Successful Derivation of Absorbing Aerosol Index from the Environmental Trace Gases Monitoring Instrument (EMI)

Abstract

We retrieved the absorbing aerosol index (AAI) based on the measured reflectance from the Environmental Trace Gases Monitoring Instrument (EMI) for the first time. EMI is a push-broom spectrometer onboard the Chinese GeoFen-5 satellite launched on 9 May 2018, which was initially developed to determine the global distribution of atmospheric composition. The EMI initial AAI results were corrected from physical stripes and yielded an offset of 5.92 as calibration errors from a background value based on the statistical method that count the EMI AAI over the Pacific Ocean under cloudless scenes. We also evaluated the consistency of the EMI AAI and data with the TROPOspheric Monitoring Instrument (TROPOMI) observations. A comparison between the monthly average EMI AAI data and TROPOMI AAI revealed regional consistencies between these instruments with a similar spatial distribution of AAI (correlation coefficient, r > 0.9). The daily-scale results demonstrated that EMI was also consistent with TROPOMI AAI (r = 0.9). The spatial distribution of EMI AAI is consistent with Aerosol Optical Depth (AOD) from TROPOMI. The daily variation of EMI AAI in an Australian wildfire event was consistent with TROPOMI (r = 0.92). Overall, we demonstrated that EMI AAI can be efficiently used to detect large aerosol events for reconstructing the spatial variability of Ultraviolet (UV) absorbing aerosols.

1. Introduction

Suspended aerosols are the most common mixtures of solid and liquid particles in the atmosphere. The effect of tropospheric aerosols on the Earth’s climate is complicated [1]. On the one hand, aerosols directly affect the radiative balance of Earth by direct scattering and absorption of incoming solar and infrared radiation [2,3]. On the other hand, aerosols indirectly affect cloud optical properties and lifetime [2]. Spaceborne remote sensing has provided abundant, long-term records of aerosol datasets for the studies on aerosol climate and human health effects [2,4].

The Absorbing Aerosol Index (AAI) was introduced with the launch of the first total ozone mapping spectrometer (TOMS) instrument in 1978. The AAI is used extensively to provide important information on the transport of tropospheric aerosols of anthropogenic origin from industrial pollution and biomass burning [5,6], natural sources from desert dust [7,8], and volcanic ash [2,9]. All these aerosols can cause regional and global air pollution. Subsequent satellite missions have advanced our understanding of absorbing aerosols by providing the measurements of the backscattered radiation, which were applied to retrieve the absorbing aerosol index [9].

TOMS series instruments have been launched into space on various platforms since November 1978, providing daily global measurements of tropospheric absorbing aerosols [10,11]. Three EPS Metop satellites, Metop-A (launched 2006), Metop-B (launched 2012), and Metop-C (launched 2018), carried the Global Ozone Monitoring Experiment-2 (GOME-2) instruments, which were designed to measure key atmospheric constituents, such as aerosols [12,13]. The AAI product from the SCanning Imaging Absorption Spectrometer for Atmospheric CHartographY (SCIAMACHY) [14,15,16], ozone monitoring instrument (OMI) [17,18,19], and TROPOspheric Monitoring Instrument (TROPOMI) [20] observations stand out with accurate and detailed retrieval of spatio-temporal distributions of aerosols and have already significantly contributed to the continuity of the long-term AAI record [21]. The Chinese Environmental Trace Gases Monitoring Instrument (EMI) is a hyperspectral satellite-borne spectrometer, flying onboard the GaoFen-5 (GF-5), which can be used to retrieve trace gases [22,23]. The EMI also provides independent global coverage AAI records at a spatial resolution of 13 × 12 km², which confirms that AAI is applicable for detection and monitoring of absorbing aerosols.

In general, AAI is a useful parameter retrieved from spaceborne-based radiances. Notably, it can be retrieved both over land and ocean because it is insensitive to the surface type. However, AAI is peculiarly sensitive to the presence of UV-absorbing aerosols, such as smoke, mineral dust, and volcanic ash [2]. The modern spaceborne retrieval algorithms for most aerosol optical parameters rely on cloud screening before performing the aerosol retrieval. The AAI retrieval is advantageous because it can be applied to cloudy scenes [16,24]. The effects of clouds on the AAI are complex because clouds represent bright surfaces with a high surface albedo when they are located below the aerosol layer. A cloud generally reflects all incident radiation when clouds are located above (e.g., attenuate) an aerosol layer, which implies that AAI is determined by cloud characteristics [16,25].

The research aim of this study was to evaluate EMI AAI by using TROPOMI datasets, including AAI and Aerosol Optical Depth (AOD) from TROPOMI Level 2 products [20]. To this end, we (a) compared the monthly average EMI AAI for smoke and dust aerosols in August and October 2019, with the AAI from TROPOMI; (b) used AOD from TROPOMI and Aerosol Robotic Network (AERONET) to provide more information about the spatial distribution of aerosol EMI satellite measurements [26]; and (c) further validated the efficiency of EMI by analyzing wildfires in Australia during November 2019. This study utilized the AAI method in the presence of clouds [27,28,29]. The AAI are retrieved by using viewing geometries and radiance measurements of EMI with the radiative transfer model SCIATRAN [30]. Since some initial AAI results affected by calibration errors in the EMI irradiance measurements [15], we corrected the algorithm for the across-track variability. In particular, we alleviated vertical stripes by using the high-frequency terms in the Fourier transform. The correction factors were determined from background values to alleviate radiometric errors in the initial AAI results based on the statistical method. The study of the temporal variations and spatial distribution of the EMI and TROPOMI AAI have demonstrated the consistency in absorbing aerosol plume transportation. Overall, EMI can be used for detecting and characterizing smoke from biomass burning and other tropospheric absorbing particulates.

2. Data

2.1. EMI Data

The EMI is part of the payload of the Chinese GaoFen-5 satellite, which was launched into a polar orbit at an altitude of ~706 km on 9 May 2018 [22,31]. EMI is a high-resolution spectrometer with a spectral resolution of 0.3–0.5 nm, and a nadir spatial resolution of 12 km × 13 km. The EMI was designed to measure spectral radiance in UV and visible (VIS) regions by covering the spectral range from 240 to 710 nm [32]. The EMI AAI retrieval was used to retrieve the AAI from 340 nm/380 nm wavelengths in the UV2 (311–403 nm) channel. As shown in Figure 1, the swath of EMI is 2600 km, which gives the daily global coverage of both particulate pollution (aerosols) and gaseous pollution (ozone, SO₂, NO₂, and various other trace gases).

2.2. Auxiliary Data

The EMI-based AAI estimates were compared with those derived from the TROPOMI instruments to assess the quality of the former observations. The TROPOMI instrument [21] is a nadir viewing spectrometer on-board the Sentinel-5 Precursor (S5P) satellite, placed into a polar orbit on 13 October 2017. The equator crossing time of the S5P satellite was 13:30. The TROPOMI AER_AI AAI product [33] was derived by using the wavelength pair of 340/380 nm and 354/388 nm reflectance. It was derived from the Level 1B earth radiance and solar irradiance measurements [34], which have a spectral resolution of 0.5 nm and a nadir spatial resolution of 5.5 km × 3.5 km. The TROPOMI Level 2 data, including AOD, were accessed from https://s5phub.copernicus.eu/dhus/#/home (accessed on 1 August 2021).

AERONET [26] is a global ground-based remote-sensing aerosol monitoring network that provides long-term aerosol optical depth (AOD) estimates at eight wavelengths (340, 380, 440, 500, 670, 870, 940, and 1020 nm). The primary aim of AERONET is to study the temporal distribution of absorbing aerosols together with space-borne data. The station nearest to northern Africa is the Tamanrasset_INM (22.79°N, 5.53°E) (https://aeronet.gsfc.nasa.gov, last access: 10 March 2022). For this study, we used AOD at 380 nm from for absorbing aerosols over the contaminated region affected by the dust storm from the version 3 Level 1.5 product of AERONET.

3. EMI AAI Retrieval Algorithm

3.1. Calculation of the AAI

AAI was used to detect UV-absorbing aerosols in the atmosphere. We calculated the EMI-based AAI by using spaceborne reflectance at 340 nm/380 nm wavelength [1].

$$\mathrm{AAI}=-100\cdot {[\log }_{10}{(\frac{{\mathrm{R}}_{340}}{{\mathrm{R}}_{380}})}^{\mathrm{meas}}-{\log }_{10}{(\frac{{\mathrm{R}}_{340}}{{\mathrm{R}}_{380}})}^{\mathrm{Ray}}]$$

(1)

where ${\mathrm{R}}_{340}^{\mathrm{meas}}$ and ${\mathrm{R}}_{380}^{\mathrm{meas}}$ are the measured reflectance for the aerosol-contaminated atmosphere. ${\mathrm{R}}_{340}^{\mathrm{Ray}}$ and ${\mathrm{R}}_{380}^{\mathrm{Ray}}$ are the measured reflectance for the aerosol- and cloud-free atmosphere. According to the radiative transfer model, such an atmosphere only contains Rayleigh scattering, molecule absorption, Lambertian surface reflection, and absorption.

In the AAI retrieval, we simulated the Rayleigh reflectance by assuming that the atmosphere was bounded by a Lambertian surface with wavelength-independent surface albedo (A_sc), which was composed of surfaces, clouds, and aerosols. This assumption allowed for the calculation of surface albedo by implying that the simulated reflectance ${\mathrm{R}}_{380}^{\mathrm{Ray}}$ was equal to the measured reflectance ${\mathrm{R}}_{380}^{\mathrm{meas}}$ at 380 nm.

$${\mathrm{R}}_{380}^{\mathrm{meas}}{=\mathrm{R}}_{380}^{\mathrm{Ray}}({\mathrm{A}}_{\mathrm{sc}})$$

(2)

$${\mathrm{A}}_{\mathrm{sc}}=\frac{\mathrm{R}-{\mathrm{R}}_{\lambda }^0}{\mathrm{T}+{\mathrm{s}}^{*}(\mathrm{R}-{\mathrm{R}}_{\lambda }^0)}$$

(3)

where R is replaced by ${\mathrm{R}}_{380}^{\mathrm{meas}}$ in Equation (3). ${\mathrm{R}}_{\lambda }^0$ is the path reflectance, which is the atmospheric contribution to the reflectance. The parameter T is the total atmospheric transmission, and s* is the spherical albedo of the atmosphere. Equation (1) was therefore reduced to:

$$\mathrm{AAI}=-100\times {\log }_{10}\frac{{\mathrm{R}}_{340}^{\mathrm{meas}}}{{\mathrm{R}}_{340}^{\mathrm{Ray}}}$$

(4)

Besides demonstrating the AAI calculation, De Graaf et al. (2005) reported that AAI is also sensitive to cloud properties [25]. Due to this, we modeled AAI in the presence of clouds by using the Lambertian cloud model.

3.2. Calculation of the Contribution of a Cloud to the AAI

Fundamentally, clouds strongly affect AAI retrieval. In particular, the cloud-induced enhanced reflectance of TROPOMI measurements yielded a remarkable AAI value anomaly. From the AAI retrieval perspective, the assumption about the surface, clouds, and aerosols as Lambertian reflectors is essential. In the presence of clouds, the assumption of the retrieval has been demonstrated to be invalid because clouds reflect incident sunlight anisotropically [21]. Therefore, the cloudy atmosphere scene was described by two Lambertian reflectors in this study, including (1) the surface Lambertian reflector and (2) the cloud Lambertian reflector. We calculated the AAI of the cloudy pixels by using an independent pixel approximation (IPA) [27]. The simulated reflectance of each ground pixel was constructed from Equation (5):

$$\begin{array}{l}R(SZA,VZA,RAA,A_s,A_c,h_s,h_c)=c\ast R^{cloud}(SZA,VZA,RAA,A_c,h_c)\\ +(1-c)\ast R^{clear}(SZA,VZA,RAA,A_s,h_s)\end{array}$$

(5)

where c is the cloud fraction, and SZA, VZA, RAA are the solar zenith angle and viewing zenith angle and the relative azimuth angle, respectively; A_s is the land surface albedo, h_s is the land surface altitude, h_c is the cloud height, A_c is the cloud albedo fixed at 0.8, R is the TOA (top of the atmosphere) reflectance, and R_cloud and R_clear are cloudy and clear sky reflectances, respectively.

The cloud fraction is defined as the measured and modeled reflectance by Equation (6):

$$c=\frac{R_{380}^{meas}-R_{380}^{clear}}{R_{380}^{cloud}-R_{380}^{clear}}$$

(6)

The reflectance at 340 nm was calculated from Equation (5). We derived the AAI by modeling clouds and using the above Lambertian cloud parameterization.

3.3. Look-Up Tables

EMI-based AAI was determined from the measured reflectances at 340 and 380 nm by using SCIATRAN. SCIATRAN calculates the TOA reflectance at 340 and 380 nm by using the scalar discrete ordinate technique in a spherical atmosphere, combined with a preconstructed lookup table (LUT). The simulated reflectance LUTs were established as a function of the SZA, VZA, RAA, surface altitude, and cloud height. The nodes of the LUTs entries are summarized in

Table 1. Parameter node settings of the look-up table (LUT).

Parameter	Node
SZA (°)	0.1, 10.0, 20.0, 30.68, 40.54, 45.57, 50.21, 55.94, 60.0, 65.17, 70.12, 72.54, 74.93, 76.11, 80.79, 84.26
VZA (°)	0.1, 10.0, 20.0, 30.68, 40.54, 45.57, 50.21, 55.94, 60.0, 65.17, 70.12
RAA (°)	0, 30, 60, 90, 120, 150, 180
Surface altitude/Cloud height (km)	0, 0.2, 0.4, 0.8, 1.2, 1.6, 2.0, 2.4, 2.8, 3.2, 3.6, 4.0, 4.6, 5.0, 5.6, 6.2, 7.0, 8.0, 9.0, 10.0, 12.0, 14.0

. For each EMI pixel, the surface altitude was determined from the Global Multi-resolution Terrain Elevation Data 2010 (GMTED2010), which was interpolated according to latitude and longitude. The cloud height of each satellite pixel was determined by using the grid with 0.25° × 0.50° (latitude × longitude) spatial resolution for the TROPOMI cloud product. In the AAI calculation, the spherical albedo s* and atmospheric transmission T were calculated by using multidimensional linear interpolation over the SZA, VZA, RAA, and the surface altitude.

3.4. Correction of AAI Errors

The across-track variability is manifested in so-called “stripes” that emerge in many atmospheric spectrometers, such as OMI, when atmospheric trace gases are retrieved from a push-broom [35]. This variability is also reflected in the AAI retrieval from the corrected reflectance because of the calibration errors in the EMI irradiance measurements. We utilized the correction method based on the Boersma’s method to remove these stripes, which generally affect the identification of absorbing aerosols. The procedure is described below.

• We selected the window (50 along track × 191 across track pixels in size) with the minimum variance in AAI to ensure that the region did not entail any aerosol absorbing pollution.
• We calculated the average AAI value of 191 ground pixels in the scanline direction.
• We performed the Fourier analysis to determine the low and high frequencies of 191 average values. The high-frequency signals were interpreted as stripe noise, which can be subtracted from the initial AAI value along the track.

AAI is a complex aerosol optical parameter that is highly sensitive to reflectance. The small variations in reflectance cause large offsets in the AAI, making it sensitive to calibration errors in reflectance [15]. In this study, we calculated the background value of the EMI AAI over the Pacific Ocean under cloudless scenarios to correct the calibration errors of the instrument, reflected in the initial EMI AAI retrieval results. We utilized the statistical method to determine the background value by filtering out cloudy pixels, sun glint, or solar eclipse pixels over the Pacific Ocean without absorbing aerosols.

Owing to the limitation of EMI data, we selected 10 orbits over the Pacific from 29 October to 7 November 2019. From 8 November, the Pacific Ocean was affected by the absorption of aerosols from an Australian wildfire (https://earthobservatory.nasa.gov/images/145859/early-season-fires-burning-in-new-south-wales (accessed on 1 October 2021). The background values, as a function of ground pixel, are shown in Figure 2. After the “stripe” correction, we retrieved a fixed correction factor of 5.92 by averaging the background value. The positive offset of the AAI can be applied for the EMI AAI results. An example of the correction is given in Figure 3, which is an EMI observation over Australia on 5 November 2019. Figure 3b shows the distribution of AAI after correction. The effect of radiometric calibration errors on the AAI is reduced after the “stripe” and background value correction. Notably, the distribution of absorbing aerosols becomes more identifiable. The background value is also sensitive to the reflectances. The multiplication factor of reflectance at 340 was 1.146 after transforming from the fixed offset of the AAI by using Equation (4). The reflectance measured by EMI was underestimated in the UV region.

4. Results and Discussion

4.1. Comparison with TROPOMI AAI

Figure 4 shows the global monthly mean AAI values from EMI and TROPOMI for August and October 2019, respectively. The global regions were divided into 0.25° latitude × 0.50° longitude grid cells. Figure 4 does not display the AAI values at the latitudes of >40°S because AAI depends on a higher viewing geometry [25], and the incident light was weak at high southern latitudes in August. The spatial distributions of AAI exhibited similar patterns between EMI and TROPOMI. The mean values of global EMI AAI were 9.6% and 24% lower than TROPOMI in August and October, respectively. The differences between the two instruments may be due to the lack of data for EMI. Another important reason is the defective correction method of initial EMI AAI results.

Both instruments identified absorbing aerosol plumes over the polluted regions, such as Saudi Arabia, Iran, the North Africa, and Southern Africa. Clear absorption was more apparent over the Sahara, Iranian, Southern African, and Thar regions, where AAI > 1 in August. There are many biomass burning activities in Southern Africa [10]. The mean value of EMI AAI in August was 65% higher than that in October in Southern Africa, corresponding with the conclusion that the absorption of aerosols from biomass burning regions was associated with the seasonal cycle of agricultural burning [36].

To further analyze the discrepancies between EMI and TROPOMI AAI, we considered four regions: Northern Africa (20W–20E, 15N–30N), Taklimakan (75E–90E, 36N–42N), Southern Africa (10W–15E, 20S–0), and the Middle East (38E–60E, 12N–30N) based on the monthly averaged AAI values for August and October 2019. In these regions, dust and biomass burning aerosols are denoted by the red rectangle in Figure 4c according to their locations. The red points denote the mean AAI in these regions, as shown in Figure 5. EMI underestimated the AAI value compared with TROPOMI AAI over the four regions, which could be due to the differences in emissions, meteorological conditions, and instrumental design for AAI retrieval. Notably, EMI AAI still captured the mineral dust and biomass burning signals for both months well.

Table 2. Summary of statistical information for the selected stations.

Monthly	August		October
Region	r	RMSE	r	RMSE	N
Northern Africa	0.92	0.67	0.92	0.66	4800
Southern Africa	0.93	0.43	0.55	0.65	4000
Middle East	0.91	0.61	0.92	0.64	3168
Taklimakan	0.94	0.47	0.90	0.27	720

presents the correlation coefficients (r), root mean square errors (RMSE), and the number of points (N). The statistical values were derived based on the monthly averaged AAI values for August and October 2019. Besides Southern Africa and the regions where low values of AAI were identified in October 2019, the EMI and TROPOMI AAI demonstrated good agreements (r = 0.90–0.94) for all the analyzed regions. The low correlation (r = 0.55) in Southern Africa may be attributed to the small dynamic range of the negative AAI value. The RMSE was only acceptable and indicated that a segment of the EMI plume could not be fit with TROPOMI plume.

To evaluate the capability of the EMI-based aerosol monitoring, a map of AAI over the Sahara region from EMI and TROPOMI on 6 August 2019 (Figure 6) was drawn. This date was selected because an intense dust storm occurred on the same day. The spatial distributions of AAI from EMI and TROPOMI were consistent, particularly over regions where high values of AAI are visible in the EMI results. Overall, EMI is capable of providing clear spatial detail of absorbing aerosols in desert areas.

This region with high values of AAI were divided into 500 boxes in 1° × 1°grid cell. The EMI AAI values agreed well with the TROPOMI measurements (r = 0.9), as shown in Figure 7. The regional averaged AAI values of EMI and TROPOMI were 1.37 and 1.32, respectively; the EMI AAI value was 3.8% higher than that of TROPOMI. The value of EMI with some faint stripes was smaller than that of TROPOMI over the region of high values (0°–3°W, 23°N–25°N). The defective correction factors for the EMI AAI offset may have triggered this adverse effect.

4.2. EMI versus AOD Measurements

The AAI is combined with AOD to study the spatial distribution of absorbing aerosols. To this end, we selected AOD data from TROPOMI as the auxiliary reference to compare the spatial monitoring of EMI AAI. The subtropical Atlantic Ocean was investigated for our domain of comparison. The retrieval of space-borne AOD data are limited by the presence of clouds. This means that a lot of data are missing over high polluted areas.

Figure 8 shows similar spatial distribution of EMI AAI and TROPOMI AOD measurements on 26 August 2019. Both show that the dust plumes were transported from the sources of Sahara towards the North Atlantic Ocean. The presence of clouds leads to a greater reduction in the availability of the TROPOMI AOD. Unlike satellite-based AOD data, EMI AAI, as a qualitative index, can be calculated in the presence of clouds to realize daily global monitoring. This is ideal for tracking the evolution of aerosol plumes from dust storm, volcanic ash, and biomass burning.

The AAI from EMI and TROPOMI can be used together with the AOD measurements at 380 nm from AERONET to better study the temporal distribution of absorbing aerosol. The daily averaged AAI values from the EMI (blue dots and lines) and TROPOMI (black dots and lines) are shown in Figure 9. The AOD measurements were the daily averaged value within the 1.0° × 0.5° (lat × lon) grid around the AERONET site (red dots and lines) for both months. A single AERONET station at Tamanrasset_INM (22.79°N, 5.53°E) was selected for this purpose to study the dust aerosols in Northern Africa.

Figure 9 shows that the trend of daily average EMI AAI agreed with TROPOMI AAI, where high AAI values corresponded to the days with high AOD on the 4 and 14 August 2019. The difference in the trend between the AOD and AAI was identified on 10 and 11 October. The values of EMI AAI are negative, but AOD is high value. This may be due to the station being affected by other aerosol types, including weakly absorbing aerosols such as industrial sulfates. The satellite-based AAI can be used in conjunction with the ground-based AOD measurements to give more information about aerosols.

4.3. Case Study

To further evaluate the capabilities of EMI, we investigated the daily transportation of absorbing aerosols over Australia. There were intense bushfire events in Australia in November, which lasted for several days. Figure 10 illustrates a satellite map from the Visible Infrared Imaging Radiometer Suite (VIIRS) on the NOAA-NASA Suomi NPP satellite on 8 November 2019. As seen, there were fires near the coast, with thick smoke transported by air masses over the Pacific.

The TROPOMI and EMI observations of the aerosol plumes from 8 to 16 November 2019 are shown in Figure 11 and Figure 12. As seen, the spatial distribution of the EMI AAI clearly exhibited the similar patterns to those of the TROPOMI measurements. The thick smoke produced by wildfires in New South Wales was exacerbated, thereby increasing the amount of emissions into the atmosphere. The flames were fanned by strong westerly winds from the continent towards the south-latitude region of the Pacific Ocean. The regions of several hundred kilometers were embraced by the resulting products from biomass burning, which significantly degraded the local air quality. Figure 13 shows a comparison of regions with high AAI values from 9 to 12 November. The regions were delimited in the 0.5° × 0.5° grid cell for high AAI values on 9 November (30°S~50°S, 160°E~180°E), 10 (40°S~60°S, 145°W~165°W), 11 (30°S~50°S, 145°W~165°W), 12 (20°S~40°S, 140°W~160°W). As seen, the average correlation coefficient between the EMI and TROPOMI was 0.92, signifying a very good agreement between the instruments.

5. Discussion and Conclusions

In this study, we introduced the EMI absorbing aerosol index (AAI) retrieval algorithm for the first time. The algorithm was based on reflectance, defined by Earth radiance and solar irradiance measurements at 340 and 380 nm (UV2 band). We first validated the EMI AAI results against the TROPOMI AAI information.

The comparison between the monthly mean EMI and TROPOMI AAI revealed similar distributions of absorbing aerosols. The difference between the EMI and TROPOMI mean AAI values was <24%. Our analysis demonstrated that the monthly mean AAI was characterized by a distinct seasonal cycle in the primary contaminated regions, with higher values during the summer season (August) and lower values during the autumn season (October). To further validate the EMI AAI retrievals, we compared EMI and TROPOMI AAI products (r = 0.9) on a single measurement with the identical spatial resolution of 1° × 1° (lat × lon) over the Sahara Desert on 6 August 2019. The spatial distribution AAI retrieved from EMI and TROPOMI AOD looked consistent. The AOD observed from Aerosol Robotic Network were used in conjunction with AAI to give more information about the temporal distribution of absorbing aerosols.

We also provided a detailed analysis of an extreme wildfire event, which delivered large black carbon loads from the Australian continent to the Pacific Ocean from 8 to 16 November 2019. The comparisons of the smoke plumes revealed the same spatial variations and distributions between EMI and TROPOMI. The quantitative agreement between the two instruments was high (r = 0.92) over the areas for four days with high AAI pollution.

To conclude, the AAI retrieval algorithm of the EMI instrument can be used for monitoring large absorbing aerosol events. Future works should deepen our understanding of EMI-based AAI because our study has not covered some important aspects of AAI retrieval. For instance, we utilized simple assumptions on cloud optical parameters to retrieve AAI in the presence of clouds. Moreover, the effects of ozone absorption on AAI were not used in the auxiliary information on EMI AAI retrieval. Despite these shortcomings, EMI is very promising for characterizing aerosols in detail. We clearly demonstrated that EMI-based radiance can be utilized to calculate AAI, thereby revealing spatio-temporal distributions of absorbing aerosols. We suggest that the AAI can also be retrieved for EMI-2 on the GaoFen-5 (02) satellites, thereby laying the foundation for establishing AAI products. Such products can be efficiently applied for long-term aerosol observations in combination with ground-based data.