Shortwave versus longwave direct radiative forcing by Taklimakan dust aerosols



[1] Six Mays from 2001 to 2006 of aerosol optical depth from the Multiangle Imaging SpectroRadiometer and short- and longwave flux from the Clouds and Earth's Radiant Energy Budget Scanner are combined to estimate radiative forcing of dust aerosols in the Taklimakan Desert (75°E–95°E, 36°N–42°N, elevation < 1600 m). The cloud-free dust shortwave versus longwave forcing per aerosol optical depth at about 05:00 UTC are −48.1 and 28.4 W m−2, respectively. Dust longwave warming offsets 58% of dust shortwave cooling and the overall dust radiative effect is to cool the Earth system. Annual shortwave and longwave forcing efficiencies vary from 26.7 to 63.8 and 18.3 to 39.3 W m−2, respectively, due to changes in surface properties. Radiative transfer model simulations also suggest Earth's system is cooled in the shortwave but warmed in the longwave by Taklimakan dust aerosols.

1. Introduction

[2] Aerosols perturb the energy budget of Earth's system by directly scattering and absorbing shortwave (SW) and longwave (LW) radiation. Among all of the natural and anthropogenic components of aerosols, desert dust is one of the most important and it plays a significant role in climate forcing [Haywood and Boucher, 2000]. Dust aerosols reflect SW radiation back to the top of the atmosphere (TOA) and thereby cool Earth's system. Dust aerosols absorb LW radiation in the atmospheric window channel and thereby warm Earth's system. Such a warming effect could be significant, as suggested by a few observation-based estimates and radiative transfer model simulations [Yu et al., 2006].

[3] Using broadband measurements from the Earth Radiation Budget Experiment (ERBE), a significant impact of dust aerosols on SW radiation is observed over the Atlantic and Arabian Oceans. However, over the Sahara and Sahel regions, satellite observations of SW radiation in clear regions were close to those in dust-laden regions [Ackerman and Chung, 1992; Hsu et al., 2000]. This is likely due to the relatively high surface albedo and large heterogeneity of the surface reflectance [Yu et al., 2006]. The LW effect of Saharan dust over the Atlantic Ocean accounts for 10% to 30% of the instantaneous SW effect [Highwood et al., 2003; Hsu et al., 2000]. The discrepancy between model simulations and satellite observations suggests that instantaneous mineral dust LW forcing in July can be as much as 50 W m−2 for 12:00 UTC in the cloud-free regions [Haywood et al., 2005]. Sahara dust LW forcing at the TOA in the six regions ranges from 1 to 21 W m−2 and the mean value is 15 W m−2 [Zhang and Christopher, 2003]. A mixture of dust, sea-salt, and pollution in the northern Indian Ocean can impose a LW warming that could reduce the SW effect by about 45% to 70% [Satheesh and Lubin, 2003]. Radiative transfer model simulations show that the aerosol LW forcing increases with the surface skin temperature and the aerosol layer height and decreases with water vapor [Liao and Seinfeld, 1998]. This implies that information about these important factors should be available in order to derive high quality aerosol LW forcing. However, they are not well characterized by either observations or simulations [Yu et al., 2006].

[4] Direct observation of dust radiative effects requires simultaneous measurements of cloud free LW and SW flux and aerosol loading. High quality satellite observations of the aerosols and fluxes have been available since 2000. The data are generally used to study the aerosols' radiative effects over the ocean [Yu et al., 2006]. An observation-based estimation of SW aerosol forcing over land is firstly presented using merged satellite observations of aerosols, clouds, and radiation [Patadia et al., 2008].

[5] The radiative impacts of dust aerosols have previously been evaluated using radiative transfer models [Wang et al., 2006]. However, simulations have rarely been corroborated empirically, especially in the Taklimakan Desert. In this study, the radiative effect of the Taklimakan dust is estimated using six years of merged satellite data for the month of May. We extend the approach by Patadia et al. [2008] to study both the SW and LW effects of dust aerosols. Thereby, the overall effect of dust aerosols is determined. Satellite retrievals of aerosol and short- and longwave flux are combined to achieve this goal. Furthermore, a radiative transfer model is used to simulate the dust radiative effects to support the observation-based estimations.

2. Data and Methodology

[6] Two satellite data sets in May from 2001 to 2006 are used. The first is the Single Scanner Footprint TOA/surface Fluxes and clouds (SSF) product that contains one hour of instantaneous CERES data for a single scanner and merged information from a higher-resolution MODIS imager on Terra [Wielicki et al., 1996]. In this product, the higher resolution MODIS data, such as scene identification along with cloud and aerosol properties are averaged over the larger CERES footprint using point spread functions. The second is the MISR level 2 daily aerosol product that contains the aerosol optical depth with 17.6 km2 resolution for each orbit. Comparisons of MISR AODs with those of ground-based remote sensing showed good correlation with no obvious systematic biases or trends over desert sites [Martonchik et al., 2004]. The MISR AODs are highly correlated with ground-based remote sensing AODs and the root-mean-square-error of MISR AODs was 0.06 [Christopher and Wang, 2004]. The MISR AOD data are collocated spatially and temporally with the CERES SSF data as follows: all instantaneous MISR pixels that fall within the CERES footprint and correspond to the same observation time are located and the arithmetic average of the AOD is computed. Given the fact that the spatial resolution of the MISR AOD is close to that of CERES and the change of respective MISR pixel sizes is little due to the narrow width of the MISR, point spread function weighting of the CERES is not assigned in the collocation algorithm. However, the change in CERES footprint size is accounted for. To account for this, the CERES pixel size in the along- and across-track dimensions is calculated as a function of viewing geometry and is used to estimate CERES footprint size [Gupta et al., 2008]. The MODIS cloud cover information from the merged CERES SSF product is used to remove cloudy pixels. Those pixels with cloud fraction >0.5% and sub-pixel clear area percentages <99.9% are removed. In order to avoid using large pixel sizes at large scan angles, the analysis is restricted to CERES pixels with satellite viewing and solar zenith angles <60°.

[7] The Santa Barbara DISORT Atmospheric Radiative Transfer model (SBDART) is used to compute TOA SW and LW fluxes for various atmospheric and surface conditions. The dust optical properties in the SW band are from AERONET data at Dunhuang [Xia et al., 2005]. In the LW spectrum, the dust optical properties are calculated using in situ measurements of refractive index [Shi et al., 2005] and AERONET size distribution. The long-range transport and vertical distribution of Asian dust aerosols were directly observed by the two-wavelength Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) [Huang et al., 2008; Liu et al., 2008]. Here, the level 1B (Ver. 2.01) CALIOP total attenuated backscatter coefficients at 532 nm are used to retrieve aerosol extinction coefficient profiles based on the Fernald's inversion method [Fernald, 1984]. The Level 1B data are averaged to a horizontal resolution of 5 km before the inversion. The Level 2 Cloud-Aerosol Discrimination data are used to screen out cloud signals. The inversion is started from an altitude of 14 km down to the ground surface and the lidar ratio of 35 sr is assumed. The daily conventional atmospheric sounding at 6 radio sounding stations surrounding the Taklimakan Desert in May from 2001 to 2006 are averaged to obtain the typical temperature and moisture profiles. Given the fact that information about surface albedo and surface skin temperature is limited, the surface albedo and surface skin temperature are set to vary from 0.18 to 0.26 with 0.1 step and 310 to 330 k with 5 k step in the model simulations, respectively.

3. Results

[8] The monthly AOD at 550 nm ranges from 0.51 in 2005 to 0.62 in 2006 and the six-year mean value is 0.56. The aerosol extinction profiles retrieved from CALIOP data for 13 days in May are presented in Figure 1. The thick black line represents the mean profile used in the model simulation. The dust aerosol layer generally extends from the surface to a height of 4 to 6 km. This is supported by limited ground-based Lidar observations and balloon measurements. For example, a dense dust layer, as observed using a ground-based depolarization Lidar, developed over the Taklimakan Desert in April from the ground up to 5.5 km [Kai et al., 2008].

Figure 1.

CALIOP dust extinction coefficient profile derived from CALIOP Level 1B attenuated backscattering coefficient data for 13 days in May of 2007 and 2008. The thick curve represents the mean profile.

[9] Figure 2 shows the MISR AOD versus the CERES SW (Figure 2, left) and LW (Figure 2, right) flux for all data points in six months. The data points are binned by the values of MISR AOD taken at regular intervals (2.5%) from the cumulative distribution function. Percentiles near 100 represent very large dust loading and those near 2.5% represent very small loading. The short horizontal and vertical bars represent the standard deviation of AOD and flux, respectively. The dashed lines represent the SBDART simulations with varying surface albedos (SW) and surface skin temperatures (LW). As the MISR AOD increases, the CERES SW flux increases, but on the contrary, the CERES LW decreases. The positive relationship between AOD and SW flux suggests more dust aerosols, more SW radiation reflected back to space and thereby the dust SW effect is to cool Earth's system. New dust optical data suggest that the single scattering albedo of pure dust at a wavelength of 670 nm ranges from 0.90 and 0.99, with a central global estimate of 0.96 [Dubovik et al., 2002]. The value at Dunhuang is 0.97 [Xia et al., 2005]. The SBDART simulations with this value, as seen from Figure 2, can produce the observed positive correlation between aerosol and SW flux at the TOA. Note that dust SW forcing in the Saharan and the Arabian regions are very low [Ackerman and Chung, 1992; Hsu et al., 2000]. These are probably due to the higher surface albedo there as compared to the Taklimakan Desert. The broadband white-sky albedo (300–5000 nm) in May 2001 is larger than 0.4 in the Saharan and Arabian deserts; however, it is lower than 0.25 in the Taklimakan Desert [Schaaf et al., 2002]. The SBDART simulations show very small dust SW effect if the model is driven by the surface albedo of 0.4 (not shown). The negative relationship between the LW flux and AOD is because dust aerosols absorb LW radiation emitted by the surface and emit to the TOA at a colder temperature when compared to the surface. The SBDART, driven by the aforementioned aerosol optical properties, can also produce the negative correlation between AOD and LW flux. We derive the dust SW and LW forcing efficiencies from relationships between the MISR AOD and the CERES SW as well as LW flux, respectively. The dust radiative forcing efficiency (DRFE) is defined as the flux perturbations due to one unit of dust optical depth. The DRFE in the SW spectrum (DREFSW) and in the LW spectrum (DREFLW) is −48.1 W m−2 and 28.4 W m−2 per AOD, respectively. This suggests the overall dust radiative effect is still to cool Earth's system, although the dust LW warming can offset 58% of the SW cooling.

Figure 2.

The relationships between the MISR aerosol optical depth at 550 nm (the x-axis) and the CERES TOA (left) shortwave flux (y-axis) and (right) longwave flux. The data points are binned by the values of MISR AOD taken at regular intervals (2.5%) from the cumulative distribution function. The short horizontal and vertical bars represent the standard deviation of AOD and flux, respectively. The dashed lines represent the SBDART simulations with varying surface albedos (SW) and surface skin temperatures (LW), respectively.

[10] Figure 3 shows similar results as Figure 2 (left) but data points in May from 2001 to 2006 are studied separately. The fact that the CERES TOA SW flux increases as the MISR AOD increases appears every year. The linear fit of the relationship between the MISR AOD and the CERES SSF TOA SW flux is also presented in Figure 3. The results show a large range of DREFSW, which varies from about −27 W m−2 per AOD in 2004 to about −64 W m−2 per AOD in 2003. The DREFSW is significantly negatively correlated to the intercept of the linear equation. The intercept represents the expected SW flux at the TOA in the case without aerosols in the atmosphere, which is determined mainly by the surface albedo, i.e., higher surface albedo, more SW flux back into space. This suggests that substantial annual variations in the DREFSW are likely due to changes in the desert surface albedo.

Figure 3.

Same as that of Figure 1 but for the collocated data points of the MISR AOD and CERES SW flux every year from 2001 to 2006.

[11] The CERES TOA LW flux, as shown in Figure 4, generally tends to decrease as the MISR AOD increases each year. However, we can also see a slight increasing tendency as the MISR AOD increases from 0.15 to 0.55 in 2005 and in 2006, although we can derive an overall decreasing tendency of TOA LW as the AOD increases. The reason for this phenomenon is not clear because information about atmospheric and surface conditions besides dust loading is very limited. The DREFLW also varies from year to year. The largest DREFLW, ∼39 W m−2, occurs in 2004 when the expected TOA LW flux without dust is also the largest, 346 W m−2. The minimum DREFLW is ∼18 W m−2, which occurs in 2006 when TOA LW without dust is relatively lower. TOA LW is determined by the surface skin temperature under the same atmospheric conditions. More TOA LW without dust, to a large extent, means higher surface skin temperature. Therefore, the year-to-year change in DREFLW is, to some extent, associated with variation in surface skin temperature. However, this is still a guess, further research on this issue is required to fully understand impacts of surface property on dust forcing.

Figure 4.

Same as that of Figure 1 but for the collocated data points of the MISR AOD and CERES LW flux every year from 2001 to 2006.

[12] We should point out that, currently, all DREFs are instantaneous evaluations at about 05:00 UTC. The diurnal mean DREFSW varies significantly with solar geometry and surface albedo. DREFSW varies with zenith angle during daytime, and to zero during dark periods. If surface albedo and aerosol optical parameters remain constant, the SBDART simulations show that the diurnal mean DREF is about 80% of the instantaneous forcing for this region. Therefore, the diurnal mean DREFSW in the Taklimakan Desert is −38.5 W m−2. Given the fact that the mean May AOD is 0.56 in the Taklimakan Desert, the dust diurnal mean SW radiative forcing in May is −21.6 W m−2. The value is larger than that over global land (−4.9 W m−2) by a factor of 4 [Yu et al., 2006]. While the dust LW forcing is continuous, it varies instead with atmospheric and surface emissivity and thermal structure. The dust LW forcing during nighttime is expected to be less than that during daytime. This is first because the higher surface skin temperature during daytime leads to larger forcing efficiency. The second reason is that the stronger vertical motion of the atmosphere during daytime can transport dust to higher levels, thereby leading to larger forcing efficiency. Both factors would lead to a larger DREFLW during daytime than during nighttime, so one would expect the diurnal-mean DREFLW would be smaller than 28.4 W m−2. The overall radiative effect of Taklimakan dust is thus to cool Earth's system.

4. Discussion and Conclusions

[13] This is the first attempt to simultaneously analyze dust SW and LW forcing in the Taklimakan Desert. The results show substantial dust direct radiative forcing, not only in the LW, but also in the SW. The notable dust LW forcing is in agreement with those in other dust source regions, such as in the Sahara and Sahel regions [Hsu et al., 2000; Zhang and Christopher, 2003]. Dust scattering leads to an increase of TOA SW flux by 21 W m−2 in the Taklimakan Desert. However, the dust SW forcing at the TOA in the African and Arabian Desert is very small [Ackerman and Chung, 1992]. A potential reason for this difference is likely due to the relatively lower surface albedo in the Taklimakan Desert. The satellite observations of aerosol and radiation, if supplemented by the surface observations of downward and upward SW and LW, can be used together to derive aerosol forcing not only to space, but also at the surface and in the atmosphere.

[14] Dust radiative effects on TOA SW and LW flux are estimated over the Taklimakan Desert by merging the MISR AOD and CERES SSF flux at the TOA. The current study clearly shows the reflected SW flux to space tends to increase whereas the emitted LW flux at the TOA tends to decrease as the dust loading increases, both over the cloud free Taklimakan desert regions in May. The regional mean dust SW and LW forcing efficiencies at 05:00 UTC are −48.1 and 28.4 W m−2. This indicates that dust instantaneous LW warming effect can offset 58% of dust instantaneous SW cooling effect. The overall radiative effect of Taklimakan dust is still to cool Earth's system even when the diurnal variation of SW and LW forcing is considered.


[15] The CERES SSF and MISR AOD data were obtained from the Atmospheric Science Data Center at the NASA Langley Research Center. The research is supported by the National Basic Research Program of China (2009CB723904) and the National Natural Science Foundation of China (40775009 and 40875084).