Skip to main content
NCBI home page
NASA Author Manuscripts logoLink to NASA Author Manuscripts
. Author manuscript; available in PMC: 2020 Jul 13.
Published in final edited form as: Geophys Res Lett. 2017 Jul 10;44(14):7544–7554. doi: 10.1002/2017gl073939

Passive remote sensing of altitude and optical depth of dust plumes using the oxygen A and B bands: first results from EPIC/DSCOVR at Lagrange-1 point

Xiaoguang Xu 1, Jun Wang 1, Yi Wang 1, Jing Zeng 1, Omar Torres 2, Yuekui Yang 3, Alexander Marshak 3, Jeffrey Reid 4, Steve Miller 5
PMCID: PMC7357207  NIHMSID: NIHMS902885  PMID: 32661445

Abstract

We presented an algorithm for inferring aerosol layer height (ALH) and optical depth (AOD) over ocean surface from radiances in oxygen A and B bands measured by the Earth Polychromatic Imaging Camera (EPIC) on the Deep Space Climate Observatory orbiting at Lagrangian-1 point. The algorithm was applied to EPIC imagery of a two-day dust outbreak over the North Atlantic Ocean. Retrieved ALHs and AODs were evaluated against counterparts observed by Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP), Moderate Resolution Imaging Spectroradiometer (MODIS), and Aerosol Robotic Network. The comparisons showed 71.5% of EPIC-retrieved ALHs were within ±0.5 km of those determined from CALIOP and 74.4% of EPIC AOD retrievals fell within a ±(0.1+10%) envelope of MODIS retrievals. This study demonstrates the potential of EPIC measurements for retrieving global aerosol height multiple times daily, which are essential for evaluating aerosol profile simulated in climate models and for better estimating aerosol radiative effects.

1. Introduction

Vertical distribution of atmospheric aerosols influences the Earth’s climate and environment in many ways. First, absorption of solar and terrestrial radiation by smoke and dust particles modifies the air thermal state and stability, and the resulting effect on the atmospheric stability depends on the altitude of aerosol layers [Babu et al., 2011; Wendisch et al., 2008; Zhu et al., 2007]. Second, the altitude of absorbing particles relative to clouds can influence cloud cover and lifetime [Koch and Del Genio, 2010; Satheesh et al., 2008; Ge et al., 2014]. Thus, aerosol vertical distribution can affect the magnitude, and even the sign of aerosol direct and indirect radiative effects [Choi and Chung, 2014; Samset et al., 2013]. However, centroid altitude of aerosol layers simulated by current climate models can differ by up to an order of magnitude in the range of 1.5 km to 3 km [Koffi et al., 2012; Kipling et al., 2016], leading to considerable uncertainty in the estimated aerosol radiative forcing. Furthermore, elevated dense aerosol plumes such as lofted mineral dust and volcanic ash, which are invisible to the aircraft radar, can pose significant hazards on aviation safety [Sears et al., 2013]. Therefore, it is critically important to observe the vertical distribution of aerosols on a global scale.

Detailed profile of attenuated backscatter can be probed by active remote sensing techniques using lidar, such as the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) [Winker et al., 2009]. However, the spatial coverage of CALIOP measurements suffers from its narrow swath. The gap between its adjacent sub-orbital “curtains” is as wide as 2200 km, especially in the tropics to subtropics. By contrast, passive remote sensing techniques provide adequate spatial coverage but poor vertical resolution, and have been mainly used to retrieve columnar aerosol quantities in cloud-free scenes. Nevertheless, various passive techniques have been developed to retrieve limited but useful information of aerosol altitude [Xu et al., 2017]. While not achieving the same level of accuracy as a lidar, passive techniques can add an important augmentation due to the better spatial coverage.

One passive technique utilizes absorption spectroscopy of molecular oxygen (O2) [e.g., Zeng et al., 2008; Dubuisson et al., 2009; Kokhanovsky and Rozanov, 2010; Sanders et al., 2015; Ding et al., 2016]. Its physical principle relies on that the aerosol layer can scatter photons back to space and reduce the chance of a photon being absorbed by the underlying O2 molecules. A risen scattering layer increases the chance for photons to be scattered, enhancing the reflectivity in the O2 absorption bands as detected by a satellite [Dubuisson et al., 2009; Wang et al., 2014]. Therefore, spectral characteristics of reflectance in O2 bands can manifest how aerosol particles interact with O2 absorption through multiple scattering at different altitudes. With this principle, attempts have been made to retrieve aerosol height using observations in the O2 A band from low-earth-orbiting (LEO) instruments including POLDER (POLarization and Directionality of the Earth’s Reflectances) and MERIS (Medium Resolution Imaging Spectrometer) [Dubuisson et al., 2009], SCIAMACHY (SCanning Imaging Absorption spectroMeter for Atmospheric CHartographY) [Kokhanovsky and Rozanov, 2010; Sanghavi et al., 2012], and GOME/GOME-2 (Global Ozone Monitoring Experiment) [Koppers and Murtagh, 1997; Sanders et al., 2015]. Overall, these studies showed the height of aerosol layers over ocean can be properly retrieved when aerosol layers are in the free troposphere. The height information for aerosols within boundary layers are limited, but can be enhanced by combining O2 A and B bands in which O2 absorption strengths are different [Pflug and Ruppert, 1993; Ding et al., 2016]; such enhancement was shown primarily through theoretical studies but with limited analysis of real data and validation [Sanghavi et al., 2012].

On February 11, 2015, NASA launched the Deep Space Climate Observatory (DSCOVR), which carries the Earth Polychromatic Imaging Camera (EPIC) to observe Earth-reflected solar radiation in ten narrow channels covering both the O2 A and B bands [Marshak and Knyazikhin, 2017]. DSCOVR flies on the Lissajous orbit about 1.5 million kilometers away at the Lagrange-1 point, giving EPIC a unique angular perspective at backscattered direction with scattering angle between 168° – 176°. EPIC scans the entire sunlit face of the Earth at a 2048 × 2048-pixel resolution, rendering the pixel size of 12 × 12 km2 at the image center. In contrast to LEO sensors that observe a location once a day, EPIC can track the change of aerosol layer height multiple times a day. Also, EPIC provides a finer spatial resolution and wider coverage than LEO sensors that include both O2 A and B bands, i.e., 80 × 40 km2 for GOME-2 and 30 × 60 km2 for SCIAMACHY.

This paper describes an algorithm that for the first time, uses real measurements in O2 A and B bands from EPIC to estimate the centroid altitude and optical depth of aerosols residing over a water background. We demonstrated the algorithm through applying it to dust plumes over the North Atlantic Ocean for two adjacent days, and evaluated the retrievals with a comprehensive use of MODIS (Moderate Resolution Imaging Spectroradiometer), AERONET (Aerosol Robotic Network), and CALIOP data. This paper is outlined as follows. The specifics of the EPIC sensor are introduced in section 2, the retrieval method is described in section 3, case study and validation are presented in section 4. Finally, we summarize the findings and consider future research in section 5.

2. EPIC measurements

EPIC measures Earth-reflected solar radiance in ten narrow bands spanning from UV to near-infrared (NIR) spectrum, which include four channels around the O2 A and B. As shown in Figure S1, two absorption channels are centered at 688 nm and 764 nm with bandwidth of 0.8 nm and 1.0 nm, respectively. Two continuum bands are placed at 680 nm and 780 nm. These bands were designed for monitoring vegetation condition [Marshak and Knyazikhin, 2017] and cloud height [Yang et al., 2013].

We used EPIC Level 1b (L1B) version-1 radiance made available from https://eosweb.larc.nasa.gov. Various pre- and post-launch calibrations were applied to the L1B EPIC data [Cede and Matt Kowalewski, 2016]. EPIC visible channels were cross-calibrated with LEO instruments like the MODIS and the Visible Infrared Imaging Radiometer Suite (VIIRS) [Haney et al., 2016]. The two O2 absorption channels were calibrated by lunar surface reflectivity [Herman et al., 2016]. The top-of-the-atmosphere (TOA) reflectance are calculated by,

ρλ=πI(λ)μ0E0(λ), (1)

where I(λ) and E0(λ) are, respectively, EPIC measured radiance and solar irradiance at wavelength of λ, μ0 is the cosine of solar zenith angle.

3. Methodology

3.1. Radiative transfer modeling and retrieval principle

To construct the relationship between EPIC measurements and aerosol vertical profile, TOA reflectance was simulated using the Unified Linearized Vector Radiative Transfer Model (UNL-VRTM) [Wang et al., 2014]. UNL-VRTM couples the VLIDORT [Spurr, 2006] radiative transfer code to the HITRAN molecular spectroscopic database [Rothman et al., 2009] and aerosol scattering codes. The solar irradiance reference spectrum from Chance and Kurucz [2010] was applied. Radiance was then obtained by convolving simulated hyperspectral monochromatic radiances with EPIC instrument filter functions for 680, 688, 764, and 780 nm bands.

We assumed a Gaussian-like aerosol profile characterized by a columnar AOD, centroid altitude, and a half-width parameter (Eq. S1 and Figure S2) [Spurr and Christi, 2014]. The centroid altitude was used to represent aerosol layer height (ALH). Aerosol single scattering properties at EPIC bands were determined from the AERONET inversion products that treat dust particles as spheroids [Holben et al., 1998; Dubovik et al., 2006]. Following Xu et al. [2015], we compiled AERONET version-2 inversion data over Capo_Verde site retrieved for 2015 and determined the climatological single scattering albedo (SSA) and phase function (Figure S3), as well as Ångstrom exponent (AE) between 675 and 870 nm, in the dusty conditions when coarse-mode volume fraction is larger than 0.6 and 440-nm AOD above 0.4.

Figure 1 shows the relationship between dust ALH and reflectance (and ratio) in EPIC O2 bands for various AOD values. Those simulations were performed for an ocean surface with reflectance of 0.015 at an EPIC geometry of [θ0, θ, ϕ] = [42°, 31°, 0°], where θ0 and θ are solar and view zenith angles, and ϕ the relative azimuth angle. As seen from the figure, while reflectance in the continuum bands are sensitive to AOD but not ALH (Figure 1a–b), ratios of reflectance of the absorption-to-continuum bands are sensitive to both AOD and ALH (Figure 1c–d). The higher the aerosol layer, the larger the reflectance ratio; this can be explained that a higher scattering layer gives more chance for photons being scattered and increase reflectance in the absorption band. Meanwhile, sensitivity of the reflectance ratio to ALH decreases as AOD decreases and diminishes as AOD approaches zero. Additional simulations with different surface reflectance and geometry were also performed (not shown), which were found to be consistent with previous studies in that the sensitivity of reflectance ratio to ALH decreased with an increase of surface reflectance and varied with geometry [Boesche et al, 2009; Dubuisson et al., 2009; Wang et al., 2014; Ding et al., 2016].

Figure 1.

Figure 1.

Open in a new tab

Relationship between ALH and TOA reflectance in two EPIC continuum bands at 680 nm and 780 nm for various AOD values (a and b, respectively). Relationship between ALH and reflectance ratio in the EPIC O2 A and B bands for various AOD values (c and d, respectively). Simulations are performed for dust aerosol optical property (Figure S3) under following conditions: a solar zenith angle of 42°, a view zenith angle of 31°, a relative azimuth angle of 0° (backscatter), and a surface reflectance of 0.015. Note the colors of dots indicate different AOD values.

Our algorithm retrieves ALH and AOD from EPIC measurements in the context of look-up tables (LUTs) that contain TOA reflectance at EPIC bands simulated for a set of predefined ALH and AOD values and for various surface reflectance and satellite geometries:

  • 680-nm AOD values of 0.01, 0.1, 0.3, 0.4, 0.6, 0.8, 1.0, 1.2, 1.5, 2.0, and 3.0;

  • ALH from 0 to 15 km with 1 km intervals and half width of 1 km;

  • Surface reflectance values of 0, 0.015, 0.03, 0.05, and 0.1;

  • θ0 and θ from 0° to 78° with 6° intervals and |θ0 – θ| < 15° (EPIC backscatter), and ϕ from 0° to 180° with 12° intervals.

The assumed half width of 1 km is representative of an average profile observed by Lidar for dust and smoke aerosols [Reid et al., 2003]. This value was also used to derive AOD from UV observations by TOMS (Total Ozone Mapping Spectrometer) and OMI (Ozone Monitoring Instrument) [Torres et al., 1998].

3.2. Retrieval procedures

Our retrieval algorithm performs a simultaneous inversion of four EPIC channels (680, 688, 764, and 780 nm) to solve AOD and ALH. In brief, it entails three steps. First, we aggregated EPIC imagery into 0.1° × 0.1° grids using nearest-neighbor interpolation for the entire 10 channels. Geo-registration of EPIC radiance is different from one band to another because monochromatic imagery is taken sequentially while the Earth is rotating. This step ensures a consistent geographical coordinate across EPIC bands. The retrieval results are also delivered at the aggregated grids.

The second step involves screening of cloud and sun-glint to identify pixels that are suitable for aerosol retrieval. The sun-glinted area over ocean with a glint angle smaller than 30° was removed [Levy et al., 2013]. Cloud pixels were screened out through brightness and homogeneity tests [Martins et al., 2002], which is illustrated in Figure S4. However, thin and small sub-pixel clouds can be easily overlooked from 12-km EPIC pixels, leading to an overestimated AOD and underestimated ALH (Figure 1c–d). Therefore, cloud screening is a potential issue to the retrieval quality, and one that may benefit from enlisting higher-resolution geostationary sensors’ cloud mask information for future efforts.

Finally, ALH and AOD were determined from EPIC observations in the context of LUTs. We first interpolated LUTs over EPIC observation geometry and surface reflectance determined by GOME-2 surface Lambert-equivalent reflectivity database [Koelemeijer et al., 2003]. Based on the interpolated LUT, we refined the AOD grid with 0.01 intervals and the ALH grid with 0.1 km intervals and interpolated reflectance for the refined AOD and ALH values. The inverse problem was then formulated in a way that the optimal values of AOD and ALH corresponded to the minimization of a scalar cost function,

∈=14∑i=14[yi−Fi(x)]2, (2)

where yi is EPIC reflectance in band i, x is a state vector comprising ALH and AOD, and is Fi(x) the LUT reflectance in band i for any given x. The ϵ parameter was reported along with the retrieved ALH and AOD and can be an indicator of retrieval quality (Figure S5).

4. Application to EPIC imagery over marine surfaces

4.1. Retrieval results

Our case study focused on dust plumes over subtropical North Atlantic Ocean. Saharan dust is often transported across this region before reaching the Caribbean Sea [Prospero et al., 2002; Yu et al., 2010, 2015]. The vertical structure and dynamics of the dust layer over this region are controlled by large-scale atmospheric circulation and regional-scale convection [Karyampudi et al., 1999; Reid et al, 2002; 2003; Huang et al., 2010; Yu et al., 2010, 2015]. In late spring and summer, dust aerosols are lofted from the Sahara deserts by strong sensible heating and low-level convergence and travel along with trade winds in a warm and dry air mass known as the Saharan Air Layer (SAL) [Prospero and Carlson, 1972]. Typically, the SAL extends from 1 km to 6 km with dust concentration peaked between 1.5 and 3.7 km. The top altitude of the SAL descends as air travels away from the African coast [Carlson and Prospero, 1972; Karyampudi et al., 1999; Maring et al., 2003; Colarco et al., 2003], a characteristic that is likely related to the subtropical subsidence [Huang et al., 2010]. Regional subsidence can also bring elevated dust aerosols into marine boundary layer [Colarco et al., 2003]. As a result, dust altitude in North Atlantic Ocean is subject to large seasonal and spatial variability [Huang et al., 2010; Yu et al., 2010, 2015].

We applied our algorithm to four EPIC scenes acquired on 17 – 18 April 2016, which provided optimal comparison with MODIS and CALIOP observations (Section 4.2). In the EPIC RGB images (Figure 2a), dust outflows appeared as light brown plumes over the dark marine surface. The retrievals of ALH, shown in Figure 2b, manifested strong spatial variability. In the northwest African coast, ALH ranged from 3 km to 5 km in the area north of 15° N but decreased to about 2 km towards south. This variation was likely controlled by large-scale circulation. As shown in Figure S6, strong upward air motions coincided with high ALH in the area north of 15° N, and low ALH at lower latitudes was associated with subtropical subsidence of ITCZ’s northern branch. This is consistent with the study by Karyampudi et al. [1999], who showed that intense sensible heating in the Sahara could produce strong dry convection uplifting dust from the surface to SAL [Carlson and Prospero, 1972]. As dust aerosols transported off the coast, SAL altitude gradually descended. On diurnal basis, higher ALH were found in the afternoon, which might be related to diurnal evolution of tropospheric convection.

Figure 2.

Figure 2.

Open in a new tab

Retrieval results for dust plumes from 4 EPIC images during 17 – 18 April 2016. Row a: enhanced RGB images created using EPIC measurements at 680 nm, 551 nm, and 440 nm. Row b: aerosol layer height (ALH) retrieved from EPIC measurements when AOD is over 0.3. Red lines indicate CALIPSO sub-orbital track for the same-day overpass. Row c: EPIC retrieved aerosol optical depth (AOD) at 680 nm. Row d: MODIS level 2 AOD (collection 6) from both Terra and Aqua satellites, which were spectrally interpolated from 0.66 μm and 0.87 μm to 680 nm using Ångstrom exponent (AE).

Figure 2c shows retrieved 680-nm AOD, which clearly illustrate the transport of dust. On April 17, dust plumes with large AOD values moved off the Saharan coast towards southeast. On the second day, dispersed dust plumes were found over Cape Verde and extended to 30° W. In addition, scattered retrievals with AOD larger than 1 were found around 12° N between 40° W and 30° W, which were possibly contaminated by sub-pixel clouds.

4.2. Comparison of AOD retrievals with MODIS and AERONET

We first compared EPIC retrieved AOD with MODIS Collection 6 level-2 10-km AOD products [Remer et al., 2005; Levy et al., 2013]. Products from both Terra and Aqua satellites were used to evaluate EPIC retrievals in local morning and afternoon, respectively. The expected uncertainty of MODIS AOD was claimed to be (+(0.04+10%), −(0.02+10%)) over ocean [Levy et al., 2013]. However, MODIS ocean AOD retrieval assumes spherical aerosols which, in the case of dust, can cause larger and angular-dependent biases [Levy et al., 2003]. A recent evaluation by Banks et al. [2017] indicated such retrieval bias can be as large as +0.13. Therefore, it should be emphasized that MODIS AODs did not server as ground truth for validating EPIC retrievals. Rather, they were used for an inter-comparison.

Figure 2d shows MODIS AOD at 680 nm, which were spectrally interpolated from 0.66 and 0.87 μm. In comparison, EPIC retrievals (Figure 2c) exhibited wider spatial coverage and captured the overall pattern of dust plumes as identified by MODIS. Meanwhile, regional disparities were present between EPIC and MODIS AODs, which could be caused by differences in the effectiveness of cloud masking, used aerosol models, and observation time. EPIC, with pixel size over ten times larger, was more likely to be contaminated by thin and small-scale clouds. Also, our EPIC algorithm used a non-spherical dust aerosol model, whereas MODIS aerosol algorithm combined a fine and a coarse spherical aerosol mode. In addition, there were up to 1 hour time differences between MODIS and EPIC observations. Rapidly evolving dust plumes could result in substantial diurnal variations of AOD [Wang et al., 2003].

Figure 3a shows a scatterplot of EPIC AOD versus MODIS AOD, both aggregated into a 0.5°x0.5° grid. In total, 74.7% of the EPIC retrievals fell within a ±(0.1+10%) envelope of MODIS AOD. EPIC AOD captured 36% variance of the MODIS AOD with a root-mean-squared error (RMSE) of 0.15. The metric of ±(0.1+10%) was selected to indicate an expected uncertainty level of EPIC AOD. This uncertainty is larger than that claimed for MODIS over-ocean AOD [Levy et al., 2013] to account for a more acute cloud contamination.

Figure 3.

Figure 3.

Open in a new tab

Comparison of EPIC retrieved AOD and ALH with the counterparts from MODIS, AERONET, and CALIOP measurements. (a) Scatterplot of EPIC retrieved 680 nm AOD versus MODIS C6 680 nm AOD, both aggregated into a 0.5° spatial resolution. Each scatter indicates an AOD pair over a 0.5°x0.5° grid. Color of scatter indicates the frequency of scatters falling within 0.025 AOD intervals. Also shown are one-by-one line (solid) and ±(0.1+10%) (dashed) envelop. (b) Scatterplot of EPIC 680-nm AOD (blue) and MODIS 680-nm AOD (red) versus AERONET 675-nm AOD. Error bars indicate standard deviation of sampled subsets. One-by-one line and ±(0.1+10%) envelope are indicated by solid and dashed lines, respectively. (c) Scatterplot of EPIC retrieved ALH versus CALIOP extinction-weighted ALH sampled for cloud-free conditions. Annotated in these scatterplots are one-by one line (solid), ±0.5km (dashed) and ±1.0km (dotted) envelops. Annotated in each panel are number of samples (N), bias, root mean squared error (RMSE), and coefficient of determination (R2).

AERONET version-2 AOD observations [Holben et al., 1998] were used to validate EPIC retrievals over Calhau, Capo_Verde, and Dakar sites. Site location and time series of AERONET AOD and collocated EPIC retrievals are illustrated in Figure S7. Observed AE at these AERONET sites were below 0.2 throughout the study period, indicating the overwhelming of dust aerosols. Figure 3b presents a scatterplot of collocated EPIC and MODIS retrievals versus AERONET AODs. The collocation method follows Ichoku et al. [2002] but was updated to associate a subset of satellite retrievals within a 0.5° × 0.5° grid centered at each site to a subset of 1-hour AERONET observations around satellite overpass time. The collocated EPIC retrievals, though with limited data samplings, all fell within the ±(0.1+10%) envelope of AERONET AOD with a RMSE of 0.07 and a coefficient of determination (R2) of 0.81. This represents a better performance than MODIS retrievals (with RMSE of 0.09 and R2 of 0.68). The positive bias in EPIC AOD was 0.02. This bias was dominated for two subsets of EPIC AODs with large spatial variation (Figure 3b and S7), which was likely caused by cloud contamination.

4.3. Validation of ALH retrievals with CALIOP observations

To validate EPIC retrieved ALH, we used CALIOP level-2 aerosol extinction profiles at a spatial resolution of 60-m vertically and 5-km horizontally, which were retrieved from CALIOP attenuated backscatter at 532 nm [Young and Vaughan, 2009]. To facilitate quantitative comparison, a mean ALH was calculated from the CALIOP extinction profile following Koffi et al. [2012],

ALHCALIOP=∑i=1nβext,iZi∑i=1nβext,i. (3)

Here, βext,i is aerosol extinction coefficient (km−1) at level i and Zi is the level altitude (km). is n 166 representing the considered top level at 10 km. Thus, ALHCALIOP represents an effective ALH weighted by aerosol extinction signal at each level, and is directly comparable with ALH defined in our retrieval algorithm.

Figure 4 compares EPIC-retrieved ALH (black) and CALIOP extinction profiles. The CALIPSO sub-orbital track was along the coastal marine surface on the April 17 and moved towards west on the next day (Figure 2b). As evident in CALIOP profiles, the spatial variability of EPIC-retrieved ALH generally agreed with that observed by CALIOP. On April 17, elevated dust layers appeared between 3 km and 6 km in the area north to 15°N. At lower latitudes, aerosol layers extended from the surface to about 4 km. Correspondingly, the ALHCALIOP was about 4 km in the north and 2 km at lower latitudes. On April 18, ALHCALIOP gradually decreased from about 3 km in the north to 2 km at lower latitudes. A scatterplot of EPIC ALH versus ALHCALIOP in cloud-free scenes is shown in Figure 3c. We found 71.5% and 98.7% of EPIC ALH retrievals were located within ± 0.5 km and ± 1.0 km envelops of ALHCALIOP, respectively. The variability of EPIC ALH explained 72% variance of ALHCALIOP with a RMSE of 0.45 km. Therefore, we expect that the accuracy of EPIC-retrieved ALH is better than 0.5 km.

Figure 4.

Figure 4.

Open in a new tab

Validation of EPIC ALH (black curve) with CALIOP profile of 532-nm extinction coefficient. Panel a and b are for CALIPSO overpasses on April 17 and 18, respectively. The corresponding sub-orbital tracks are indicated in Figure 2b. Cloud layers are indicated by gray color. The red curve depicts CALIOP extinction-weighted ALH calculated by (Eq. 3).

5. Discussion and Conclusion

In this paper, we presented a new retrieval algorithm that infers aerosol layer height (ALH) and optical depth (AOD) from backscattered radiation in O2 A and B bands observed by the EPIC/DSCOVR orbiting around the Lagrangian-1 point. The algorithm was applied to EPIC imagery for dust plumes acquired over North Atlantic Ocean during 17 – 18 April 2016. We found that EPIC-retrieved ALHs were in good agreement with CALIOP-probed aerosol extinction profile, with 71.5 % of EPIC retrievals falling within ±0.5 km of CALIOP extinction-weighted altitude and a RMSE of 0.45 km. Thus, retrieval accuracy of EPIC ALH is expected to be better than 0.5 km. EPIC AOD retrievals were well validated by AERONET AOD. The AOD retrievals agreed with MODIS AOD in that 74.7% of EPIC AOD retrievals are in a ±(0.1+10%) envelope of MODIS AOD.

The implication of this study is twofold. First, EPIC observations can provide global aerosol height information multiple times a day, which will allow further study of diurnal variation of aerosol vertical structure. Such information is valuable for evaluating the vertical distribution of aerosols estimated by climate models [Koffi et al., 2012] and for better estimating aerosol radiative effects [Zhang et al., 2013]. Second, the retrieved ALH can provide complementary height information for determining absorbing aerosol properties from UV bands. EPIC also measures backscattered UV radiances at 340 nm and 388 nm, which was designed to detect UV-absorbing aerosols like mineral dust and smoke [Torres et al., 2016]. However, inferring aerosol properties from those UV bands requires the characterization of aerosol height, because UV radiance is sensitive to aerosol vertical distribution [Torres et al., 1998]. For example, Jeong and Hsu [2008] retrieved SSA from OMI radiance with synergic use of AOD from MODIS and aerosol height from CALIOP. With the aerosol height and loading available from EPIC O2 A and B bands, these closures are now possible with measurements from a single instrument.

The current algorithm incorporates a dust aerosol model, thus restricts its application to dust-laden scenes over a marine surface. The presence of non-dust aerosols, such as transported biomass-burning smoke, could bias the retrieved AOD and ALH. The retrieval accuracy could also be affected by sea salt particles, though the effect of which tends to be small as sea salt particles usually place close to marine surface and has a much lower AOD than dust. Considering that the combined O2 A and B bands can allow aerosol height retrieval over bright surface [Sanghavi et al., 2012] and that EPIC measurements in the UV and visible bands can be potentially used to discriminate aerosol types, a subsequent algorithm enhancement will include other aerosol types and land surfaces by utilizing more EPIC channels. Another limitation of the current algorithm exists for the case of multiple aerosol layers due to the assumption of a single Gaussian-like profile. This limitation is intrinsic to EPIC’s narrow-band measurements that contain limited information for retrieving detailed aerosol vertical structure. Nevertheless, the retrieval of multiple-layered vertical distribution would be possible from hyperspectral radiometric and polarimetric measurements in the O2 A and B bands [Ding et al., 2016], such as those measured by the TROPOspheric Monitoring Instrument (TROPOMI) on board of the Copernicus Sentinel-5 Precursor satellite [Sanders et al., 2015].

Supplementary Material

Supplementary Information

Key Points.

  • Algorithm to retrieve dust optical depth and centroid height using the O2 A and B bands is developed.

  • First retrieval results of dust optical depth and altitude from EPIC/DSCOVR are shown in good agreement with the counterparts from MODIS and CALIPSO.

  • Passive remote sensing of aerosol height multiple times within a day is demonstrated with EPIC and discussed for future studies.

Acknowledgements

This research is in part supported by NASA’s DSCOVR Earth Science Algorithms Program (Grant No. NNX17AB05G managed by Richard S. Eckman) and in part supported by Office of Naval Research (ONR’s) Multidisciplinary University Research Initiatives (MURI) Program under the award No. N00014-16-1-2040. We acknowledge the computational support from the High Performance Computing group at University of Iowa. All the data presented in this manuscript will be made available through Coalition on Publishing Data in the Earth and Space Sciences (https://copdessdirectory.osf.io), and please email J. Wang (jun-wang-1@uiowa.edu) and X. Xu (xiaoguang-xu@uiowa.edu) for details.

References:

  1. Babu SS, Moorthy KK, Manchanda RK, Sinha PR, Satheesh SK, Vajja DP, Srinivasan S, and Kumar VHA (2011), Free tropospheric black carbon aerosol measurements using high altitude balloon: Do BC layers build “their own homes” up in the atmosphere?, Geophysical Research Letters, 38(8), L08803. [Google Scholar]
  2. Banks JR, Brindley HE, Stenchikov G, and Schepanski K (2017), Satellite retrievals of dust aerosol over the Red Sea and the Persian Gulf (2005–2015), Atmos. Chem. Phys, 17(6), 3987–4003. [Google Scholar]
  3. Boesche E, Stammes P, and Bennartz R (2009), Aerosol influence on polarization and intensity in near-infrared O2 and CO2 absorption bands observed from space, Journal of Quantitative Spectroscopy and Radiative Transfer, 110(3), 223–239. [Google Scholar]
  4. Carlson TN, and Prospero JM (1972), The Large-Scale Movement of Saharan Air Outbreaks over the Northern Equatorial Atlantic, Journal of Applied Meteorology, 11(2), 283–297. [Google Scholar]
  5. Cede A and Kowalewski M (2016), EPIC optical design, calibration, and data correction (talk), A21K-04, presented in The AGU 2016 Fall Meeting, San Francisco, CA, Dec 12–16, 2016. [Google Scholar]
  6. Chance K, and Kurucz RL (2010), An improved high-resolution solar reference spectrum for earth’s atmosphere measurements in the ultraviolet, visible, and near infrared, Journal of Quantitative Spectroscopy and Radiative Transfer, 111(9), 1289–1295. [Google Scholar]
  7. Choi J-O, and Chung CE (2014), Sensitivity of aerosol direct radiative forcing to aerosol vertical profile, Tellus B, 66, 24376. [Google Scholar]
  8. Colarco PR (2003), Saharan dust transport to the Caribbean during PRIDE: 2. Transport, vertical profiles, and deposition in simulations of in situ and remote sensing observations, Journal of Geophysical Research, 108(D19), 8590. [Google Scholar]
  9. Ding S, Wang J, and Xu X (2016), Polarimetric remote sensing in oxygen A and B bands: sensitivity study and information content analysis for vertical profile of aerosols, Atmos. Meas. Tech, 9(5), 2077–2092. [Google Scholar]
  10. Dubovik O, et al. (2006), Application of spheroid models to account for aerosol particle nonsphericity in remote sensing of desert dust, Journal of Geophysical Research: Atmospheres, 111, D11208. [Google Scholar]
  11. Dubuisson P, Frouin R, Dessailly D, Duforêt L, Léon J-F, Voss K, and Antoine D (2009), Estimating the altitude of aerosol plumes over the ocean from reflectance ratio measurements in the O2 A-band, Remote Sensing of Environment, 113(9), 1899–1911. [Google Scholar]
  12. Haney C, Doelling D, Minnis P, Bhatt R, Scarino B, and Gopalan A (2016), The calibration of the DSCOVR EPIC multiple visible channel instrument using MODIS and VIIRS as a reference, in Proc. SPIE 9972, Earth Observing Systems XXI, 99720P. [Google Scholar]
  13. Herman J, Szabo A, and Marshak A (2016), DSCOVR/EPIC: Calibration and related issues (talk), presented in DSCOVR EPIC & NISTAR Science Team Meeting, NASA Goddard Space Flight Center, Greenbelt, MD, 8–9 Feb 2016. [Google Scholar]
  14. Holben BN, et al. (1998), AERONET - A Federated Instrument Network and Data Archive for Aerosol Characterization, Remote Sensing of Environment, 66(1), 1–16. [Google Scholar]
  15. Huang J, Zhang C, and Prospero JM (2010), African dust outbreaks: A satellite perspective of temporal and spatial variability over the tropical Atlantic Ocean, Journal of Geophysical Research, 115, D05202. [Google Scholar]
  16. Ichoku C, Chu DA, Mattoo S, Kaufman YJ, Remer LA, Tanré D, Slutsker I, and Holben BN (2002), A spatio-temporal approach for global validation and analysis of MODIS aerosol products, Geophysical Research Letters, 29(12), 1616, doi: 10.1029/2001GL013206. [DOI] [Google Scholar]
  17. Jeong M-J, and Hsu NC (2008), Retrievals of aerosol single-scattering albedo and effective aerosol layer height for biomass-burning smoke: Synergy derived from “A-Train” sensors, Geophysical Research Letters, 35(24), L24801. [Google Scholar]
  18. Karyampudi VM, et al. (1999), Validation of the Saharan Dust Plume Conceptual Model Using Lidar, Meteosat, and ECMWF Data, Bulletin of the American Meteorological Society, 80(6), 1045–1075. [Google Scholar]
  19. Kipling Z, et al. (2016), What controls the vertical distribution of aerosol? Relationships between process sensitivity in HadGEM3–UKCA and inter-model variation from AeroCom Phase II, Atmospheric Chemistry and Physics, 16(4), 2221–2241. [Google Scholar]
  20. Koch D and Del Genio AD (2010), Black carbon semi-direct effects on cloud cover: review and synthesis, Atmos. Chem. Phys, 10, 7685–7696. [Google Scholar]
  21. Koelemeijer RBA, de Haan J, and Stammes P (2003), A database of spectral surface reflectivity in the range 335–772 nm derived from 5.5 years of GOME observations, Journal of Geophysical Research, 108(D2), 4070. [Google Scholar]
  22. Koffi B, et al. (2012), Application of the CALIOP layer product to evaluate the vertical distribution of aerosols estimated by global models: AeroCom phase I results, Journal of Geophysical Research: Atmospheres, 117(D10), D10201. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Kokhanovsky AA, and Rozanov VV (2010), The determination of dust cloud altitudes from a satellite using hyperspectral measurements in the gaseous absorption band, International Journal of Remote Sensing, 31(10), 2729–2744. [Google Scholar]
  24. Koppers GAA and Murtagh DP (1997), Retrieval of height resolved aerosol optical thickness in the atmospheric band, in: Radiative Transfer in the Absorption Bands of Oxygen: Studies of their Significance in Ozone Chemistry and Potential for Aerosol Remote Sensing, edited by: Koppers GAA, Stockholm University, Stockholm, Sweden, 1–24. [Google Scholar]
  25. Levy RC, Remer LA, Tanré D, Kaufman YJ, Ichoku C, Holben BN, Livingston JM, Russell PB, and Maring H (2003), Evaluation of the Moderate-Resolution Imaging Spectroradiometer (MODIS) retrievals of dust aerosol over the ocean during PRIDE, Journal of Geophysical Research, 108(D19), 8594. [Google Scholar]
  26. Levy RC, Mattoo S, Munchak LA, Remer LA, Sayer AM, Patadia F, and Hsu NC (2013), The Collection 6 MODIS aerosol products over land and ocean, Atmospheric Measurement Techniques, 6, 2989–3034. [Google Scholar]
  27. Maring H, Savioe DL, Izaguirre MA, Custals L, and Reid JS (2003), Vertical distributions of dust and sea-salt aerosols over Puerto Rico during PRIDE measured from a light aircraft, Journal of Geophysical Research, 108(D19), 8587. [Google Scholar]
  28. Marshak A, and Knyazikhin Y (2017), The spectral invariant approximation within canopy radiative transfer to support the use of the EPIC/DSCOVR oxygen B-band for monitoring vegetation, Journal of Quantitative Spectroscopy and Radiative Transfer, 191, 7–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Martins JV, Tanré D, Remer L, Kaufman Y, Mattoo S, and Levy R (2002), MODIS Cloud screening for remote sensing of aerosols over oceans using spatial variability, Geophysical Research Letters, 29(12), 1619. [Google Scholar]
  30. Pflug BM, and Ruppert T (1993), Information content of measurements in the O2A- and O2B-bands for monitoring of aerosols from space, SPIE Vol. 1968 Atmospheric Propagation and Remote Sensing II, 533–544. [Google Scholar]
  31. Prospero JM, and Carlson TN (1972), Vertical and areal distribution of Saharan dust over the western equatorial north Atlantic Ocean, Journal of Geophysical Research, 77(27), 5255–5265. [Google Scholar]
  32. Prospero JM, Ginoux P, Torres O, Nicholson SE, and Gill TE (2002), Environmental characterization of global sources of atmospheric soil dust identified with the NIMBUS 7 Total Ozone Mapping Spectrometer (TOMS) absorbing aerosol product, Rev. Geophys, 40(1), 1002. [Google Scholar]
  33. Reid JS, Kinney JE, Westphal DL, Holben BN, Welton EJ, Tsay S-C, Christopher SA, Eleuterio DP, Jonsson HH, Livingston JM, Maring HB, Meier M, Pilewskie P, Reid EA, Russell PB, Savoie D, Smironov A, and Tanré D (2003), Analysis of measurements of Saharan dust by airborne and ground-based remote sensing methods during the Puerto Rico Dust Experiment (PRIDE), J. Geophys. Res, 108(D19), 8586. [Google Scholar]
  34. Reid JS, Westphal DL, Livingston J, Savoie DS, Maring HB, Pilewskie P, Eleuterio D (2002), The vertical distribution of dust transported into the Caribbean during the Puerto Rico Dust Experiment, Geophys. Res. Lett, 29(7), 1151. [Google Scholar]
  35. Remer LA, et al. (2005), The MODIS Aerosol Algorithm, Products, and Validation, Journal of the Atmospheric Sciences, 62(4), 947–973. [Google Scholar]
  36. Samset BH, et al. (2013), Black carbon vertical profiles strongly affect its radiative forcing uncertainty, Atmos. Chem. Phys, 13(5), 2423–2434. [Google Scholar]
  37. Sanders AFJ, de Haan JF, Sneep M, Apituley A, Stammes P, Vieitez MO, Tilstra LG, Tuinder ONE, Koning CE, and Veefkind JP (2015), Evaluation of the operational Aerosol Layer Height retrieval algorithm for Sentinel-5 Precursor: application to O2 A band observations from GOME-2A, Atmospheric Measurement Techniques, 8(11), 4947–4977. [Google Scholar]
  38. Sanghavi S, Martonchik JV, Landgraf J, and Platt U (2012), Retrieval of aerosol optical depth and vertical distribution using O2 A- and B-band SCIAMACHY observations over Kanpur: a case study, Atmos. Meas. Tech, 5(5), 1099–1119. [Google Scholar]
  39. Satheesh SK, Moorthy KK, Babu SS, Vinoj V, and Dutt CBS (2008), Climate implications of large warming by elevated aerosol over India, Geophysical Research Letters, 35, L19809. [Google Scholar]
  40. Sears TM, Thomas GE, Carboni E, Smith AJA and Grainger RG (2013), SO2 as a possible proxy for volcanic ash in aviation hazard avoidance, J. Geophys. Res. Atmos, 118, 5698–570. [Google Scholar]
  41. Spurr R (2006), VLIDORT: A linearized pseudo-spherical vector discrete ordinate radiative transfer code for forward model and retrieval studies in multilayer multiple scattering media, Journal of Quantitative Spectroscopy & Radiative Transfer, 102, 316–342. [Google Scholar]
  42. Spurr R, and Christi M (2014), On the generation of atmospheric property Jacobians from the (V)LIDORT linearized radiative transfer models, Journal of Quantitative Spectroscopy and Radiative Transfer, 142(0), 109–115. [Google Scholar]
  43. Torres O, Levy RC, Remer L, Jethva H, and Wang J (2016), Aerosol particle properties using EPIC UV-VIS observations (talk), presented in DSCOVR Earth Science Team Meeting, NASA Goddard Space Flight Center, Greenbelt, MD, Oct 31 – Nov 1, 2016. [Google Scholar]
  44. Torres O, Bhartia PK, Herman JR, Ahmad Z, and Gleason J (1998), Derivation of aerosol properties from satellite measurements of backscattered ultraviolet radiation: Theoretical basis, J. Geophys. Res, 103(D14), 17099–17110. [Google Scholar]
  45. Wang J, Christopher SA, Reid JS, Maring H, Savoie D, Holben BH, Livingston JM, Russell PB, and Yang SK (2003), GOES-8 retrieval of dust aerosol optical thickness over the Atlantic Ocean during PRIDE, J. Geophys. Res, 108, 8595. [Google Scholar]
  46. Wang J, Xu X, Ding S, Zeng J, Spurr R, Liu X, Chance K, and Mishchenko M (2014), A numerical testbed for remote sensing of aerosols, and its demonstration for evaluating retrieval synergy from a geostationary satellite constellation of GEO-CAPE and GOES-R, Journal of Quantitative Spectroscopy and Radiative Transfer, 146(0), 510–528. [Google Scholar]
  47. Wendisch M, et al. (2008), Radiative and dynamic effects of absorbing aerosol particles over the Pearl River Delta, China, Atmospheric Environment, 42(25), 6405–6416. [Google Scholar]
  48. Winker DM, Vaughan MA, Omar A, Hu Y, Powell KA, Liu Z, Hunt WH, and Young SA (2009), Overview of the CALIPSO Mission and CALIOP Data Processing Algorithms, Journal of Atmospheric and Oceanic Technology, 26(11), 2310–2323. [Google Scholar]
  49. Xu X, Wang J, Zeng J, Spurr R, Liu X, Dubovik O, Li L, Li Z, Mishchenko MI, Siniuk A, and Holben BN (2015), Retrieval of aerosol microphysical properties from AERONET photopolarimetric measurements: 2. A new research algorithm and case demonstration, Journal of Geophysical Research: Atmospheres, 120, 7079–7098. [Google Scholar]
  50. Xu X, Wang J, Wang Y, and Kokhanovsky A (2017), Passive remote sensing of aerosol height, in Remote Sensing of Aerosols, Clouds, and Precipitation, edited by Islam T, Hu Y, Kokhanovsky A, and Wang J, Elsevier; DOI: 10.1016/B978-0-12-810437-8.00001-3. [DOI] [Google Scholar]
  51. Yang Y, Marshak A, Mao J, Lyapustin A, and Herman J (2013), A method of retrieving cloud top height and cloud geometrical thickness with oxygen A and B bands for the Deep Space Climate Observatory (DSCOVR) mission: Radiative transfer simulations, Journal of Quantitative Spectroscopy and Radiative Transfer, 122, 141–149. [Google Scholar]
  52. Young SA, and Vaughan MA (2009), The Retrieval of Profiles of Particulate Extinction from Cloud-Aerosol Lidar Infrared Pathfinder Satellite Observations (CALIPSO) Data: Algorithm Description, Journal of Atmospheric and Oceanic Technology, 26(6), 1105–1119. [Google Scholar]
  53. Yu H, et al. (2015), Quantification of trans-Atlantic dust transport from seven-year (2007–2013) record of CALIPSO lidar measurements, Remote Sensing of Environment, 159, 232–249. [Google Scholar]
  54. Yu H, Chin M, Winker DM, Omar AH, Liu Z, Kittaka C, and Diehl T (2010), Global view of aerosol vertical distributions from CALIPSO lidar measurements and GOCART simulations: Regional and seasonal variations, Journal of Geophysical Research, 115, D00H30. [Google Scholar]
  55. Zeng J, Han Q, and Wang J (2008), High-spectral resolution simulation of polarization of skylight: Sensitivity to aerosol vertical profile, Geophys. Res. Lett, 35(20), L20801. [Google Scholar]
  56. Zhang L, Li QB, Gu Y, Liou KN, and Meland B (2013), Dust vertical profile impact on global radiative forcing estimation using a coupled chemical-transport–radiative-transfer model, Atmospheric Chemistry and Physics, 13(14), 7097–7114, doi: 10.5194/acp-13-7097-2013. [DOI] [Google Scholar]
  57. Zhu A, Ramanathan V, Li F, and Kim D (2007), Dust plumes over the Pacific, Indian, and Atlantic oceans: Climatology and radiative impact, Journal of Geophysical Research: Atmospheres, 112, D16208. [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Information
NIHMS902885-supplement-Supplementary_Information.docx (2.5MB, docx)

RESOURCES