Estimation of aerosol direct radiative effects for all-sky conditions from CERES and MODIS observations

Journal of Atmospheric and Solar-Terrestrial Physics 102 (2013) 311–320

Contents lists available at SciVerse ScienceDirect

Journal of Atmospheric and Solar-Terrestrial Physics journal homepage: www.elsevier.com/locate/jastp

Estimation of aerosol direct radiative effects for all-sky conditions from CERES and MODIS observations Hye-Ryun Oh a, Yong-Sang Choi b,n, Chang-Hoi Ho a, Myeong-Jae Jeong c a

School of Earth and Environmental Sciences, Seoul National University, Seoul, South Korea Department of Environmental Science and Engineering, Ewha Womans University, Seoul, South Korea c Department of Atmospheric and Environmental Sciences, Gangneung-Wonju National University, Gangneung, South Korea b

art ic l e i nf o

a b s t r a c t

Article history: Received 14 January 2013 Received in revised form 6 June 2013 Accepted 17 June 2013 Available online 28 June 2013

Satellite observations have shown the global average of the aerosol direct radiative effect (DRE) at the top of the atmosphere to be approximately −5.0 W m−2. Although there is a general consensus on this quantity, it is essentially biased toward clear-sky conditions. To circumvent this limitation, the present study introduces a new method for retrieving the global DRE of aerosol over the region of 601S–601N for all-sky conditions (both clear and cloudy skies). The all-sky DRE was calculated on a monthly basis by combining the measured DRE for a clear sky and the simulated DRE for a cloudy sky in 11
11 grids. For the measured clear-sky DRE, we employed aerosol, cloud, and radiation fluxes from the Cloud and Earth's Radiant Energy System (CERES) instrument and the Moderate Resolution Imaging Spectroradiometer (MODIS) onboard the Terra satellite for May 2000–December 2005. For the simulated cloudy-sky DRE, we performed radiative transfer modeling with the MODIS cloud properties in addition to the aerosol optical properties independently estimated in this study that include asymmetry factor and single scattering albedo. The results show that the global mean7 standard deviation of DRE for the all-sky scene is −3.1 7 1.0 W m−2, which is weaker than that for the clear-sky only. This is in good agreement with the global estimates from previous studies based on different methods. The main advantage of our method is near-real-time estimation of monthly global all-sky DRE that has physical consistency with the CERES data. & 2013 Elsevier Ltd. All rights reserved.

Keywords: All-sky aerosol direct radiative effect Satellite observation Cloud

1. Introduction Atmospheric aerosols are known to directly alter the Earth's radiation budget by scattering and absorbing solar energy (Ramanathan et al., 2001; Kaufman et al., 2002). As an important measure of the radiative impact of aerosols, the aerosol direct radiative effect (hereafter DRE) has been defined as the difference between the values of reflected solar energy with and without the presence of aerosols. Clearly, quantification of DRE is essential for understanding the state of the present-day climate and for enhancing the reliability of future climate simulations (Bates et al., 2006; Patadia et al., 2008; Myhre, 2009). Numerous attempts have been made to quantify DRE by using a suite of instruments on the ground, ships, and aircrafts as well as satellites (Haywood et al., 1999; Bellouin et al., 2005; Remer and Kaufman, 2006). Among these measurements, only satellites can provide an extensive spatial coverage. A mean7standard error of the global DRE measured from satellites is generally acknowledged to be approximately

n

Corresponding author. Tel.: +82 2 3277 4461. E-mail address: [email protected] (Y.-S. Choi).

1364-6826/$ - see front matter & 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.jastp.2013.06.009

−5.570.2 W m−2 over ocean and −4.970.7 W m−2 over land. Strictly speaking, however, these values are confined to clear-sky, cloud-free conditions (Yu et al., 2006; Intergovernmental Panel on Climate Change, 2007). Such estimations of DRE are restricted to clear-sky (DREclear ) scenes as a result of fundamental limitations of measurement technology to distinguish the radiative contribution of the aerosols from that of clouds under cloudy conditions. Therefore, it is important to distinguish the differences in the current satellite retrieval of global DRE from global DRE for all-sky (i.e., both clearsky and cloudy-sky) scenes (DREall ). It should be noted that global cloud coverage is more than 60% (Rossow et al., 1993; Choi and Ho, 2009), and an enormous amount of aerosols is overlapped with clouds (Devasthale and Thomas, 2011). These factors imply that the inclusion of cloudy-sky scenes in the estimation of DRE could result in significant changes in DRE magnitude. Several studies reported that because clouds can immediately affect DRE by attenuating light reflected back to space by aerosols, DREall may differ substantially from DREclear (Podgorny and Ramanathan, 2001; Keil and Haywood, 2003; Chand et al., 2009; Chung, 2012). Chand et al. (2009) discovered that the negative sign of DRE turns positive when cloud fraction is more than 40% over

312

H.-R. Oh et al. / Journal of Atmospheric and Solar-Terrestrial Physics 102 (2013) 311–320

the Atlantic Ocean. Similar results were reported in some regions following dust storms and biomass burning (e.g., Podgorny and Ramanathan, 2001; Chand et al., 2009). Expansion to a “global” scale by Chen et al. (2011) showed better estimation of anthropogenic DREall over the ocean when the effects of cloud fractions were considered. However, their method remains insufficient for explaining DREall owing to their assumption of negligible DRE of anthropogenic aerosols for cloudy regions (Oh et al., 2013). The objective of this study is to estimate bulk DREall of aerosols at the top-of-the-atmosphere (TOA) on a monthly basis by accounting for cloud influence on DRE. The computation of is carried out by our new method based on satellite remote sensing data. Our method requires satellite observations of energy flux, aerosols, and clouds. Though these data are subject to various uncertainty sources in the respective algorithms, our method would eventually give estimates of DRE similar to previous studies that basically depended on the aid of aerosol chemical transport models. We begin with brief explanations of data in the following section.

2. Satellite data We used daily level-3 data from the Clouds and the Earth's Energy System (CERES, edition 2C) and Moderate Resolution Imaging Spectroradiometer (MODIS, collection 5) onboard the Terra satellite from its launch time for about 6 years (May 2000 to December 2005). From 2006 the present version of the CERES data is no longer available. The domain was confined to 601S–601N since the satellite aerosol data contain relatively large uncertainties in high latitudes (4601) owing to high surface albedo. In this study, the domain of 601S–601N will be referred to as ‘the globe’ for convenience sake. This section contains detailed descriptions of each variable used in the analysis. Shortwave flux data at both the TOA and surface were obtained from the CERES Monthly Gridded Radiative Fluxes and Clouds (FSW). These radiation fluxes available for pristine (without clouds and aerosols),1 clear-sky (without clouds and with aerosols) and cloudy-sky (with clouds and aerosols) conditions were averaged over a 11
11 latitude–longitude grid. The flux for pristine conditions was based on Fu and Liou radiative transfer model (RTM) calculations (Charlock et al., 2005). Clear-sky conditions are defined as the cases in which cloud amounts of less than 0.1% occur within CERES's footprints (20 km at nadir). The bias in the instantaneous shortwave flux at the TOA is known to be 2–5% and 4–6% for clear-sky and cloudy-sky conditions, respectively (http:// eosweb.larc.nasa.gov/PRODOCS/ceres/SSF/Quality_Summaries). Aerosol optical depth (AOD)2 at 0.63 μm, surface albedo, and cloud products in the CERES FSW were also used. These data were essentially obtained from MODIS, and then collocated inside CERES flux footprints to retain consistency with the shortwave flux. Surface albedo was estimated from MODIS imager pixels, precalculated lookup tables, and Fu and Liou RTM (Fu and Liou, 1993), which is approximately 0.02 lower than that of the ground observation owing to a spatial resolution difference (Rutan et al., 2009). Cloud properties—cloud fraction (CF), cloud top temperature (CTT), cloud optical depth (COD), and cloud droplet effective radius 1 The fluxes for pristine conditions were calculated from the instantaneous Cloud and Radiative Swath (CRS) using a 4-stream radiative transfer model (Fu and Liou, 1993). 2 If MODIS AOD was not available for the footprint with cloudiness more than 50% or deserts, AOD was created from the National Center for Atmospheric Research (NCAR) Model for Atmospheric Transport and Chemistry (MATCH) (http://eosweb.larc.nasa.gov/PRODOCS/ceres/CRS/Quality_Summaries/CER_CRS_Ter ra_Edition2B.html).

(ER)—were obtained from the MODIS/Terra Level-3 gridded atmospheric data product (MOD08), with a 11
11 latitude and longitude resolution.3 The major source of errors for cloud fraction is associated with the cloud mask with an uncertainty of 15% (Ackerman et al., 2008). Dong et al. (2008) reported that the uncertainty of COD and ER retrieved from MODIS is 8% and 15%, respectively. Since the present study focuses on estimating monthly DREall , all flux, cloud, and aerosol data used were the monthly averages. Virtually, the monthly averages indicate the arithmetical means of instantaneous values during the Terra satellite overpass time (10:30 am local time). In addition, we used a fixed aerosol vertical profile that is the four-year (2007−2010) global mean. The profile was obtained from the Cloud–Aerosol Lidar with Orthogonal Polarization (CALIOP) level-2 vertical feature mask (Winker et al., 2007) data.4 For calculation of DRE at TOA only, the detailed change of aerosol profiles may not be a primary concern. Rather, as we will show later, whether aerosols are above clouds or not is much more important. In fact, this profile is used only to constitute a look-up table to relate cloud properties with cloud effects on aerosols, and regional variation in DRE can still be realistically derived by vertically-integrated aerosol and cloud properties observed from MODIS. More importantly, the use of the fixed aerosol profile enables us to assess the uncertainty in DRE due to incorrect aerosol profiles. The result will be shown in Section 5. Meanwhile, cloud profiles were also employed to analyze the case that aerosols are above clouds. For CALIOP, clouds are detected by the layer-integrated particle depolarization ratio at 532 nm in addition to cloud-top and cloud-bottom temperatures (Hu et al., 2009). The CALIOP cloud mask was in good agreement with that from MODIS at more than 85% on the global scale (Holz et al., 2008). The horizontal (vertical) resolution of the data is 333 m (30 m) below 8.2 km, and 1000 m (60 m) from 8.2 km to 20.2 km.

3. Calculating grid-point DRE 3.1. DRE for all-sky Let us consider a 11
11 domain with fractional cloud coverage f. Recall that the DRE for the cloudy-sky scene cannot be measured by satellite instruments, while that for the clear-sky scenes can. Thus, the DRE over all-sky scenes DREall can be derived through the following equation: cloud DREall ðiÞ ¼ DREclear obs ðiÞð1−f ðiÞÞ þ DRERTM ðiÞf ðiÞ;

ð1Þ

DREclear obs

is the satellite-observed bulk DRE for the clear-sky where scene in the ith grid point and DREcloud RTM is the RTM-simulated DRE for the cloudy-sky scene. DREclear obs is defined as the difference in the net irradiances with and without the aerosol, and it is obtained by the difference between TOA shortwave fluxes in clear-sky conditions and pristine conditions from CERES. Here, the clear-sky 3 The CF is produced by using a cloud mask obtained from various threshold tests of spectral reflectance and brightness temperature (King et al., 1997; Platnick et al., 2003). The CTT developed by Menzel et al. (1983) was converted from the cloud top pressure retrieved by using a CO2 slicing technique with CO2 absorption channels within 13.2–14.4 μm. The uncertainty in CTT was less than 1 K considering error due to the MODIS 10.8 μm calibration and humidity profile (Dong et al., 2008). The total-column COD and ER were determined by using a combination of visible and near-IR channels (King et al., 1997). 4 The aerosol algorithm in CALIOP is based on the cluster analysis of the Aerosol Robotic Network (AERONET) (1993−2002) data to determine extinction-tobackscatter ratio for each aerosol type (Omar et al., 2009). The cluster was classified into six aerosol types—desert dust, smoke, clean continental, polluted continental, marine, and polluted dust—allowing for an uncertainty of 30% of extinction-tobackscatter ratio (Omar et al., 2009).

H.-R. Oh et al. / Journal of Atmospheric and Solar-Terrestrial Physics 102 (2013) 311–320

condition indicates the cloud amount less than 0.1%. DREcloud RTM may be calculated by including cloud effects on DRE. However, the DRE for the clear-sky scene should be used as a surrogate for DRE for the cloudy scene, assuming the DREs are the same within a given 11-grid. Hence DREcloud RTM can be replaced by the summation of aerosol DREclear obs and ΔDRE (i.e., the RTM-simulated cloud effect on DRE). Therefore, Eq. (1) becomes clear DREall ðiÞ ¼ DREclear obs ðiÞð1−f ðiÞÞ þ ðDREobs ðiÞ þ ΔDREðiÞÞf ðiÞ: all

ð2Þ

DREclear obs ,

indicating no In the case of f¼ 0, DRE is equal to cloud effect on aerosol DRE. It should be also noted that the DREclear obs is only available for cloud coverage of f≠1, though there was no case of f ¼1 in the monthly mean. Eq. (2) can be rearranged in a simpler form DREall ðiÞ ¼ DREclear obs ðiÞ þ ΔDREðiÞf ðiÞ

ðf ðiÞ≠1Þ;

ð3Þ

all

DREclear obs ,

f, that is, DRE is finally determined by three variables: and ΔDRE. The former two variables (DREclear and f) can be obs obtained from satellite observations, but the last variable (ΔDRE) should be supplied by radiative transfer modeling. A method to calculate ΔDRE will be introduced in the following section. 3.2. Calculation of the cloud effect on DRE (ΔDRE) The method to calculate ΔDRE consists of three parts: (1) the look-up table composition from RTM simulations, (2) the input data preparation of satellite observations, and (3) the fitting of the input data into the look-up table (Fig. 1). The look-up table search is actually to relate the observed cloud properties with ΔDRE. In composing the look-up table, we used the Santa Barbara DISORT Atmospheric Radiative Transfer (SBDART) model (Ricchiazzi et al., 1998). The integrated spectral range was set to 0.3–5.0 μm, and

313

the central wavelength to 0.63 μm, which is equivalent to those of the CERES spectrometer. The prescribed shortwave insolation in the SBDART model was scaled to the monthly observations of solar insolation in the FSW data. Composition of the look-up table requires characterization of various cloud and aerosol properties: clouds with COD (0−20), ER (0−30 μm), and surface temperature (ST) minus CTT (0−60 K) at step size of 1, 1, and 5, respectively; and aerosols with AOD (¼0.2), aerosol single scattering albedo (ASSA¼0.87 or 0.99), aerosol asymmetry factor (AAF¼0.52 or 0.76). The simulated clouds are single-layered overcast clouds with spherical droplets since most of aerosols are not radiatively interacting with high-level cold clouds. The low (i.e., 0.87 and 0.55) and high (i.e., 0.99 and 0.76) values for ASSA and AAF represent various aerosol extents (Chin et al., 2002; Andrews et al., 2006), and an AOD of 0.2 used in the simulation is comparable to the global mean value (Yu et al., 2006). Later, ΔDREs for the other ASSA and AAF values are calculated by bilinear interpolation at step size of 0.01 (0.01) due to the liner correspondence of ASSA (AAF) and ΔDRE (see Choi et al., 2009). And also, ΔDRE simulated for AOD¼ 0.2 is scaled by the ratio of (1−exp(AOD) to 1−exp(0.2)) where AOD is the observed grid-mean value, to account for regionally varying AOD of non- or weakly-absorbing aerosols (Zhao et al., 2011). The aerosol vertical profile is determined by the ratio of the maximum pixel number of total aerosols among all layers to the pixel number of total aerosols in each layer, represented in Fig. 3 as a black solid line. Here, the unit of the aerosol profile does not matter since the overall profile is exponentially scaled by the given AOD. For the atmospheric vertical profile, we set the temperature, water vapor, and ozone profiles for mid-latitude summer supplied in the RT model. However, these values do not matter because ΔDRE is nearly insensitive to these atmospheric profiles.

Step 1. RTM Pre-calculation Model boundary conditions Shortwave spectral range: 0.3–5.0 μm Cloud fraction (CF): 1 Cloud optical thickness (COT): 0–20,1 step size Cloud effective radius (ER): 0–30,1 step size Surface temperature minus Cloud top temperature (ST–CTT): 0–60,5 step size Single-layer cloud with spherical droplets Aerosol optical depth (AOD): 0.2 Aerosol single scattering albedo (ASSA): 0.87 or 0.99 Aerosol asymmetry factor (AAF): 0.52 or 0.76 Aerosol vertical profile: 4-year (2007–2010) mean aerosol profile observed from the CALIOP

Lookup table (LUT) setting Three-order polynomial regression function between cloud properties (COT, ER, ST minus CTT) and ∆DRE

Step 2. Input data Satellite observations CERES: shortwave flux, AOD, surface albedo MODIS: COD, ER, ST minus CTT Monthly average for all properties clear DRE satellite ( ) AOD

COD, ER, ST minus CTT AOD, surface albedo

ASSA, AAF

Step 3. Fitting input data into LUT Output ∆DRE Fig. 1. Flowchart for the methodology of estimating ΔDRE.

314

H.-R. Oh et al. / Journal of Atmospheric and Solar-Terrestrial Physics 102 (2013) 311–320

Using ΔDREs and the corresponding cloud and aerosol conditions in the look-up table, we can get a third-order polynomial function relating all the variables ΔDREω; g ¼ aω;g x3 þ bω; g x2 þ cω; g x þ dω; g

and high uncertainties in observed cloud and aerosol properties at higher latitudes (4601) preclude the calculation of ASSA, AAF, and finally ΔDRE. Over the relatively bright surface in the range of approximately 0.2–0.4, it is difficult to retrieve AOD which is essential to obtain ASSA and AAF because reflectance is virtually insensitive to AOD (Seidel et al., 2012). Hence the present method is limited to low and middle latitudes and dark surfaces due to the relatively large retrieved uncertainty in AOD and DREclear efficiency for high surface albedo.

ð4Þ

where ω is the ASSA, g the AAF, and x a cloud property (i.e., COD, ER, or ST minus CTT) in each grid cell. The coefficients a, b, c, and d are given in Table 1. Eq. (4) indicates that the calculation of ΔDRE requires a grid-mean value of ASSA, AAF, and the cloud properties (COD, ER, and ST minus CTT) from the CERES/MODIS. Among these variables, cloud properties are directly measured by the satellite instruments, but the monthly-mean bulk aerosol properties (ASSA and AAF) should be retrieved in this study. This can be done by using shortwave flux and AOD from CERES (Choi et al., 2009). In detail, DREclear efficiency (defined as aerosol DREclear per AOD) at both TOA and the surface as a function of ASSA and AAF is used to retrieve a pair (ASSA, AAF). The assumptions and conditions for the ASSA and AAF retrieval follow Choi et al. (2009). Since the ASSA and AAF values were derived from CERES observations, they should be physically consistent with the DREs. In comparison with ASSA and AAF from the AERONET ground observation, the uncertainties in ASSA and AAF are 70.03 and 70.05, respectively (Choi et al., 2009). The effects of these uncertainties on our estimate of aerosol DRE will be discussed in Section 5. Note that high albedo

3.3. Sensitivity of ΔDRE to aerosol and cloud properties The sensitivity of ΔDRE to various aerosol and cloud properties was tested for quantitative examination of the cloud effects on aerosol DREall . Fig. 2 shows ΔDRE–COD relation for different effective radius (ER; 10 and 30 μm), the difference between surface temperature and cloud top temperature (ST minus CTT; 10 and 30 K) and surface albedo (0.1 and 0.3). Fig. 2a is for ASSA¼0.87 and AAF ¼0.52 (i.e., strong absorbing and weak forward scattering). ΔDRE for COD¼0.0 refers to a zero cloud effect (i.e., DREall ¼DREclear obs in Eq. (3)). With the exception of very thin clouds (CODo 2), ΔDRE increases with increasing COD (i.e., optically thicker clouds), and a positive value for ΔDRE indicates offsetting the cooling effect of aerosols (Liao and Seinfeld, 1998). The negative slope for CODo2 may be associated with solar radiation that is reflected or transmitted by thin clouds, and is scattered secondarily by aerosols. Particularly, more upward shortwave fluxes can pass through thin clouds for brighter surfaces at lower solar zenith angles, which induces negative ΔDRE. Moreover, relatively smaller ER or smaller ST minus CTT induce a larger ΔDRE for the same COD. It should be noted that the ΔDRE is mostly positive for low surface albedo (0.1; black lines), while negative for high surface albedo (0.3; gray lines). This is because the shortwave radiation reflected from a bright surface acts to strengthen the secondary scattering by aerosols. Conversely, Fig. 2b is for ASSA ¼ 0.99 and AAF ¼ 0.76 (i.e., weak absorbing and strong forward scattering). The figure also shows the positive correspondence between COD and ΔDRE for COD42. For larger COD ( 410), the positive slope is diminished because a COD of more than 10 refers to thick clouds that completely block incoming solar radiation. However, ΔDRE is nearly insensitive to changes in ER and ST minus CTT.

Table 1 Regression coefficients for estimating ΔDRE as a function of the cloud optical depth (COD), effective radius (ER), and surface temperature (ST) minus cloud top temperature (CTT) for four pairs of (ASSA, AAF). Each value for (ASSA, AAF) at 0.63 μm is (0.87, 0.52), (0.99, 0.52), (0.87, 0.76), and (0.99, 0.76). (ASSA, AAF)

x

a

b

c

d

−4

(0.87, 0.52)

COD ER ST−CTT

−5
10 −6
10−4 −2
10−4

−0.051 0.047 0.037

3.448 −1.272 −1.832

−30.62 13.09 18.37

(0.99, 0.52)

COD ER ST−CTT

−1
10−3 −4
10−4 −5
10−6

0.012 0.034 0.000

1.944 −0.906 −0.022

−37.85 −8.680 −15.56

(0.87, 0.76)

COD ER ST−CTT

−5
10−4 −4
10−4 −3
10−4

−0.026 0.032 0.039

2.227 −0.897 −1.910

−12.62 17.42 26.28

(0.99, 0.76)

COD ER ST−CTT

−1
10−3 −2
10−4 −2
10−5

0.046 0.017 0.002

0.449 −0.472 −0.115

−16.35 −3.421 −6.117

Simulated ΔDRE from RTM 30

30

20

ER=10 µm, ST-CTT=10 K ER=30 µm, ST-CTT=10 K ER=10 µm, ST-CTT=30 K ER=30 µm, ST-CTT=30 K

20

[W m-2]

Albedo =0.1 10

10

0

0

Albedo =0.1

Albedo =0.3 -10

Albedo =0.3

-10 0

5

10 COD

15

20

0

5

10 COD

15

20

Fig. 2. Simulated ΔDRE against COD for various ER, ST minus CTT and surface albedo for (a) ASSA¼0.87 and AAF ¼0.52 and for (b) ASSA¼ 0.99 and AAF ¼0.76. Black and grey lines denote surface albedos of 0.1 and 0.3, respectively.

H.-R. Oh et al. / Journal of Atmospheric and Solar-Terrestrial Physics 102 (2013) 311–320

8

315

20 ASSA=0.96, AAF=0.70 Profile 4

15

[km]

Profile 3 Land

4

Profile 2

2

cloud

Profile 1

[W m-2]

6

10 5 Profile1 Profile2 Profile3 Profile4

0

Ocean

0 0.0

-5 0.2

0.4 0.6 Ratio

0.8

1.0

0

5

10 COD

15

20

Fig. 3. (a) Four aerosol vertical profiles and cloud layer at 2 km altitude (blue area) as the conditions of RTM experiments and (b) the simulated ΔDRE against COD for the different profiles. The ratio in x-axis indicates the value for the number of aerosol in each layer divided by the maximum number of aerosol in a layer among the all layers. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

ΔDRE from CERES and RTM 10 0 [W m-2]

Fig. 3a shows various aerosol vertical profiles (lines) and the cloud layer at 2 km (blue shading). Profile 1 (black solid line) is the globally-averaged aerosol profile from CALIOP, characterized by two large values in the low troposphere below 1 km (i.e., 0.1 km and 0.4 km); in this profile, most of aerosols are below the cloud. Separating a global ocean (dark gray line with open square) from global continent (gray line with open circle), we see their difference (gray shaded area). The aerosol maximum is at 0.1 km and 1 km over the ocean and the land, respectively. In addition, to understand sensitivity of ΔDRE to aerosol vertical profiles, we designated three other artificial aerosol profiles such that the aerosol ratios (defined as a ratio of the number of aerosol pixels at a level to the maximum over all levels) at 0.1–1.0 km in profile 1 are replaced at 2, 4, and 6 km. Hence, profile 2 (black dashed line) is of aerosols within the cloud, and profiles 3 (black dotted line) and 4 (black dashed–dotted line) are of aerosols above the cloud. With the different aerosol profiles, the ΔDRE–COD relation was investigated for ASSA¼0.96 and AAF ¼0.70 (Fig. 3b). Similar to the results in Fig. 2, a positive slope is shown in all of the cases for COD4 2. The slope is larger when aerosols exist at higher altitudes (profile 3 4profile 2 4profile 1). However, there is no difference between profiles 3 and 4, implying that ΔDRE is not subject to the altitude of aerosol located above the cloud. While not shown in the figure, the same is true for aerosol below the cloud. Simply put, the detailed aerosol profile is not so crucial in determining ΔDRE. The ΔDRE–COD relation purely simulated by the RTM in Fig. 3 needs to be compared with the observations (Fig. 4). Table 2 provides the geographic coordinates of the domain for the comparison and the CERES aerosol optical properties averaged over that domain. Since ΔDREs cannot be obtained purely by observations, one needs to have aid of RTM simulation using the values in Table 2. In detail, TOA shortwave fluxes for all-sky scenes in each COD bin were first obtained from CERES. Then from the TOA shortwave fluxes, the ΔDRE was derived by removing RTMsimulated DREclear ; we refer it to as the observed ΔDRE hereafter. Fig. 4 shows that the observed (box plots) and simulated (black dots) ΔDREs are moderately similar. The differences between the observed and simulated ΔDREs also have to be addressed here. Compared with the median values of the observed ΔDRE, the simulation appears to largely underestimate the ΔDRE for a relatively small COD over all regions. This underestimation may be attributed to inaccurate detection of cirrus clouds. Moreover, the observed ΔDREs have a negatively skewed wide range for all regions. This is because the observed ΔDRE is based on regionally

-10 -20 90%

-30

75% Median

CERES CERES RTMRTM

-40 0

5

10 COD

25% 10%

15

20

Fig. 4. ΔDRE as a function of global COD from CERES observation (box plots) and RT model (solid circles). The box plots sum up the distribution, median, and variability of ΔDRE.

Table 2 Regional-averaged values of the CERES observed and estimated input parameters to simulate ΔDRE. Longitude (deg.)

Globe 01–3601

Latitude (deg.)

601S–601N

CERES observed parameters

Estimated parameters

Solar zenith Surface angle (deg.) albedo

ASSA AAF

43.51

0.96

0.10

0.70

averaged bulk values in Table 2, and moreover is influenced by other factors such as aerosol semi- and indirect effects.

4. Expanding a grid-point DRE to the globe 4.1. Spatial distribution of aerosol and cloud properties To understand the mechanisms by which aerosol and cloud properties determine ΔDRE in different regions, the global distributions of these variables should be examined. The estimated

316

H.-R. Oh et al. / Journal of Atmospheric and Solar-Terrestrial Physics 102 (2013) 311–320

Fig. 5c displays the 5-year mean of AOD from the CERES. In contrast to the distributions of ASSA and AAF, AOD is relatively larger over the land and coastal regions than over the ocean. Particularly, AODs are higher than 0.4 over India, East Asia, and the West coast of Africa, reflecting significant influence of natural and anthropogenic aerosols. Furthermore, the larger AOD, the larger ΔDRE; recall that ΔDRE calculated for fixed AOD (i.e., 0.2) in the look-up table is scaled by the observed grid mean value of AOD. Along with ASSA, AAF, and AOD, cloud properties such as CF (a), COD (b), ER (c), and CTT (d) from the MODIS is necessary for ΔDRE calculation. The 5-year mean of the cloud properties are shown in Fig. 6. Overall, CF and COD exhibit a similar spatial pattern (Fig. 6a and b), and large values are found over East Asia and the midlatitude ocean. A relatively smaller ER is distributed over the land and neighboring coastal regions downwind of land masses than over the ocean (Fig. 6c). Such a land–ocean contrast in ER is attributed to differences in dynamics of cloud formation and in abundance of aerosols that can act as CCN (Kawamoto et al., 2001). The CTT distribution shows that more lower-level clouds exist over the ocean than over land (Fig. 6d). CTTs are lower than 260 K over the western Pacific where deep convection frequently occurs. 4.2. Aerosol direct radiative effect for all-sky

Fig. 5. Spatial distributions of the estimated, (a) ASSA and (b) AAF averaged over the entire data period of May 2000 to December 2005 (c) AOD.

annual averages of monthly ASSA, AAF, and AOD over the globe are shown in Fig. 5. The global mean 7standard deviations of ASSA and AAF are 0.96 70.03 and 0.7070.08, respectively. Regionally, ASSA and AAF are relatively smaller over land and coastal regions than over ocean (Fig. 5a and b). These results suggest that land and coastal sea areas can have large ΔDRE due to more absorbing and backward scattering aerosols if all other factors are equal according to Fig. 2. The present estimates of ASSA and AAF were compared with those by Kim and Ramanathan (2008), who operate the Georgia Tech/Goddard Global Ozone Chemistry Aerosol Radiation and Transport (GOCART) model and assimilation with the AERONET data. The overall pattern of the two results is similar, though the detailed regional magnitudes differ (Fig. 5a and b versus Figs. A3a and A3b in Kim and Ramanathan, 2008). Our ASSA and AAF estimates over the ocean are higher by 2.1–9.0% (ASSA¼0.95– 0.99, AAF ¼0.57–0.78) than the results obtained by Kim and Ramanathan (2008), while those over mid-latitude land areas are lower by 2.0–8.3% (ASSA¼0.87–0.95, AAF ¼0.57–0.78). These discrepancies may result from differences in season, methodology, and measurement uncertainties between the two studies. In this study, the estimated ASSA and AAF are the 5-year mean, while the results from Kim and Ramanathan (2008) are the mean of March to May when a large amount of dust and smoke aerosols arise in the Northern Hemisphere. Methodologically, our estimation was radiatively determined by the observed and RTM-simulated SW fluxes, whereas their results were based on the chemical transport model.

In this section, our final estimates of DREall are discussed. Fig. 7 depicts the spatial patterns of the averages of DREall (a), DREclear (b), and their differences (c). The area-averaged DREall value 7 standard deviation is −3.17 1.0 W m−2 over the globe, indicating a net aerosol radiative effect of cooling. Our result is in a very good agreement with the value from the model-based estimate for 901S–901N by Kim and Ramanathan (2008) (−3.0 71.0 W m−2) and that from a combined approach from CERES/MODIS and GOCART by Zhao et al. (2011) (−2.9 70.7 W m−2). Therefore, the aerosol cooling effect is dominant over most regions except for the equatorial ocean of 1201E–1801 and extra-tropical oceans (Fig. 7a). DREclear from the CERES was found to be below −5.0 W m−2 in most regions (Fig. 7b). The area-averaged aerosol DREclear values 7standard deviation were −7.3 71.6 W m−2 over the globe. This result is slightly lower than that of Zhao et al. (2011), who reported that DREclear is −6.8 7 1.7 W m−2 from the combined approach of CERES/MODIS and GOCART models. The difference between the two results is inevitable, and may have resulted from the estimating flux for pristine conditions; we used a flux for pristine conditions in each grid from the CERES/FSW, while Zhao et al. (2011) calculated the value by extrapolation of the flux for AOD¼0. The stronger aerosol cooling effect over the continents and neighboring coastal areas in the northern hemisphere reflects massive aerosol loading and its transport. The difference between DREall and DREclear (Fig. 7c) shows a positive sign over most regions (4.2 7 1.4 W m−2 on the global average), which indicates that clouds generally contribute to a weakening of the aerosol cooling effect. The negative values for some land areas may be attributed to a relatively high surface albedo. As indicated in Fig. 6b and d, these regions contain relatively thin (CODo10) and high clouds (CTT o260 K), leading to negative ΔDRE (refer to short dashed or dotted gray lines in Fig. 2). This result indicates that more shortwave radiation is reflected over brighter surfaces, which reinforces scattering by aerosols and their cooling effect. On the contrary, for dark surfaces such as the ocean, cloud effects compensate for a negative DREclear . Particularly over East Asia and the ocean at latitudes higher than 301 in both hemispheres, the differences between DREall and DREclear are very large (4 6 W m−2), primarily because of the combined influence of large CF and COD (refer to Fig. 6a and b). From Eq. (3), it is clear that DREall can increase with an increase in either f (i.e., CF) or ΔDRE. For the west coast of South Africa and

H.-R. Oh et al. / Journal of Atmospheric and Solar-Terrestrial Physics 102 (2013) 311–320

317

Fig. 6. MODIS-observed (a) cloud fraction, (b) cloud optical depth COD, (c) cloud effective radius ER (μm), and (d) cloud top temperature CTT (K) averaged over the entire data period.

a

c

d

e

b

f

g

h i

Fig. 7. Spatial distributions of (a) DREall , (b) DREclear , and (c) DREall minus DREclear averaged over the entire data period. DREall could not be derived over bright surface because ASSA and AAF were not available (white area).

South America, large DRE differences (4 9 W m−2) occur despite relatively low COD ( o10). These results from low-level clouds with large CF (Fig. 6a and d), indicating that aerosols are often located above the clouds. Finally, we further examined monthly variations in the DREs in the selected nine regions indicated by the boxes in Fig. 7c. Indeed, these monthly variations provide important implications on

regional aerosol–cloud radiative interactions. Overall, DREall (○) is larger than DREclear (●) (Fig. 8). For Mediterranean Basin, India, East Asia, and the west coast of North Africa, both DREclear and DREall show relatively large monthly variations (Fig. 8a–c, and f). For the Mediterranean Basin, the difference between DREclear and DREall is relatively large during December–January–February (DJF) and relatively small during May–June–July–August (MJJA); however, the opposite is true for India and the west coast of North Africa. Meanwhile, for East Asia, the DRE difference is larger than 10 W m−2 throughout the year, indicating a persistent influence of clouds on DREall . These DRE differences follow the monthly variation in the corresponding COD and CF (Bergamo et al., 2008; Dey and Tripathi, 2008). On the contrary, for the remaining regions (Fig. 8d, e, g–i) the seasonal variation in DREclear is relatively small. Among these regions, for West Coast of the United States, the differences between the two DREs remain nearly unchanged. For the U.S. East Coast and Indonesia, the large DRE differences during DJF are mainly due to an increase in CF. Particularly during the winter season (e.g., July and August) of South America, negatively strong DREclear occurred owing to an increase in biomass-burning aerosols (Torres et al., 2010). Since smoke aerosols by biomass burning can ascend to high-level altitudes, these aerosols can be located over prevailing low-level clouds, which may result in relatively weaker DREall during the winter season. Over the Southern Hemispheric ocean, the monthly variation in DREall is larger than that of DREclear owing to storms in the Southern Hemisphere, implying that clouds can modify DREclear in this region. Therefore, the monthly variation in DREall in each region is highly affected by CF and COD.

5. Uncertainty in aerosol direct radiative effect in all-sky scenes Though our global estimates of aerosol all-sky DRE (DREall ) are generally in good agreement with those from previous studies based on different methods (e.g., Kim and Ramanathan, 2008; Zhao et al., 2011), several sources of uncertainty associated with our estimation should be discussed. We summarized the uncertainty sources in Table 3. These values are based on the change in

318

H.-R. Oh et al. / Journal of Atmospheric and Solar-Terrestrial Physics 102 (2013) 311–320

Aerosol DRE for clear-sky and all-sky 10

10

0

0

-10

-10

-10

-20

-20

-20

-30 J F MA MJ J A S O N D

-30 J F MA MJ J A S O N D

-30 J F MA MJ J A S O N D

10 DRE all

[W m-2]

[W m-2]

[W m-2]

0

DRE clear

10

10

10

0

0

0

-10

-10

-10

-20

-20

-20

-30 J F MA MJ J A S O N D

-30 J F MA MJ J A S O N D

-30 J F MA MJ J A S O N D

10

10

10

0

0

0

-10

-10

-10

-20

-20

-20

-30 J F MA MJ J A S O N D month

-30 J F MA MJ J A S O N D month

-30 J F MA MJ J A S O N D month

Fig. 8. Monthly variations in DREall (open circles) and DREclear (filled circles) over the nine regions indicated by the box in Fig. 7c.

globally averaged DREall (%) due to an uncertainty of each factor in all grid points. The sources of uncertainty are roughly classified into three categories: RTM, MODIS/CERES, and the assumed aerosol vertical profile. The first source of uncertainty is the difference in shortwave fluxes from between RTM and CERES. RTM used in the study has relatively low absorptivity, which results in underestimation of DREclear by approximately 10% (Oh et al., 2013). The bias leads to uncertainty in DREall of 7.6%. The second source of uncertainty is retrieval errors of the various products related to the aerosol and cloud. Remer et al. (2005) expected that the retrieval uncertainty of AOD over land and ocean is 0.05 70.15 AOD and 0.03 70.05 AOD; this amount can cause approximately 22.9% and 30.9% of uncertainty in DREall , respectively. Consequently, 30.5% of uncertainty in DREall can be occurred. Indeed, AOD is the largest influencing factor among several error sources. Owing to the measurement errors by a conversion from CERES radiances to fluxes and by high surface

albedo in high latitudes, ASSA and AAF have retrieval uncertainties of 70.03 and 70.05, respectively. The associated uncertainty in DREall is 11.5% and 24.2% due to ASSA and AAF, respectively. A maximum retrieval uncertainty of 15% appears in error sources related to various cloud properties, which can affect DREall by up to approximately 18%. As another influencing factor, surface albedo has 10% of retrieval uncertainty (Jin et al., 2003), which may cause a 4.1% change in DREall . The third source of uncertainty is associated with the aerosol and cloud altitudes. According to Fig. 3, whether aerosols are above clouds or not is a key factor in determining DREall . Therefore, we further examined the frequency of aerosol occurrence above clouds that have a relatively large influence on ΔDRE from CALIOP VFM data during 2007–2010. The frequency was determined by comparing maximum heights of aerosols and clouds from their vertical profiles. Table 4 shows the percentage of six aerosol types (A), and the frequency for aerosols above clouds (B). The resulting total frequency (%) can be calculated by

H.-R. Oh et al. / Journal of Atmospheric and Solar-Terrestrial Physics 102 (2013) 311–320

319

Table 3 Error sources in DRE for all-sky and its uncertainty. Source of uncertainty Model error source Retrieval error source

Relative altitude of aerosol and cloud n

Extent Difference in shortwave fluxes from between RTM and CERES MODIS AOD Aerosol single scattering albedo Aerosol asymmetry factor MODIS cloud mask MODIS cloud optical depth MODIS cloud effective radius MODIS cloud top temperature Surface albedo Aerosol above cloud

10% of DRE

Reference clear

Uncertainty in DREall (%)

Oh et al. (2013)

0.05 7 0.15τ in land 0.03 7 0.05τ in ocean 70.03 70.05 15% 8% 15% 1K 10% 20.33% of the total case

7.6

Remer et al. (2005)

30.5, 22.9n, 30.9nn

Choi et al. (2009) Choi et al. (2009) Ackerman et al. (2008) Dong et al. (2008) Dong et al. (2008) Dong et al. (2008) Jin et al. (2003) Table 4 in this study

11.5 24.2 17.9 8.4 0.03 0.9 4.1 15.7

This value is for land. This value is for ocean.

nn

multiplication (i.e., A
B) of all aerosol contributions, yielding approximately 20% over the globe. Among the total frequency, aerosols classified as dust (i.e., polluted dust and dust) occupied more than half. The absolute frequency of aerosol occurrence above clouds was generally highest (31.3%) for clean continental aerosols; however, its influence is insignificant owing to the smallest ratio (9.3%) to the total aerosol. Allowing for the total frequency (20.3%) for aerosols above clouds, DREall can be changed by 15.7% (last line in Table 3).

6. Conclusions and discussions By considering the influence of clouds on aerosol DRE, this study has estimated aerosol DRE for both cloudy and cloud-free scenes over 601S–601N. In this study we attempted to circumvent conventional limitation of the aerosol DRE estimations for cloudy scenes based on satellite observations which are confined to cloud-free scenes. Our method uses shortwave flux, surface albedo, aerosol from CERES, and cloud properties from MODIS. Our method using monthly satellite observations and RTM is a fast and cost-efficient way to determine monthly varying aerosol radiative effects, and is distinguished from previous methods using chemical transport models. The results provided the consistent evidence that clouds can suppress or enhance direct radiative cooling by aerosols, which also depends on aerosol optical properties and surface albedo. On the basis of our calculation, the average7standard deviation of DREall is −3.1 71.0 W m−2 over the globe, which is approximately one third of the DRE for cloud-free scenes (−7.37 1.6 W m−2). The value of DREall minus DREclear is positive for all oceanic regions but can be negative for land regions with high surface albedos. Particularly for 301S–601S, the aerosol radiative cooling was largely canceled out. In addition, the monthly variation in DREall in this region differed significantly from that of DREclear , implying that cloud variation can alter the time variation in DRE. Because of the observational limitations in satellite remote sensing, modeling studies have been performed by using climate and chemical models to globally estimate DREall , which improved our understanding of the influence of clouds on the DRE. Nevertheless, imperfect parameterizations in these models induce a large uncertainty in quantification of the all-sky DRE, and its nearreal-time estimation is currently impossible owing to the heavy computing demand. However, this study has straightforwardly estimated all-sky DRE by using satellite observations to the fullest capacity with the aid of the pre-calculated look-up table from RTM. This methodology is more likely to be a satellite remote

Table 4 Frequency (%) for aerosols above clouds from the CALIOP VFM data observed over 4 years (2007–10). Frequency for aerosols above clouds (%), B

Total frequency (%), A
B

Aerosol type

Percentage (%), A

Marine aerosol Polluted continental Clean continental Polluted dust Dust Smoke

22.27 12.50

6.23 7.91

1.38 0.98

9.31 26.91 16.26 12.71

31.39 24.67 28.43 29.67

2.92 6.63 4.62 3.77

Total

99.96

20.40

20.33

sensing technique commonly used to retrieve level-2 products. The key for this methodology was the observational extraction of aerosol optical properties, ASSA and AAF, which enabled calculation of a reliable DRE for all-sky scenes. On the global average all-sky aerosol DRE values are −2.0 W m−2 in Loeb and Manalo-Smith (2005), −3.0 W m−2 in Kim and Ramanathan (2008), −2.9 W m−2 in Zhao et al. (2011), −1.0 W m−2 in Chen et al. (2011) (for anthropogenic aerosol only), and −3.1 W m−2 in the present study. These estimates were obtained by different methods, but are fairly comparable with our result. All is lower than clear-sky DREs (about −5.0 W m−2 in the global average) summarized in IPCC AR4 (2007), consistently supporting that aerosol DREs generally weakened once cloud effects were considered. However, caution must be exercised in interpreting the results because of various error sources. For this reason, further studies should be performed to reduce a retrieval uncertainty with more reliable observations.

Acknowledgements This study was supported by Korea Ministry of Environment as ``Climate Change Correspondence R&D Program''. Choi, Y.-S. was supported by the Korean Ministry of Environment as part of the Eco-Innovation Project. References Ackerman, S.A., Holz, R.E., Frey, R., Eloranta, E.W., Maddux, B.C., McGill, M., 2008. Cloud detection with MODIS. Part II: validation. Journal of Atmospheric and Oceanic Technology 25, 1073–1086.

320

H.-R. Oh et al. / Journal of Atmospheric and Solar-Terrestrial Physics 102 (2013) 311–320

Andrews, E., et al., 2006. Comparison of methods for deriving aerosol asymmetry parameter. Journal of Geophysical Research 111, D05S04, http://dx.doi.org/ 10.1029/2004JD005734. Bates, T.S., et al., 2006. Aerosol direct radiative effects over the northwest Atlantic, northwest Pacific, and North Indian Oceans: estimates based on in-situ chemical and optical measurements and chemical transport modeling. Atmospheric Chemistry and Physics 6, 1657–1732. Bellouin, N., Boucher, O., Haywood, J., Reddy, M.S., 2005. Global estimate of aerosol direct radiative forcing from satellite measurements. Nature 438, 1138–1141. Bergamo, A., Tafuro, A.M., Kinne, S., De Tomasi, F., Perrone, M.R., 2008. Monthlyaveraged anthropogenic aerosol direct radiative forcing over the Mediterranean based on AERONET aerosol properties. Atmospheric Chemistry and Physics 8, 6995–7014. Chand, D., Wood, R., Anderson, T.L., Satheesh, S.K., Charlson, R.J., 2009. Satellitederived direct radiative effect of aerosols dependent on cloud cover. Nature Geoscience 2, 181–184. Charlock, T.P., Rose, F.G., Rutan, D.A., Coleman, L.H., Caldwell, T., Zentz, S., 2005. Global, multi-year analysis of clouds and Earth's radiant energy system Terra observations and radiative transfer calculations. 15th ARM science team meeting, Daytona Beach, Florida. Available from: 〈http://www-cave.larc.nasa.gov/ cave/pdfs/Charlock.15ARM05.pdf〉. Chen, L., Shi, G.Y., Qin, S.G., Yang, S., Zhang, P., 2011. Direct radiative forcing of anthropogenic aerosols over oceans from satellite observations. Advances in Atmospheric Sciences 28, 973–984, http://dx.doi.org/10.1007/s00376-0109210-4. Chin, M., Ginoux, P., Kinne, S., Torres, O., Holben, B.N., Duncan, B.N., Martin, R.V., Logan, J.A., Higurashi, A., Nakajima, T., 2002. Tropospheric aerosol optical thickness from the GOCART model and comparisons with satellite and sunphotometer measurements. Journal of the Atmospheric Sciences 59, 461–483. Choi, Y.S., Ho, C.H., 2009. Validation of the cloud property retrievals from the MTSAT-1R imagery using MODIS observations. International Journal of Remote Sensing 30, 5935–5958. Choi, Y.S., Ho, C.H., Oh, H.R., Park, R.J., Song, C.G., 2009. Estimating bulk optical properties of aerosols over the western North Pacific by using MODIS and CERES measurements. Atmospheric Environment 43, 5654–5660. Chung, C.E., 2012. Aerosol direct radiative forcing: a review. Atmospheric aerosols— regional characteristics—chemistry and physics. In: Hayder, A.-R. et al., (Eds.), InTech. Available from: 〈http://www.intechopen.com/books/atmospheric-aero sols-direct-radiative-forcing-characteristics-chemistry-and-physics/ aerosol-direct-radiavtive-forcing-a-revies〉. Devasthale, A., Thomas, M.A., 2011. A global survey of aerosol–liquid water cloud overlap based on four years of CALIPSO–CALIOP data. Atmospheric Chemistry and Physics 11, 1143–1154. Dey, S., Tripathi, S.N., 2008. Aerosol direct radiative effects over Kanpur in the IndoGangetic basin, northern India: long-term (2001–2005) observations and implications to regional climate. Journal of Geophysical Research 113, D04212, http://dx.doi.org/10.1029/2007JD009029. Dong, X., Minnis, P., Xi, B., Sun-Mack, S., Chen, Y., 2008. Comparison of CERES– MODIS stratus cloud properties with ground-based measurements at the DOE ARM Southern Great Plains site. Journal of Geophysical Research 113, D03204, http://dx.doi.org/10.1029/2007JD008438. Fu, Q., Liou, K.N., 1993. Parameterization of the radiative properties of cirrus clouds. Journal of the Atmospheric Sciences 50, 2008–2025. Haywood, J.M., Ramaswamy, V., Soden, B.J., 1999. Tropospheric aerosol climate forcing in clear-sky satellite observations over the oceans. Science 283, 1299–1303. Holz, R.E., Ackerman, S.A., Nagle, F.W., Frey, R., Dutcher, S., Kuehn, R.E., Vaughan, M. A., Baum, B., 2008. Global Moderate Resolution Imaging Spectroradiometer (MODIS) cloud detection and height evaluation using CALIOP. Journal of Geophysical Research 113, D00A19, http://dx.doi.org/10.1029/2008JD009837. Hu, Y., Winker, D., Vaughan, M., Lin, B., Omar, A., Trepte, C., Flittner, D., Yang, P., Nasiri, S.L., Baum, B., Holz, R., Sun, W., Liu, Z., Wang, Z., Young, S., Stamnes, K., Huang, J., Kuehn, R., 2009. CALIPSO/CALIOP cloud phase discrimination algorithm. Journal of Atmospheric and Oceanic Technology 26, 2293–2309. Intergovernmental Panel on Climate Change, 2007. Climate Change 2007: The Physical Science Basis. Contribution of Working Group Ⅰ to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. In: Solomon, S. et al., (Eds.), Cambridge University Press, Cambridge, U.K. Jin, Y., Schaaf, C.B., Woodcock, C.E., Gao, F., Li, X., Strahler, A.H., Lucht, W., Liang, S., 2003. Consistency of MODIS surface bidirectional reflectance distribution

function and albedo retrievals: 2. Validation. Journal of Geophysical Research 108, 4159, http://dx.doi.org/10.1029/2002JD002804. Kaufman, Y.J., Tanre, D., Boucher, O., 2002. A satellite view of aerosols in the climate system. Nature 419, 215–223. Kawamoto, K., Nakajima, T., Nakajima, T.Y., 2001. A global determination of cloud microphysics with AVHRR remote sensing. Journal of Climate 14, 2054–2068. Keil, A., Haywood, J.M., 2003. Solar radiative forcing by biomass burning aerosol particles during SAFARI 2000: a case study based on measured aerosol and cloud properties. Journal of Geophysical Research 108, 8467, http://dx.doi.org/ 10.1029/2002JD002315. Kim, D.Y., Ramanathan, V., 2008. Solar radiation budget and radiative forcing due to aerosols and clouds. Journal of Geophysical Research 113, D02203, http://dx. doi.org/10.1029/2007JD008434. King, M.D., Tsay, S.C., Plantnick, S.E., Wang, M., Liou, K.N., 1997. Cloud retrieval algorithms for MODIS, MODIS algorithm theoretical basis document. NASA. Liao, H., Seinfeld, J.H., 1998. Effect of clouds on direct aerosol radiative forcing of climate. Journal of Geophysical Research 103, 3781–3788. Loeb, N.G., Manalo-Smith, N., 2005. Top-of-atmosphere direct radiative effect of aerosols over global oceans from merged CERES and MODIS observations. Journal of Climate 18, 3506–3526. Menzel, W.P., Smith, W.L., Stewart, T.R., 1983. Improved cloud motion wind vector and altitude assignment using vas. Journal of Climate and Applied Meteorology 22, 377–384. Myhre, G., 2009. Consistency between satellite-derived and modeled estimates of the direct aerosol effect. Science 325, 187–190. Oh, H.R., Ho, C.H., Choi, Y.S., 2013. Comments on “Direct radiative forcing of anthropogenic aerosols over oceans from satellite observation”. Advances in Atmospheric Sciences 30, 10–14. Omar, A.H., Winker, D.M., Kittaka, C., Vaughan, M.A., Liu, Z., Hu, Y., Trepte, C.R., Rogers, R.R., Ferrare, R.A., Lee, K.-P., Kuehn, R.E., Hostetler, C.A., 2009. The CALIPSO automated aerosol classification and lidar ratio selection algorithm. Journal of Atmospheric and Oceanic Technology 26, 1994–2014. Patadia, F., Gupta, P., Christopher, S.A., 2008. First observational estimates of global clear sky shortwave aerosol direct radiative effect over land. Geophysical Research Letters 35, L04810, http://dx.doi.org/10.1029/2007GL032314. Platnick, S., King, M.D., Ackerman, S.A., Menzel, W.P., Baum, B.A., Riedi, J.C., Frey, R. A., 2003. The MODIS cloud products: algorithms and examples from Terra. IEEE Transactions on Geoscience and Remote Sensing 41, 459–473. Podgorny, I.A., Ramanathan, V., 2001. A modeling study of the direct effect of aerosols over the tropical Indian Ocean. Journal of Geophysical Research 106 24,097–24,105. Ramanathan, V., Crutzen, P.J., Kiehl, J.T., Rosenfeld, D., 2001. Atmosphere–aerosols, climate, and the hydrological cycle. Science 294, 2119–2124. Remer, L.A., et al., 2005. The MODIS aerosol algorithm, products and validation. Journal of the Atmospheric Sciences 62, 947–973. Remer, L.A., Kaufman, Y.J., 2006. Aerosol direct radiative effect at the top of the atmosphere over cloud free ocean derived from four years of MODIS data. Atmospheric Chemistry and Physics 6, 237–253. Ricchiazzi, P., Yang, S.R., Gautier, C., Sowle, D., 1998. SBDART: a research and teaching software tool for plane-parallel radiative transfer in the Earth's atmosphere. Bulletin of the American Meteorological Society 79, 2101–2114. Rossow, W.B., Walker, A.W., Garder, L.C., 1993. Comparison of ISCCP and other cloud amounts. Journal of Climate 6, 2394–2418. Rutan, D., Rose, F., Roman, M., Manalo-Smith, N., Schaaf, C., Charlock, T., 2009. Development and assessment of broadband surface albedo from clouds and the Earth's radiant energy system clouds and radiation swath data product. Journal of Geophysical Research 114, D08125, http://dx.doi.org/10.1029/2008jd010669. Seidel, F.C., Kokhanovsky, A.A., Schaepman, M.E., 2012. Fast retrieval of aerosol optical depth and its sensitivity to surface albedo using remote sensing data. Atmospheric Research 116, 22–32. Torres, O., Chen, Z., Jethva, H., Ahn, C., Freitas, S.R., Bhartia, P.K., 2010. OMI and MODIS observations of the anomalous 2008–2009 Southern Hemisphere biomass burning seasons. Atmospheric Chemistry and Physics 10, 3505–3513. Yu, H., et al., 2006. A review of measurement-based assessments of the aerosol direct radiative effect and forcing. Atmospheric Chemistry and Physics 6, 613–666. Winker, D.M., Hunt, W.H., McGill, M.J., 2007. Initial performance assessment of CALIOP. Geophysical Research Letters 34, L19803, http://dx.doi.org/10.1029/ 2007GL030135. Zhao, T.X.-P., Loeb, N.G., Laszlo, I., Zhou, M., 2011. Global component aerosol direct radiative effect at the top of atmosphere. International Journal of Remote Sensing 32, 633–655.