosol seen in the mean
(Figs. 8a and 10f) mainly occurs in the rain rate range of 5 – 15 mm/day, with a
maximum reduction in monthly precipitation of 120 mm at the rain rate of 11 m/day (Fig.
10a). There is a sign of precipitation enhancement due to aerosol for very light rain (<3
mm/day) and there is no obvious change in heavy rain (> 25 mm/day). In the center of the
ITCZ (subdomain B in Fig. 10f), there are both obvious precipitation reduction and
enhancement in different ranges of the rain rate. The reduction occurs for moderate rain
of 9 – 20 mm/day and the enhancement for light rain less than 7 mm/day. Their combined
contribution gives a weak reduction in the mean (Fig. 8a and 10f). The same features of
aerosol related reduction in moderate rain and enhancement in light rain also occur along
the northern edge of the ITCZ (subdomains C and D), with the peak reduction shifting
slightly to the heavier rain rate (20 mm/day). Comparing the aerosol related changes to
subdomain A, the total amount of reduction is about the same north and south of the
ITCZ, but the enhancement in light rain is much greater to the north. This is especially
true in subdomain C, where the cancellation between the reduction in moderate rain and
19
enhancement in light rain yields a negligible change in the mean (Fig. 8a and 10f). To the
south of the ITCZ in subdomain E (Fig. 11e), we find a different feature: light rain (<10
mm/day) is reduced as opposed to enhanced north of the ITCZ in subdomain D. While
the complication in aerosol-related changes in precipitation at different rain rates will be
discussed more in section 4, an important message here is that effects of aerosol on
precipitation cannot always be measured solely in terms of changes in mean
precipitation.
(FIG. 10 approximately goes here)
e. Variance
The importance of the aerosol to precipitation needs to be understood in the
context of total variability in precipitation. This can be addressed from two perspectives.
First, the variance in precipitation explainable by aerosol should be compared to those by
other known climate factors, such as ENSO, tropical Atlantic SST and NAO. This has
been discussed in section 3b. In certain months of the year, aerosol-related variance in
monthly mean precipitation is at the same order of magnitude as those related to the other
climate factors (Fig. 4). Similarly, we may ask how much of the variance in precipitation
that is linearly independent of the other climate factors can be explained by aerosol. To
quantify this, we calculated the “residual variance” of precipitation, which is the variance
of precipitation after the influences by ENSO, tropical Atlantic SST and NAO were
linearly removed by the multi-variable regression as described in section 3b. The aerosolrelated fraction of this residual variance was then calculated at each grid point using
aerosol anomalies averaged over the AA domain. As expected, the large fractional
20
variance due to aerosol is mainly south of the ITCZ (Fig. 11) where the largest aerosolrelated reduction in precipitation is found. Aerosol-related variance explains up to 15% of
the residue variance. This aerosol-related fractional variance can be as high as 45% in
certain months (e.g., January). All these observations suggest that precipitation variability
related to aerosol is a non-negligible component of its total variability.
(FIG. 11 approximately goes here)
f. Analysis for the second period (1988-2000): Water vapor effect
An outstanding problem not yet addressed so far is the possibility that the
observed precipitation change associated with aerosol is caused not by aerosol but by
anomalous water vapor accompanying anomalous aerosol. It is known that during much
of the year dusty air masses emerge from North Africa and pass over the tropical Atlantic
north of the equator and into the Caribbean. During the summer, these outbreaks are
characterized by the presence of an elevated layer of hot, dry, dust-laden air; because of
these characteristics it is commonly referred to as the SAL (Carlson and Prospero 1972;
Karyampudi et al.1999). South of the equator, dry air outbreaks from Africa have also
been observed (Zhang and Pennington 2004). Some of them may have high aerosol
concentrations. To quantitatively isolate effects of aerosol from those of water vapor, we
made two similar sets of composites of precipitation between top and bottom aerosol
tercile months for the shorter second period (1988 – 2000) when SSM/I water vapor data
are available. In one set of composites, only the known climate factors are removed
through a multi-variable regression as done previously for the longer first period (19792000). In another, the SSM/I water vapor averaged over the AA domain is included in the
21
multi-variable regression and its linear association with precipitation is removed along
with the other climate factors. The composite difference for the second period, with or
without water vapor effect removed, is almost identical (Fig. 12), which is similar to its
counterpart in Fig. 8a for the first period. The domain averaged reduction in mean
precipitation attributable to water vapor is only about 4% of that due to aerosol. The
fractional variance of monthly precipitation that is explainable by water vapor variability
is about 1 % in comparison to 9% by aerosol. We can therefore confidently say that the
aerosol-related changes in precipitation observed here cannot be attributed to
accompanying water vapor variability.
(FIG. 12 approximately goes here)
g. Analysis for the third period (2000-2006): Validation by high-resolution data
The long-term satellite data we have used thus far in this study (GPCP and
TOMS) may suffer from inaccuracies for various reasons. To validate the trends we
perform the same analysis using more recent satellite products that provide mare accurate
data with higher resolutions, TRMM and MODIS. Because of the short data record length
of the latter sensors, we cannot develop reliable statistics using monthly mean data. On
the other hand, when data of higher temporal (e.g., daily) resolution are used, some of the
assumptions underlining our analysis method (e.g., that the AA domain mean aerosol
represents the variability that directly affects ITCZ precipitation) may not be valid.
Consequently the complication of wet deposit vs. aerosol-induced changes in
precipitation may not be clearly resolved solely from an observational point of view
alone. At this stage, what we can ask is whether the same analysis method applied to
22
monthly mean TRMM and MODIS data would yield results consistent with those
obtained from the GPCP and TOMS data even though the results are not identical.
Following this line of thinking, we repeated the composites for the top and bottom tercile
months of aerosol using MODIS aerosol optical depth (AOD) data and TRMM
precipitation data for the third period of 2000-2006. Despite the limited temporal
coverage of data and detailed discrepancies, we obtain a relatively consistent result for
the main regions where significant precipitation reduction is found (Fig. 8a and 13a).
When aerosol becomes anomalous high, precipitation is reduced in the same region as
found from the GPCP and TOMS data: south of the ITCZ in the western tropical Atlantic.
The strongest reduction there also occurs during boreal winter and spring (Fig. 13b). It is
noticed, however, that TRMM precipitation is obviously enhanced north of the ITCZ due
to increases in the MODIS aerosol concentration. This enhancement, very weak in the
GPCP-TOMS data, is worth of further scrutiny but is outside the scope of this study.
(FIG. 13 approximately goes here)
h. Case study: Backward trajectory analysis
A remaining question from the previous analysis is, what is the source of the
aerosol that may directly induce the precipitation reduction? To identify the aerosol
origin, we conducted backward trajectory analyses using the HYSPLIT4 model applied to
the NCAR reanalysis data (Draxler and Rolph). The starting points of the back tracking
are set at three levels (500, 1500, and 3000 m) at the grid of 0˚E and 35°W, where the
aerosol-related precipitation reduction is largest. This point will be referred to as the
target point of the backward trajectory analysis. Because the precipitation reduction is
23
most significant in boreal winter and spring (section 3c), two outstanding cases were
selected for the backward trajectory tracking: January-February of 1997 and April-May
of 1987. In the time series of normalized anomalies (Fig. 5), these two seasons are
observed with highest anomalous aerosol concentration and significant precipitation
reductions. The backward trajectories are shown in Fig.14. It is clear that aerosols
reaching the target area are from the east, with the most consistent transport occurring at
the 3000 m level; the 500 m trajectories arrive mostly from over-ocean regions. In
January and February of 1997, the most possible aerosol sources are in central Africa
where the mixture of dust and smoke is prevailing (Torres, 2002; Prospero et al. 2002;
Husar et al. 1997). In April and May 1987, there appear to be two separated paths of
aerosol from Africa, in stead one as in January and February. But such details need to be
verified using a more systematic approach covering a sufficiently long period for reliable
statistics. Nevertheless, these two cases make us confident that aerosols directly
attributable to the precipitation reduction south of the ITCZ are from African sources.
(FIG. 14 approximately goes here)
4. Summary and conclusions
In an attempt at exploring whether aerosol can induce large-scale fluctuations in
tropical precipitation on climate timescales, we have documented statistically significant
co-variability between African aerosol and precipitation in the tropical Atlantic using
long-term multiyear satellite data of aerosol (TOMS aerosol index) and precipitation
(GPCP). The strongest signature of this co-variability lies south of the Atlantic ITCZ,
especially over the western part of the tropical Atlantic, where precipitation is reduced in
24
months with anomalously high aerosol concentrations in a large area. This aerosol-related
reduction in precipitation is most prominent during boreal winter and spring. The aerosolinduced fluctuations in precipitation are linearly independent of known climate and
weather factors (e.g., ENSO, NAO, tropical Atlantic SST, and water vapor) and are of the
same order of magnitudes as those induced by these factors. Aerosol effects on
precipitation are not the same for different precipitating systems represented by their rain
rates. In our observations, African aerosol tends to reduce precipitation at moderate rain
rates (10 – 20 mm/day) but may enhance light rain (< 10 mm/day). There is no obvious
aerosol effect on heavy rain (> 20 mm/day).
Based on these results, we confidently reject the null hypothesis H0 (there is no
coherent large-scale variability between aerosol and precipitation in the tropics) proposed
in section 1 of this article. These results, to the best of our knowledge, demonstrate for
the first time the climatic effect of aerosol on precipitation in the tropical Atlantic from
observational point of view. On the other hand, our work shows that the aerosol effects
are not uniform. More investigations, from both observational and modeling perspectives,
are needed to further quantify the aerosol effect on tropical precipitation and to
understand its mechanisms. Specifically germane to the results from this study, the
following issues need to be addressed:
(1) As discussed in section 1, aerosol may affect precipitation through their
radiative effects or their roles as CCN or IN. Our observations, however statistically
significant they are, provide no information on the mechanisms of aerosol – precipitation
interactions.. It is notable, however, that global climate models that incorporate the
radiative effects of black carbon aerosol have produced patterns of reduction in
25
precipitation that are similar to those found in our analysis (Wang 2007; Chung and
Seinfeld 2005). The consistent results between modeling and observations lend
confidence to the modeling results and provide a large-scale context to observations.
(2) One intriguing result from this study is the opposite effects of aerosol on
precipitation at light and moderate rain rates. The observed reduction in precipitation at
moderate rain rates is consistent with many previous studies reporting suppression of
warm rain (see discussion in section 1). But the accompanying enhancement of light rain
by aerosol is unexpected. Sometimes, it is this enhancement of light rain that partially off
sets the reduction in moderate rain and leads to a trivial change in the total (or mean)
precipitation. This has two related implications. The effects of aerosol on precipitation
cannot be solely judged by changes in the mean. While mean precipitation is a main
concern for the hydrological cycle at the surface, how precipitation at various rain rates
may vary is directly relevant to dynamics. If changes in different rain rates are associated
with different precipitating cloud systems, the aerosol-induced fluctuations in
precipitation would also mean fluctuations in the vertical latent heating structure. It is
well known that responses of the large-scale circulation to tropical diabatic heating
sensitively depend on its vertical profiles (Hartmann et al. 1984; Bergman and Hendon
2000; Schumacher et al. 2004). TRMM retrievals of tropical latent heating profiles (Tao
et al. 2007) may provide insights to this problem.
(3) It is clear from this study that the largest fluctuations in precipitation
associated with aerosol are not in regions where the largest aerosol concentrations are
found. One reason for this is simply that north of the ITCZ, where aerosol concentrations
can be very high, there is not much precipitation. The other reason might be related to
26
background aerosol. Wang (2005) found in a cloud-resolving model that higher CCN
concentrations does not necessarily lead to increased precipitation efficiency; the aerosol
effect on precipitation is more easily seen in a pristine environment than in a polluted
one. Other large-scale background environment may also play a role in determining the
pattern of precipitation responding to aerosol effects. For example, one puzzling result
from this study is that the largest reduction in precipitation by aerosol is located in the
western tropical Atlantic, far from the sources of African aerosol. Even though the
significant correlation pattern relates the upstream sources and this response area (Fig. 7),
the great distance between the largest precipitation response and the aerosol source region
needs to be explained. One possible reason for this is that the dynamic forcing of
precipitation (e.g., SST meridional gradient) in the eastern part of the AMI domain is
much stronger than in the west part, which may make precipitation in the eastern part less
susceptible to external perturbations such as those from aerosol. Fig. 15 superposes the
precipitation difference composite (as seen in Fig.8a) onto the climatological mean of
tropical Atlantic SST. The largest SST merdional gradient is found where the aerosol
effect on mean precipitation change is relatively weak.
(FIG. 15 approximately goes here)
(4) The asymmetric aerosol effects on precipitation north and south of the ITCZ
found in this study may reflect the differential response of precipitation to mineral dust
and biomass burning smoke. Even though the aerosol data we have used cannot
distinguish mineral dust from biomass burning smoke, there are reasons to believe that
biomass burning aerosol from Africa might have played a major role in the observed
reduction of precipitation. Our backward trajectory analysis suggests sources of aerosol
27
in areas of the largest aerosol-related reduction in precipitation are over equatorial Africa,
where biomass burning is most intense in boreal winter and spring (Hoelzemann et al.
2004), the same season of our observed largest reduction in precipitation. Other studies
have shown that aerosol released from biomass burning over equatorial Africa can spread
over the entire tropical Atlantic (Duncan et al. 2003; Hoelzemann et al. 2004; Ito and
Penner 2004; Ito and Penner 2005). Previous studies seem agree that the biomass burning
smoke reduces coalescence efficiency and consequently suppresses warm rain (Rosenfeld
1999; Kaufman et al 2005a). The aerosol effect on warm rain, whose typical rain rate is
moderate (<20 mm/day), might be the main reason for the observed reduction in
precipitation in this study.
A role of mineral dust in the observed reduction of precipitation cannot be ruled
out. During boreal winter and spring large quantities of mineral dust are carried from
sources in the Sahel-Soudano region of Africa into the tropical Atlantic (Mahowald et al.
2004; Washington and Todd 2005). Precipitation reduction by desert dust in other parts
of the world has been reported (Kaufman et al. 2005a; Rosenfeld et al. 2001). Human
activities are known to be influential to biomass burning (Dwyer et al. 2000; Kaufman et
al. 2005) and mineral dust (Moulin and Chiapello, 2006). Our results point to a possible
role of human activity in modulating precipitation even in remote areas,
The results from this study cannot be generalized for other tropical regions
because of different precipitation environments and mechanisms, and different aerosol
characteristics. With a given temperature distribution, aerosol should not produce a net
increase or decrease in global precipitation without altering total surface evaporation.
Regionally, aerosol may help redistribute precipitation and change rain characteristics
28
(types, rate, duration, etc.). Therefore, an effect of aerosol on the global hydrological
cycle, if any, must be manifested in terms its regional signatures. Studies of aerosol
effects on the global hydrological cycle must therefore be conducted with regional
focuses (Haywood and Boucher 2000; Lohmann and Feichter, 2005). Our study provides
a basis for a regionally focused exercise aimed at understanding the possible climatic
effects of aerosol on the global hydrological cycle.
Acknowledgments.
We thank Omar Torres for advice on the use of aerosol products; Daniel
Rosenfeld for critical discussions on the interpretation of precipitation change; and
Chunzai Wang for Atlantic SST indices. Authors acknowledge NCDC of NOAA
(http://lwf.ncdc.noaa.gov/oa/wmo/wdcamet-ncdc.html) for data provision of GPCP
version 2 data, GSFC of NASA (http://toms.gsfc.nasa.gov/aerosols/aerosols_v8.html) for
data provision of TOMS AI, the Data and Information Services Center (DSIC) of NASA
(http://disc.sci.gsfc.nasa.gov/) for data provision of TRMM and MODIS data, and
Remote Sensing Systems (http://www.remss.com/) for data provision of SSM/I data. This
study is funded by the NOAA office of Global Programs through the Cooperative
Institute for Marine and Atmospheric Studies (CIMAS).
29
References
Adler, R. F., G. J. Huffman, A. Chang, R. Ferraro, P. Xie, J. Janowiak, B. Rudolf, U.
Schneider, S. Curtis, D. Bolvin, A. Gruber, J. Susskind, P. Arkin, and E. Nelkin,
2003: The version-2 global precipitation climatology project (GPCP) monthly
precipitation analysis (1979-present), J. Hydrometeor., 4, 1147-1167.
Albrecht, B., 1989: Aerosols, Cloud Microphysics, and Fractional Cloudiness, Science,
245, 1227–1230.
Alexander, M. A., and J. D. Scott, 2002: The influence of ENSO on airsea interaction in
the Atlantic, Geophys. Res. Lett., 29(14), 1701, doi: 10.1029/2001GL014347.
Alpert, P., Y. J. Kaufman, Y. Shay-El, D. Tanre, A. da Silva and J. H. Joseph, 1998:
Quantification of dust- forced heating of the lower troposphere, Nature, 395, 367-370.
Anderson, B. E., and Coauthors, 1996: Aerosols from biomass burning over the tropical
South Atlantic region: distributions and impacts, J. Geophys. Res., 101, 24 117-24
137.
Bergman, J. W., and H. H. Hendon, 2000: Cloud radiative forcing of the low-latitude
tropospheric circulation: Linear calculations. J. Atmos. Sci., 57, 2225–2245.
Cakmur, R. V., R. L. Miller, and I. Tegen, 2001: A comparison of seasonal and
interannual variability of soil dust aerosols over the Atlantic Ocean as inferred by the
TOMS AI and AVHRR AOT retrievals, J. Geophys. Res., 106(D16), 18,287–18,303.
Carlson, T. N., 1979: Atmospheric turbudity in Saharan dust outbreaks as determined by
analysis of satellite brightness data. Mon. Wea. Rev., 107, 322-355.
Carlson, TN, and J. M. Prospero, 1972: The large-scale movement of Saharan air
outbreaks over the northern equatorial Atlantic. J. Appl. Meteor., 11, 283–297.
30
Chiapello, I. And C. Moulin, 2002: TOMS and METEOSAT satellite records of the
variability of Saharan dust transport over the Atlantic during the last two decades
(1979-1997), Geophys. Res. Lett. 29, 17-20.
Chiapello, I., C. Moulin, and J.M. Prospero, 2005: Understanding the long-term
variability of African dust transport across the Atlantic as recorded in both Barbados
surface concentrations and large-scale Total Ozone Mapping Spectrometer (TOMS)
optical thickness. J. Geophys. Res., 110, D18S10, doi:10.1029/2004JD005132.
Chiapello I., J. M. Prospero, J. R. Herman, and N. C. Hsu, 1999: Detection of mineral
dust over the North Atlantic Ocean and Africa with Nimbus 7 TOMS, Journal of
Geophysical Research, 104, 9277–9291
Chung, S.H., and J.H. Seinfeld, 2005: Climate response of direct radiative forcing of
anthropogenic black carbon. J. Geophys. Res., 110, D11102,
doi:10.1029/2004JD005441.
DeMott, P. J., K. Sassen, M. R. Poellot, D. Baumgardner, D. C. Rogers, S. D. Brooks, A.
J. Prenni, and S. M. Kreidenweis, 2003: African dust aerosols as atmospheric ice
nuclei, Geophys. Res. Lett., 30, 1732.
Diaz J. P., F. J. Exposito, C. J. Torres, F. Herrera, J. M. Prospero, and M. C. Romero,
2001: Radiative properties of aerosols in Saharan dust outbreaks using ground-based
and satellite data: Applications to radiative forcing. J. Geophys. Res, 106, 18403–
18416.
Draxler, R. R., and Rolph, G.D., HYSPLIT (HYbrid Single-Particle Lagrangian
Integrated Trajectory) Model access via NOAA ARL READY Website
31
(http://www.arl.noaa.gov/ready/hysplit4.html), NOAA Air Resources Laboratory,
Silver Spring, MD.
Duncan, B. N., R. V. Martin, A. C. Staudt, R. Yevich, and J. A. Logan, 2003: Interannual
and seasonal variability of biomass burning emissions constrained by satellite
observations, J. Geophys. Res., 108, 4100, doi:10.1029/2002JD002378.
Dunion, J.P., and C.S. Velden, 2004: The impact of the Saharan Air Layer on Atlantic
tropical cyclone activity. Bull. Amer. Meteor. Soc., 85, 353-365.
Dwyer, E., and Coauthors, 2000: Global spatial and temporal distribution of vegetation
fire as determined from satellite observations, Int. J. of Remote Sensing, 21, 1289 1302.
Fisher, L. B., 2004: Climatological validation of TRMM TMI and PR monthly rain
products over Oklahoma. J. Appl. Meteor., 43, 519–535.
Ginoux, P., J.M. Prospero, O. Torres, and M. Chin, 2004: Long-term simulation of global
dust distribution with the GOCART model: correlation with North Atlantic
Oscillation, Environmental Modelling and Software, 19, 113-128.
Gu, G., and R. F. Adler, 2006: Interannual rainfall variability in the tropical Atlantic
region, J. Geophys. Res., 111, D02106, doi:10.1029/2005JD005944.
Gu. G. and C. Zhang, 2002: Westward-propagating synoptic-scale disturbances and the
ITCZ. J. Atmos. Sci., 59, 1062-1075.
Hagos, S.M. and K.H. Cook, 2005: Influence of surface processes over Africa on the
Atlantic marine ITCZ and South American precipitation. J. Climate, 18, 4993-5010
Hao, W. M., and M. Liu, 1994: Spatial and temporal distribution of tropical biomass
burning. Global Biogeochem. Cycles, 8, 495–504.
32
Hartmann, D. L., H. H. Hendon, and R. A. Houze Jr., 1984: Some implications of the
mesoscale circulations in tropical cloud clusters for large-scale dynamics and climate.
J. Atmos. Sci., 41,113–121.
Hastenrath S., and L. Heller, 1977: Dynamics of climate hazards in northeast Brazil.
Quart. J. Roy. Meteor. Soc., 103, 77–92.
Haywood, J. and Boucher, O., 2000: Estimates of the direct and indirect radiative forcing
due to tropospheric aerosols: A review. Rev Geophys, 38, 513-543,
10.1029/1999RG000078.
Herman, J. R. and coauthors, 1997: Global distribution of UV-absorbing aerosols from
Nimbus 7/TOMS Data. J. Geophys. Res., 102, 16,911– 16,922.
Hoelzemann, J.J., M.G. Schultz, G.P. Brasseur, C. Granier, and M. Simon, 2004: Global
Wildland Fire Emission Model (GWEM): Evaluating the use of global area burnt
satellite data. J. Geophys. Res., 109, 1-18.
Huang, Y., W. L. Chameides, and R. E. Dickinson, 2007: Direct and indirect effects of
anthropogenic aerosols on regional precipitation over east Asia. J. Geophys. Res., 112,
D03212, doi:10.1029/2006JD007114.
Husar, R. B., J. M. Prospero and L. L. Stowe, 1997: Characterization of tropospheric
aerosols over the oceans with the NOAA advanced very high resolution radiometer
optical thickness operational product, J. Geophys. Res., 102, 16,889-16,909.
Huffman, G. J., and Coauthors, 1997: The Global Precipitation Climatology Project
(GPCP) Combined Precipitation Dataset. Bulletin of the American Meteorological
Society, 78, 5-20.
33
Hurrell, J. M. Decadal trends in the North Atlantic Oscillation regional temperatures and
precipitation. Science 269, 6760-679 (1995).
Intergovernmental Panel on Climate Change (IPCC), Climate Change 2007: The
Physical Sciences Basis, Cambridge University Press, New York, 2007.
Ito, A., and J. E. Penner, 2004: Global estimates of biomass burning emissions based on
satellite imagery for the year 2000. J. Geophys. Res., 109, D14S05,
doi:10.1029/2003JD004423.
Ito, A., and J. E. Penner, 2005: Historical emissions of carbonaceous aerosols from
biomass and fossil fuel burning for the period 1870–2000, Global Biogeochem.
Cycles, 19, GB2028, doi:10.1029/2004GB002374.
Jackson, D.L., and G.L. Stephens, 1995: A Study of SSM/I-Derived Columnar Water
Vapor over the Global Oceans. J. Climate, 8, 2025–2038.
Janicot S., A. Harzallah, B. Fontaine, and V. Moron, 1998: West African monsoon
dynamics and eastern equatorial Atlantic and Pacific SST anomalies (1970–88). J.
Climate, 11, 1874–1882.
Johnson, D. B., 1982: The role of giant and ultragiant aerosol particles in warm Rain
initiation. J. Atmos. Sci., 39, 448–460.
Karyampudi, VM, et al. 1999: Validation of the Saharan Dust Plume Conceptual Model
Using Lidar, Meteosat, and ECMWF Data. Bull. Amer. Meteor. Soc., 80, 1045-1075.
Kaufman, Y., O. Boucher, D. Tanre, M. Chin, L. Remer, and T. Takemura, 2005:
Aerosol anthropogenic component estimated from satellite data. Geophys. Res. Lett,
32, L17804, doi:10.1029/2005GL023125.
34
Kaufman, Y. J., and R. S. Fraser, 1997: The effect of smoke particles on clouds and
climate forcing. Science, 277, 1636-1639.
Kaufman, Y. J., I. Koren, L. A. Remer, D. Rosenfeld, and Y. Rudich, 2005a: The Effect
of smoke, dust and pollution aerosol on shallow cloud development over the Atlantic
ocean. Proceedings of the National Academy of Sciences, 102, 11 207-11 212.
Kaufman, Y. J., I. Koren, L. A. Remer, D. Tanre´, P. Ginoux, and S. Fan, 2005b: Dust
transport and deposition observed from the Terra-Moderate Resolution Imaging
Spectroradiometer (MODIS) spacecraft over the Atlantic Ocean, J. Geophys. Res.,
110, D10S12, doi:10.1029/2003JD004436.
Kaufman, Y. J., D. Tanre and O. Boucher, 2002: A satellite view of aerosols in the
climate system. Review for Nature, 419, 215-223.
Kaufman Y. J., D. Tanre, L. A. Remer, E. F. Vermote, A. Chu, and B. N. Holben, 1997:
Operational remote sensing of tropospheric aerosol over land from EOS moderate
resolution imaging spectrometer. J. Geophys. Res., 102, 17051–17067.
Koren, I., Y. J. Kaufman, D. Rosenfeld, L. A. Remer, and Y. Rudich, 2005: Aerosol
invigoration and restructuring of Atlantic convective clouds, Geophys. Res. Lett., 32,
L14828, doi:10.1029/2005GL023187.
Kyte, E.A., G.D. Quartly, M.A. Srokosz and M.N. Tsimplis, 2006: Interannual variations
in precipitation: The effect of the North Atlantic and Southern Oscillations as seen in
a satellite precipitation data set and in models, J. Geophys. Res., 111, D24113, doi:
101029/2006JD007138.
Lamb P. J., 1978: Large-scale tropical Atlantic surface circulation patterns associated
with Subsaharan weather anomalies. Tellus, 30, 240–251.
35
Lohmann, U. and J. Feichter, 2005: Global indirect aerosol effects: a review. Atmos.
Chem. Phys., 5, 715–737.
Lohmann, U., and G. Lesins, 2002: Stronger constraints on the anthropogenic indirect
aerosol effect. Science, 298, 1012–1016.
Matsui, T., H. Masunaga, R. A. Pielke Sr., and W.-K. Tao, 2004: Impact of aerosols and
atmospheric thermodynamics on cloud properties within the climate system, Geophys.
Res. Lett., 31, L06109, doi:10.1029/2003GL019287.
Menon, S., and A. Del Genio, 2007: Evaluating the impacts of carbonaceous aerosols on
clouds and climate. In: An Interdisciplinary Assessment: Human-Induced Climate
Change [Schlesinger, M., et al. (eds.)]. Cambridge University Press, Cambridge, UK.
Menon, S., 2004: Current uncertainties in assessing aerosol effects on climate. Ann. Rev.
Environ. Resources, 29, 1-30, doi:10.1146/annurev.energy.29.063003.132549.
Menon, S., J.E. Hansen, L. Nazarenko, and Y. Luo, 2002: Climate effects of black carbon
aerosols in China and India. Science, 297, 2250-2253.
Miller, R.L., and I. Tegen, 1998: Climate response to soil dust aerosols. J. Climate, 11,
3247-3267.
Moulin, C., and I. Chiapello, 2006: Impact of human-induced desertification on the
intensification of Sahel dust emission and export over the last decades, Geophys. Res.
Lett., 33, L18808, doi:10.1029/2006GL025923.
Philander, S.G.H. El Niño, La Niña and the Southern Oscillation (Academic Press, San
Diego, CA, 1990).
36
Prospero, J. M. 1996: Saharan dust transport over the North Atlantic Ocean and
Mediterranean: An overview, in The Impact of Desert Dust Across the Mediterranean,
edited by S. Guerzoni and R. Chester, Kluwer, Dordrecht.
Prospero, J. M.1999: Long-term measurements of the transport of African mineral dust to
the southeastern United States: Implications for regional air quality, J. Geophys. Res.,
104, 15,917– 15,927.
Prospero, J. M., and J. P. Lamb, 2003: African droughts and dust transport to the
Caribbean: Climate change and implications, Science, 302, 1024-1027.
Prospero J M, Glaccum R A, Nees R T, 1981: Atmospheric transport of soil dust from
Africa to south America. Nature, 289, 570-572
Prospero, J.M. and Nees, R.T., 1986: Impact of North African drought and El Nino on
mineral dust in the Barbados trade winds. Nature, 320, 735–738.
Ramanathan, V., Crutzen, P. J., Kiehl, J. T., and Rosenfeld, D., 2001: Aerosols, climate,
and the hydrological cycle, Science, 294, 2119-2124.
Robertson A. W., S. Kirshner, and P. Smyth, 2004: Downscaling of daily rainfall
occurrence over northeast Brazil using a Hidden Markov Model. J. Climate, 17,
4407–4424.
Rosenfeld, D., 1999: TRMM observed first direct evidence of smoke from forest fires
inhibiting rainfall. Geophys. Res. Lett., 26, 3105-3108.
Rosenfeld, D., 2000: Suppression of rain and snow by urban and industrial air pollution,
Science, 287, 1793-1796
Rosenfeld D. and H. Farbstein, 1992: Possible influence of desert dust on seedability of
clouds in Israel. Journal of Applied Meteorology, 31, 722-731.
37
Rosenfeld, D., Y. Rudich, and R. Lahav, 2001: Desert dust suppressing precipitation -- a
possible desertification feedback loop. Proceedings of the National Academy of
Sciences, 98, 5975-5980.
Sassen, K., P. J. DeMott, J. M. Prospero, and M. R. Poellot, 2003: Saharan dust storms
and indirect aerosol effects on clouds: CRYSTAL-FACE results, Geophys. Res. Lett.,
30, 1633.
Schumacher, C., R.A. Houze, and I. Kraucunas, 2004: The Tropical Dynamical Response
to Latent Heating Estimates Derived from the TRMM Precipitation Radar. J. Atmos.
Sci., 61, 1341–1358. Segal Y, A. Khain, M. Pinsky, and A. Sterkin: 2004: Sensitivity
of raindrop formation in ascending cloud parcels to cloud condensation nuclei and
thermodynamic conditions. Quart. J. Roy. Met Soc., 130, 561-581.
Servain, J. Simple climate indices for the tropical Atlantic Ocean and some applications.
J. Geophys. Res. 96, 15137-15146 (1991).
Swap, R. J., and Coauthors, 1996: Long-range transport of southern African aerosols to
the tropical South Atlantic, J. Geophys. Res., 101 23777-23793.
Takemura, T., T. Nozawa, S. Emori, T. Y. Nakajima, and T. Nakajima, 2005: Simulation
of climate response to aerosol direct and indirect effects with aerosol transportradiation model. J. Geophys. Res., 110, D02202, doi:10.1029/2004JD005029.
Tao, W.-K., X. Li, A. Khain, T. Matsui, S. Lang, and J. Simpson, 2007: Role of
atmospheric aerosol concentration on deep convective precipitation: Cloud-resolving
model simulations, J. Geophys. Res., 112, D24S18, doi:10.1029/2007JD008728.
Tao, W.K., E.A. Smith, R.F. Adler, Z.S. Haddad, A.Y. Hou, T. Iguchi, R. Kakar, T.N.
Krishnamurti, C.D. Kummerow, S. Lang, R. Meneghini, K. Nakamura, T. Nakazawa,
38
K. Okamoto, W.S. Olson, S. Satoh, S. Shige, J. Simpson, Y. Takayabu, G.J. Tripoli,
and S. Yang, 2006: Retrieval of Latent Heating from TRMM Measurements. Bull.
Amer. Meteor. Soc., 87, 1555–1572.
Thompson, R.M., S.W. Payne, E.E. Recker, and R.J. Reed, 1979: Structure and
Properties of Synoptic-Scale Wave Disturbances in the Intertropical Convergence
Zone of the Eastern Atlantic. J. Atmos. Sci., 36, 53–72.
Torres, O., P.K. Bhartia, J.R. Herman, A. Sinyuk, P. Ginoux, and B. Holben, 2002: A
Long-Term Record of Aerosol Optical Depth from TOMS Observations and
Comparison to AERONET Measurements. J. Atmos. Sci., 59, 398-413.
Torres, O., P. K. Bhartia, J. R. Herman, Z. Ahmad and J. Gleason, 1998: Derivation of
aerosol properties from satellite measurements of backscattered ultraviolet radiation:
theoretical basis. J. Geophys. Res., 103, 17 099–17 110.
Twomey, S. A., M. Piepgrass, and T. L. Wolfe, 1984: An assessment of the impact of
pollution on the global albedo, Tellus, 36, 356–366.
Wang, C., 2002: Atlantic climate variability and its associated atmospheric circulation
cells. Journal of Climate, 15, 1516-1536.
Wang, C., 2005: A modeling study of the response of tropical deep convection to the
increase of cloud condensation nuclei concentration: 1. Dynamics and microphysics,
J. Geophys. Res., 110, D21211, doi:10.1029/2004JD005720.
Wang, C., 2007: Impact of direct radiative forcing of black carbon aerosols on tropical
convective precipitation, Geophys. Res. Lett., 34, L05709,
doi:10.1029/2006GL028416.
39
Washington, R. and M.C. Todd, 2005: Atmospheric controls on mineral dust emission
from the Bodélé Depression, Chad: the role of the low level jet, Geophys. Res. Lett.,
32, L17701.
Wentz F. J. 1997: A well-calibrated ocean algorithm for SSM/I, J. Geophys. Res., 102,
8703-8718.
Wu, L., and Z. Liu 2002: Is tropical Atlantic variability driven by the North Atlantic
Oscillation? Geophys. Res. Lett. 29(13), 1653, 10.1029/2002GL014939.
Wurzler, S, T.G. Reisin, Z. Levin, 2000: Modification of mineral dust particles by clouds
processing and subsequent effects on drop size distributions, J. Geophys. Res., 105,
4501-4512.
Xie, T. –P, and J. Carton, 2004: Tropical Atlantic variability: Patterns, mechanisms, and
impacts. Earth Climate: The Ocean–Atmosphere Interaction, Geophys. Monogr., 147,
121–142.
Yin, Y. and Chen, L.: Long-range transport of mineral aerosols and its absorbing and
heating effects on cloud and precipitation: a numerical study, Atmos. Chem. Phys.
Discuss., 7, 3203-3228, 2007.
Yoshioka M., N. Mahowald, A. Conley, W. Collins, D. Fillmore, C. Zender, and D.
Coleman, 2007: Impact of Desert Dust Radiative Forcing on Sahel Precipitation:
Relative importance of dust compared to sea surface temperature variations,
vegetation changes and greenhouse gas warming, Journal of Climate, 20,
DOI:10.1175/JCLI4056.1,1445-1467
40
Yin, X. G., A. Gruber, and P. Arkin, 2004: Comparison of the GPCP and CMAP merged
gauge-satellite monthly precipitation products for the period 1979-2001. Journal of
Hydrometeorology, 5, 1207-1222
Zebiak, S. E., 1993: Air-Sea Interaction in the Equatorial Atlantic Region. J. Climate, 6,
1567-1568.
Zhang, C. and J. Pennington, 2004: African dry-air outbreaks. J. Geophys. Res., 109,
D20108, doi:10.1029/2003JD003978.
41
List of Figures
FIG. 1 Analysis domains and climatological means of precipitation (mm/day, colors) and
aerosols (contours, 10 × Aerosol Index, no unit). The Atlantic aerosol domain (AA,
solid boundaries) is used for domain-averaged aerosol. The Atlantic marine ITCZ
domain (AMI, centered at the dashed line at 5˚N and bounded to the north and
south by dotted lines at 15˚N and 5˚S) is used for domain-averaged precipitation.
The AA domain is divided into two subdomains, separated at 5˚N (dashed line):
NAA to the north and SAA to the south.
FIG. 2 Time-Latitude diagrams of (a) mean seasonal cycles and (b) variance in
precipitation (mm/day, colors) and aerosol index (contours) zonally averaged over
the AA domain (see Fig. 1). The center of the ITCZ rain band is marked by black
dashed lines.
FIG. 3 Monthly mean time series (1979-2007) of (a) AA-domain mean TOMS Aerosol
Index (AI) in black and MODIS Aerosol Optical Depth (AOD) in gray, and (b) the
AMI-domain mean GPCP precipitation in black and TRMM precipitation in gray.
Three analysis periods are marked above panel (a): the 1st period of 1979-2000; 2nd
period of 1988-2000, and 3rd period of 2000-2007.
FIG. 4 (a) Fractional variance (%) of precipitation anomalies related to aerosol and
ENSO, NAO, tropical Atlantic zonal SST mode (ZonSST) and meridional SST
mode (MerSST), and AA-domain averaged aerosol as function of month. The
aerosol-related variance is shown for both cases with aerosol anomalies detrended
and not detrended; (b) Fractional variance (%) of aerosol anomalies related to the
climate factors.
42
FIG. 5 Time series (1979-2000) of AMI-domain averaged precipitation normalized
anomalies (dashed line) and AA-domain averaged aerosol normalized anomalies
(solid line).
FIG.6 Scatter diagram of AA-domain averaged aerosol normalized anomalies and AMIdomain averaged precipitation normalized anomalies. Solid line indicates a linear fit
(correlation coefficient R = 0.324, significant at the 99% confidence level).
FIG.7 Correlation coefficient (colors) and confidence level (95 and 99%, white contours)
between AA-domain averaged aerosol normalized anomalies and precipitation
normalized anomalies at each grid.
FIG.8 Difference composites of precipitation and aerosol normalized anomalies between
the top and bottom aerosol tercile months: (a) and (b) are the annual mean and
seasonal cycle of the precipitation composite, respectively; (c) and (d) are the
annual mean and seasonal cycle of the aerosol composite, respectively. Dashed
lines mark the center of the ITCZ rain band.
FIG.9 Probability distribution functions (PDFs) of precipitation normalized anomalies
from the composite difference (bars), random sampling (solid line), and interannual
variability (dashed line). The mean and one standard deviation for the difference
composite are marked at the top by arrows. The mean of the random sample is zero,
and its one standard deviation is marked by squares. The mean and one standard
deviation for the interannual variability are marked by straight vertical dashed line
and circled crosses, respectively. The bin width for the PDFs was determined as the
difference of data maximum and minimum divided by square root of sample size.
43
FIG. 10 (a) – (e) Differences in monthly precipitation between the top and bottom aerosol
tercile months as a function of rain rate averaged in five sub-domains (A-E) shown
in (f). Total difference in monthly precipitation (mm) is given for each sub-domain.
The thick solid lines in (a)-(e) are 2-mm/d moving averages.
FIG. 11 Spatial distribution of annual mean fractional variance (%) in precipitation
normalized anomalies related to aerosol. Contour interval 5.
FIG. 12 Precipitation difference composite with water vapor effect removed along with
climate factors through a multi-variable regression for the second analysis period.
FIG.13 Difference composite of TRMM precipitation normalized anomalies between top
and bottom tercile months of MODIS AOD normalized anomalies: (a) Annual mean;
(b) seasonal cycle. The center of the ITCZ rain band is marked by black dashed
lines.
FIG.14 Backward trajectory analysis for 10 days using HYSPLIT4 model with NOAA
reanalysis data for (a) January and February,1997 and (b) April and May, 1987.
Starting point is [0, 35°W] with three height levels: 500 m (in red lines), 1500 m (in
light green lines), and 3000 m (in blue lines).
FIG.15 Same as Fig.8a except climatology mean sea surface temperature (˚C) contours
are added.
List of Tables
TABLE 1 Testing cases for the selection of aerosol and precipitation domains
44
15
21
7
6
4
3
3
3
3
8
2
3
6
7
3
17
20
SAA
5
5
3
-15
-60
4
9
4
2
-5
6
10
4
4
0
-10
13
8
13
5
10
8
16
15
11
3
5
9
22
18
4
112
11
AMI
6 7
1
Latitude
13
8
5
NAA
10
18
12
7
6
15
10
23
9
8
18
18
17
16
20
19
19
1517
20
7
12
6
14
5
18
20
25
11
16
List of Figures
-50
-40
-30
-20
Longitude
-10
0
FIG. 1 Analysis domains and climatological means of precipitation (mm/day, colors) and
aerosols (contours, 10 × Aerosol Index, no unit). The Atlantic aerosol domain (AA, solid
boundaries) is used for domain-averaged aerosol. The Atlantic marine ITCZ domain
(AMI, centered at the dashed line at 5˚N and bounded to the north and south by dotted
lines at 15˚N and 5˚S) is used for domain-averaged precipitation. The AA domain is
divided into two subdomains, separated at 5˚N (dashed line): NAA to the north and SAA
to the south.
45
(a)
12
8
10
7
3.5 4
5
7
4.5
6
4
5
4
8
3
7
7
5
6
2
2
7
6
9
5
-15
1
4
6
5
4.5
5
3.
5
4.
9
-10
6
4 56
3
6
5
7
3.
5
4
8
8
12
16
1614
18
20
4.5
4.5
2.5
2
5
3.3
Latitude
4
910
8 7
6
4.5
5
5
3. 3
4
8
6
5
7
-5
9
14
12
4.5 4
3
3.45
5 4. 5
4
4
056
12
5
3.
5
18
10
7
8
9
10
9
8
5 6 7
1.5
9
10
6
3
44.55
15
7
8
9
14
20
2
5
2.
3.5
1
25
5
6 7 8
Months
9 10 11 12 1
(b)
896
14
2
12
3. 5
5
5
2.
2.5
4
3
1.5
3
5
3
43. 5
4
2.5
4
3
4.55
3
2
6
4.5
3.5
55
6 7 8
Months
4
4.5
3
2.5. 5 3
2.5
3.5
2
2. 5
4
4.
5
9
12 14
9
22
12 10
7
7
5
4. 5
5
3
2
3.5
Latitude
7
8
3
1. 5
2
2
3
2
5
3
4
6
4
3.5
2.5
5
6
2
2.5
5
87 89
4.54
5
4.5 4
4.5
5
-15
1
4
6
-5
-10
6
3
3.5
4.5
10
87
6 4
12
4.5
10 5 3.5 2.5
78
2
3
2. 5 3
2
53
3.
0
4
2
10
5
5
1
14 18
16
9
5
6
4
10
10
12
18
8
10
20
4.5
6
15
16
1.5
2
5
2. 5
3.
20
4. 5
7
8 9
4
5
25
1
5
45
9 10 11 12 1
FIG. 2 Time-Latitude diagrams of (a) mean seasonal cycles and (b) variance in
precipitation (mm/day, colors) and aerosol index (contours) zonally averaged over the
AA domain (see Fig. 1). The center of the ITCZ rain band is marked by black dashed
lines.
46
(a)
1st Period
2nd Period
3rd Period
0.4
TOMS AI
1.2
1
0.3
0.8
0.2
0.6
0.4
TOMS AI
0.2
0.1
MODIS AOD
08
07
20
20
05
20
06
04
20
20
02
20
20
03
01
20
20
00
99
98
19
19
96
97
19
19
19
95
94
93
19
19
91
19
92
90
19
19
88
19
89
87
19
86
19
19
84
19
19
85
83
82
19
19
80
19
19
19
81
0
79
0
MODIS AOD
0.5
1.4
5
4
3
2
1
GPCP Precipitation
TRMM Precipitation
08
07
20
20
06
20
05
04
20
20
03
02
20
20
00
01
20
20
99
98
19
19
97
19
96
95
19
19
94
93
19
19
92
19
91
90
19
19
89
19
88
19
87
86
19
19
85
84
19
19
83
82
19
19
81
19
19
19
80
0
79
Precipitation (mm/day)
(b)
FIG. 3 Monthly mean time series (1979-2007) of (a) AA-domain mean TOMS Aerosol
Index (AI) in black and MODIS Aerosol Optical Depth (AOD) in gray, and (b) the AMIdomain mean GPCP precipitation in black and TRMM precipitation in gray. Three
analysis periods are marked above panel (a): the 1st period of 1979-2000; 2nd period of
1988-2000, and 3rd period of 2000-2007.
47
(a)
Fractional Variance (%)
50
ENSO
ZonSST
MerSST
NAO
Aerosol
Aerosol-detrended
40
30
20
10
0
1
2
3
4
5
6 7
Month
8
9 10 11 12
(b)
Fractional Variance (%)
50
ENSO
ZonSST
MerSST
NAO
40
30
20
10
0
1
2
3
4
5
6 7
Month
8
9 10 11 12
FIG. 4 (a) Fractional variance (%) of precipitation anomalies related to aerosol and
ENSO, NAO, tropical Atlantic zonal SST mode (ZonSST) and meridional SST mode
(MerSST), and AA-domain averaged aerosol as function of month. The aerosol-related
variance is shown for both cases with aerosol anomalies detrended and not detrended; (b)
Fractional variance (%) of aerosol anomalies related to the climate factors.
48
4
2
Aerosol
1
2
0
1
-1
0
-2
-1
-3
-2
-4
19
79
19
80
19
81
19
82
19
83
19
84
19
85
19
86
19
87
19
88
19
89
19
90
19
91
19
92
19
93
19
94
19
95
19
96
19
97
19
98
19
99
20
00
20
01
Precipitation
3
FIG. 5 Time series (1979-2000) of AMI-domain averaged precipitation normalized
anomalies (dashed line) and AA-domain averaged aerosol normalized anomalies (solid
line).
49
Aerosols
Precipitation
Precipitation Normalized Anomalies
2
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
1.5
1
0.5
0
-0.5
-1
-1.5
-2
-2
-1
0
1
Aerosol Normalized Anomalies
2
FIG.6 Scatter diagram of AA-domain averaged aerosol normalized anomalies and AMIdomain averaged precipitation normalized anomalies. Solid line indicates a linear fit
(correlation coefficient R = 0.324, significant at the 99% confidence level).
50
95
95
0.2
99
95
99
5
99
99
-0.1
-0.2
95
95
-0.3
95
99
95
-15
-60
99
95
95
99
-5
99
95
0
99
99
0
-10
0.1
95
95
10
95
Latitude
0.3
95
99
15
95
20
99
25
-40
-20
0
Longitude
FIG.7 Correlation coefficient (colors) and confidence level (95 and 99%, white contours)
between AA-domain averaged aerosol normalized anomalies and precipitation
normalized anomalies at each grid.
51
(a)
Latitude
25
20
1.5
15
1
10
0.5
5
0
0
-0.5
-5
-1
-10
-15
-60
-1.5
-40
-20
0
Longitude
(b)
Latitude
25
20
1.5
15
1
10
0.5
5
0
0
-0.5
-5
-1
-10
-1.5
-15
1 2 3 4 5 6 7 8 9 10 11 12 1
Longitude
52
(c)
Latitude
25
20
1.5
15
1
10
0.5
5
0
0
-0.5
-5
-1
-10
-15
-60
-1.5
-40
-20
0
Longitude
(d)
Latitude
25
20
1.5
15
1
10
0.5
5
0
0
-0.5
-5
-1
-10
-1.5
-15
1 2 3 4 5 6 7 8 9 10 11 12 1
Month
FIG.8 Difference composites of precipitation and aerosol normalized anomalies between
the top and bottom aerosol tercile months: (a) and (b) are the annual mean and seasonal
cycle of the precipitation composite, respectively; (c) and (d) are the annual mean and
seasonal cycle of the aerosol composite, respectively. Dashed lines mark the center of the
ITCZ rain band.
53
Probability Distributions (%)
30
Difference Composite
Interannual Variability
Random Sampling
25
20
15
10
5
0
-3
-2
-1
0
1
2
Precipitation Normalized Anomalies
3
FIG.9 Probability distribution functions (PDFs) of precipitation normalized anomalies
from the composite difference (bars), random sampling (solid line), and interannual
variability (dashed line). The mean and one standard deviation for the difference
composite are marked at the top by arrows. The mean of the random sample is zero, and
its one standard deviation is marked by squares. The mean and one standard deviation for
the interannual variability are marked by straight vertical dashed line and circled crosses,
respectively. The bin width for the PDFs was determined as the difference of data
maximum and minimum divided by square root of sample size.
54
100
0
-100
-200
0
10
20
Rain Rate (mm/d)
30
Diff. in Monthly Precip. (mm)
(c) For subdomain C: -174 mm
0
-100
10
20
Rain Rate (mm/d)
30
(e) For subdomain E: -1234 mm
0
-100
-200
0
10
20
Rain Rate (mm/d)
30
100
0
-100
-200
0
10
20
Rain Rate (mm/d)
30
(f) Subdomains definition
25
100
1.5
20
1
15
Latitude
Diff. in Monthly Precip. (mm)
100
(d) For subdomain D: -594 mm
100
-200
0
Diff. in Monthly Precip. (mm)
(b) For subdomain B: -2007 mm
Diff. in Monthly Precip. (mm)
Diff. in Monthly Precip. (mm)
(a) For subdomain A: -3127 mm
0
10
C
5
D
B
A
0
0.5
0
E
-0.5
-5
-100
-1
-10
-200
0
-15
-60
10
20
Rain Rate (mm/d)
-1.5
-40
-20
0
Longitude
30
FIG. 10 (a) – (e) Differences in monthly precipitation between the top and bottom aerosol
tercile months as a function of rain rate averaged in five sub-domains (A-E) shown in (f).
Total difference in monthly precipitation (mm) is given for each sub-domain. The thick
solid lines in (a)-(e) are 2-mm/d moving averages.
55
25
20
15
Latitude
15
10
5
10
0
-5
5
-10
-15
-60
-40
-20
0
Longitude
FIG. 11 Spatial distribution of annual mean fractional variance (%) in precipitation
normalized anomalies related to aerosol. Contour interval 5.
56
Latitude
25
20
1.5
15
1
10
0.5
5
0
0
-0.5
-5
-1
-10
-15
-60
-1.5
-40
-20
0
Longitude
FIG. 12 Precipitation difference composite with water vapor effect removed along with
climate factors through a multi-variable regression for the second analysis period.
57
(a)
Latitude
25
20
1.5
15
1
10
0.5
5
0
0
-0.5
-5
-1
-10
-1.5
-15
-60
-40
-20
0
Longitude
(b)
Latitude
25
20
1.5
15
1
10
0.5
5
0
0
-0.5
-5
-1
-10
-15
-1.5
2
4
6
8
Month
10
12
FIG.13 Difference composite of TRMM precipitation normalized anomalies between top
and bottom tercile months of MODIS AOD normalized anomalies: (a) Annual mean; (b)
seasonal cycle. The center of the ITCZ rain band is marked by black dashed lines.
58
(a)
30oN
15oN
0o
15oS
30oS
75oW
50oW
25oW
0o
25oE
50oE
75oW
50oW
25oW
0o
25oE
50oE
(b)
30oN
15oN
0o
15oS
30oS
FIG.14 Backward trajectory analysis for 10 days using HYSPLIT4 model with NOAA
reanalysis data for (a) January and February,1997 and (b) April and May, 1987. Starting
point is [0, 35°W] with three height levels: 500 m (in red lines), 1500 m (in light green
lines), and 3000 m (in blue lines).
59
26
.4
Latitude
10
.6
25
15
26
.8
27
.2
26
26 .
4
26 .8
27 .2
27 .6
22
1 .6
212.2
20
8
22 .
24 .6
2
24 4 .4
25 .8
.2
25 5 .626
1.5
22
.4
23 .6
24
2244.8.4
25
25.6.2
26
26 .4
1
0.5
28
27 .6
5
0
0
27.628
27 .2 26.8
26.4
27 .6
26
-5
27
.2
.8
26 26 .4 6
2
8
-40
6
25 .
2
.
25
-20
24 .8
24 .4
246
-0.5
-1
26
-10
-15
-60
27
-1.5
0
Longitude
FIG.15 Same as Fig.8a except climatology mean sea surface temperature (˚C) contours
are added.
60
TABLE 1 Testing cases for the selection of aerosol and precipitation domains
Case Number
Aerosol
Precipitation
Domain
Domain
A
AA
AMI
> 99%
B
AA
AA
> 99%
C
NAA
AMI
> 99%
D
SAA
AMI
> 99%
E
NAA
NAA
< 95%
F
SAA
SAA
> 99%
61
Confidence Level