Total Ozone Columns from the Environmental Trace Gases Monitoring Instrument (EMI) Using the DOAS Method

: Global measurements of total ozone are necessary to evaluate ozone hole recovery above Antarctica. The Environmental Trace Gases Monitoring Instrument (EMI) onboard GaoFen 5, launched in May 2018, was developed to measure and monitor the global total ozone column (TOC) and distributions of other trace gases. In this study, some of the ﬁrst global TOC results of the EMI using the differential optical absorption spectroscopy (DOAS) method and validation with ground-based TOC measurements and data derived from Ozone Monitoring Instrument (OMI) and TROPOspheric Monitoring Instrument (TROPOMI) observations are presented. Results show that monthly average EMI TOC data had a similar spatial distribution and a high correlation coefﬁcient (R ≥ 0.99) with both OMI and TROPOMI TOC. Comparisons with ground-based measurements from the World Ozone and Ultraviolet Radiation Data Centre also revealed strong correlations (R > 0.9). Continuous zenith sky measurements from zenith scattered light differential optical absorption spectroscopy instruments in Antarctica were also used for validation (R = 0.9). The EMI-derived observations were able to account for the rapid change in TOC associated with the sudden stratospheric warming event in October 2019; monthly average TOC in October 2019 was 45% higher compared to October 2018. These results indicate that EMI TOC derived using the DOAS method is reliable and has the potential to be used for global TOC monitoring.


Introduction
Stratospheric ozone, distributed at an altitude of approximately 20-35 km, plays an important role in protecting human health and the Earth's ecological balance [1,2] by providing a shield against strong UV radiation (200-300 nm) [3]. Stratospheric ozone is involved in numerous photochemical reactions, especially the activation of Br and Cl, which can destroy the ozone layer [4,5]. The hole in the ozone layer was first found by Farman in the Argentine Islands (65 • S, 64 • W) and Halley Bay (76 • S, 27 • W) [6]. Chlorofluorocarbons (CFCs), a leading cause of ozone depletion, were phased out by the Montreal Protocol [7], and in response, the total ozone column (TOC) above Antarctica is recovering and the ozone hole is reducing [8][9][10].
In the past 30 years, satellite observations have been widely applied to retrieve the TOC by measuring backscattered light using the nadir viewing method [11][12][13]. Satelliteborne instruments with a high spatial resolution can obtain more accurate information on regional trace gases, which have facilitated research in atmospheric chemistry, including ozone hole monitoring and the analysis of stratospheric dynamics [11]. The Global Ozone Monitoring Experiment (GOME) onboard the European Remote Sensing-2 satellite was 2. Data

EMI Data
The GF-5 satellite, at an altitude of 705 km, is part of the Chinese high-resolution Earth observation system and has an ascending equator cross time of 13:30 local standard time (LST) [22,23]. The ultraviolet (UV) and visible (VIS) wavelength bands for the EMI are UV1 (240-315 nm), UV2 (311-403 nm), VIS1 (401-550 nm), and VIS2 (545-710 nm) [24]. In this study, we used the UV2 band to retrieve EMI TOC. The wide field of view (FOV) of the EMI (114 • ) ensures daily global coverage. Each channel has two-dimensional charge-coupled device (CCD) detectors, with a spectral resolution of 0.3-0.5 nm. The EMI scanning method is shown in Figure 1.

Satellite Data
The OMI is onboard the EOS Aura satellite, launched on 15 July 2004, with an overpass time of 13:30 LST. The OMI FOV can reach 114°, providing daily global coverage, and TOC can be retrieved using both UV (270-380 nm) and VIS (350-500 nm) wavelengths [25]. The nadir spatial resolution of OMI is 13 × 14 km 2 , with a spectral resolution of 0.5 nm, and TOC data are available at https://disc.gsfc.nasa.gov/datasets/OMDOAO3_003/summary (accessed on January 1, 2021).

Satellite Data
The OMI is onboard the EOS Aura satellite, launched on 15 July 2004, with an overpass time of 13:30 LST. The OMI FOV can reach 114 • , providing daily global coverage, and TOC can be retrieved using both UV (270-380 nm) and VIS (350-500 nm) wavelengths [25]. The nadir spatial resolution of OMI is 13 × 14 km 2 , with a spectral resolution of 0.5 nm, and TOC data are available at https://disc.gsfc.nasa.gov/datasets/OMDOAO3_003/summary (accessed on 1 January 2021).
TROPOMI is onboard the Copernicus Sentinel-5 Precursor (S5P) satellite, launched in October 2017, with an overpass time of 13:30 LST. TROPOMI has a nadir pixel size of 3.5 × 5.5 km 2 [13,26], daily coverage with approximately 14 orbits per day, and a spectral resolution of 0.25-1 nm. Spectral bands cover UV-Vis (270-550 nm), NIR (675-775 nm), and SWIR (2305-2385 nm) [19], and TOC data are available at https://s5phub.copernicus.eu/ dhus/#/home (accessed on 7 January 2021). The detailed instrument performances of EMI, OMI, and TROPOMI are shown in Table 1.  [27] and we used TOC measurements from Brewer and Dobson in this study due to their reliability. Furthermore, ZSL-DOAS observations (http://www.ndaccdemo.org/, accessed on 26 December 2020) from the Système d'Analysepar Observation Zénitale (SAOZ) network were also used for validation. Two TOC values in the morning and afternoon can be retrieved using the ZSL-DOAS method from the GB measurements of the SAOZ network.

SCD Retrieval
The DOAS method, based on the Lambert-Beer law, retrieves the concentrations of interest trace gases through their characteristic absorptions and the measured intensity. From the Lambert-Beer law and derivation: ln where I*(λ) means the measured intensity, I o (λ) means the original luminous intensity, σ * i means the cross section of trace gas i, c i means the averaged concentration of trace gas i, L means the length of optical path, SCD i means the slant columns density of trace gas i, ln I o (λ) means the differential optical density, and λ means the wavelength. The SCDs of desired trace gases can be retrieved by least-squares fitting from Equation (1).
Ozone SCDs were retrieved using QDOAS software developed by the Royal Belgian Institute for Space Aeronomy (BIRA-IASB) (http://uv-vis.aeronomie.be/software/QDOAS/, accessed on 20 October 2020), with a retrieval wavelength range of 313-320 nm. Solar irradiance measured on 12 June 2018 was used as the reference spectrum. We note that a fitting window of 325-335 nm is more appropriate for ozone SCDs; however, owing to the limitations of the instrument and reference spectrum, the results of ozone SCDs in the 325-335 nm fitting window were not satisfactory. Therefore, we chose the 313-320 nm fitting window to retrieve ozone SCDs from the EMI. This lower wavelength may lead to larger AMF biases when the SZA is large.
O 3 , NO 2 , HCHO, BrO, SO 2 , and ring cross-sections (the rotational Raman scattering effect, calculated by Ring.exe of QDOAS) were considered in the retrieval algorithm, and the detailed parameters are listed in Table 2. From the Algorithm Theoretical Basis Document (ATBD) of TROPOMI ozone product (https://sentinels.copernicus.eu/documents/ 247904/2476257/Sentinel-5P-TROPOMI-ATBD-Total-Ozone, accessed on 18 May 2021), a difference of 1-2% appeared when using the cross-sections with different temperatures.  picted the reference spectrum while the black curve represents the measured spectrum. Panels (b)-(h) show the differential cross-sections (black curves) included in the analysis and fitting functions (in red). Panel (i) depicts the residual of the DOAS fitting. From Figure 2, the retrieved ozone SCD and SCD error were 1.515 × 10 19 molecule/cm 2 and 1.334 × 10 17 molecule/cm 2 , respectively, with the root mean square (RMS) of the spectral fitting residual of 1.909 × 10 −3 . Therefore, the relative SCD error, calculated by SCD error /SCD, was 0.88%. ens. 2021, 13, x FOR PEER REVIEW 5 of 17 Figure 2 shows an example of the fitting results carried out with the QDOAS tool for one orbit (orbit number 002590) on 2 November 2018. For panel (a), in red is depicted the reference spectrum while the black curve represents the measured spectrum. Panels (b)-(h) show the differential cross-sections (black curves) included in the analysis and fitting functions (in red). Panel (i) depicts the residual of the DOAS fitting. From Figure 2, the retrieved ozone SCD and SCD error were 1.515 × 10 19 molecule/cm 2 and 1.334 × 10 17 molecule/cm 2 , respectively, with the root mean square (RMS) of the spectral fitting residual of 1.909 × 10 −3 . Therefore, the relative SCD error, calculated by SCDerror/SCD, was 0.88%.

AMF Retrieval
The AMF, calculated by the SCIATRAN radiative transfer model, was used to convert the SCD to VCD. In this study, the AMF was determined using multidimensional linear interpolation from a precalculated AMF lookup table (LUT), and 316.3 nm was used to establish the AMF LUT. The parameter nodes for the establishment of the AMF LUT are listed in Table 3. In this study, column-and latitude-dependent profiles were used as a priori profiles to reduce the uncertainty of the AMF calculation caused by variations in ozone profiles. The a priori profiles were obtained from the TOMS V8 climatology [33]. Figure 3 shows the dependence of AMF on the ozone profiles with different VCDs. It is apparent that ozone profiles have a large influence on the AMF calculation, especially when the SZA is large-ozone profiles can cause 20% differences in the AMF when the SZA is 75 • . Hence, the appropriate ozone profiles should be selected using precalculated VCDs to reduce the uncertainty of the AMF calculation. The AMF of the EMI TOC was determined using a two-step AMF calculation method. First, the precalculated VCDs were obtained through SCDs and AMFs, which were determined using multidimensional linear interpolation from the precalculated rough AMF LUT (without the parameter node of VCD), and the ozone profiles used for rough AMF LUT were selected by month and latitude from the SCIATRAN database. Then, the accurate AMFs were calculated by multidimensional linear interpolation from the precalculated accurate AMF LUT. The VCD, SZA, relative azimuth angle (RAA), VZA, surface albedo, latitude, month, and cloud pressure were the parameter nodes of the accurate AMF LUT. To reduce uncertainties in the AMF calculations, we only used data with SZAs less than 82 • in this study. The surface albedo was obtained from the OMLER monthly climatology derived from several years of OMI observations, with a resolution of 0.5 • × 0.5 • (lat × lon) [34]. The effects of clouds, which are related to the calculation of the AMF, should be considered when a ground pixel is covered by clouds. The cloud information (cloud pressure and cloud fraction) used for the AMF calculation of EMI TOC was obtained from the TROPOMI cloud product (S5P_L2_CLOUD_1 and S5P_L2_CLOUD_HiR1), with a resampled spatial resolution of 0.25 • × 0.25 • (lat × lon). The independent pixel approximation method was used to calculate the AMF by splitting the partly cloudy ground pixels into a weighted sum for clear (M clear ) and cloudy conditions (M cloud ): where w denotes the weight, namely the effective cloud fraction obtained from TROPOMI. tude from the SCIATRAN database. Then, the accurate AMFs were calculated by multidimensional linear interpolation from the precalculated accurate AMF LUT. The VCD, SZA, relative azimuth angle (RAA), VZA, surface albedo, latitude, month, and cloud pressure were the parameter nodes of the accurate AMF LUT. To reduce uncertainties in the AMF calculations, we only used data with SZAs less than 82° in this study. The surface albedo was obtained from the OMLER monthly climatology derived from several years of OMI observations, with a resolution of 0.5° × 0.5° (lat × lon) [34].

De-Stripe
Obvious stripes, due to irradiance calibration error, in the EMI TOC retrieval (Figure 4a) needed to be removed before validation, and we used spatial filtering following the Fourier transform method to remove the stripes following Boersma's method [35]. Taking the TOC retrieval of one orbit on 2 November 2018 as an example, the steps are as follows:  The TOC following de-striping is shown in Figure 4b.

Error Analysis
In this study, we did not consider systematic errors because they have a limited influence on trends [11]. The relative SCD error of the EMI TOC retrieval algorithm was less The TOC following de-striping is shown in Figure 4b.

Error Analysis
In this study, we did not consider systematic errors because they have a limited influence on trends [11]. The relative SCD error of the EMI TOC retrieval algorithm was less than 2%. The relative error of a priori ozone profiles on the AMF, which have a large influence on AMF calculations, were evaluated through extensive comparisons with a year of ozonesonde data from the WOUDC database on a global scale. The detailed error sources and relative errors are listed in Table 4. The total error of the EMI TOC, calculated using E VCD = E 2 SCD + E 2 albedo + E 2 aerosol + E 2 cloud_pressure + E 2 cloud_fraction + E 2 profile , was less than 4.5% (excluding systematic errors).

EMI Versus OMI and TROPOMI
Global TOCs in October 2018 from EMI, TROPOMI, and OMI are shown in Figure 5a,c,e, at a resampled spatial resolution of 0.25 • × 0.5 • (lat × lon). Figure 5b shows that the relative SCD error was less than 2%. The relative differences in EMI from TROPOMI and OMI, calculated using 100% × VCD TROPOMI −VCD EMI VCD EMI and 100% × VCD OMI −VCD EMI VCD EMI , are shown in Figure  5d,f. It is apparent that global TOC retrieved from EMI shows a similar spatial distribution to OMI and TROPOMI. High ozone values are concentrated between 30-60 • S, and low values are apparent in the high-latitude region of Antarctica. The relative differences between EMI and TROPOMI ranged from 0.27-0.29% (at a 99% confidence interval), with an average absolute difference of 0.28%, which was calculated using 100% × 1 N ∑ VCD TROPOMI −VCD EMI VCD EMI . As shown in Figure 5d,f, EMI TOC shows a 10% difference compared with TROPOMI and OMI in some high-latitude areas, where the SZAs are large. As mentioned in Section 3.1, larger biases are likely in the AMF calculation when the SZA is large, and we attribute the large biases of EMI TOC in some high-latitude areas to this. Furthermore, the distribution of snow and ice may influence the retrieval of TOC [12], and we will consider these parameters in future research. Additional results for global TOCs from EMI, OMI, and TROPOMI and their relative differences are provided in the Appendix A ( Figures A1 and A2). Our results show that the EMI TOC is highly consistent with the OMI and TROPOMI TOCs.
The linear fitting of global TOC between EMI and OMI and TROPOMI in October 2018 are shown in Figure 6a,b. Regression analysis shows that the EMI TOC has a good correlation with both OMI and TROPOMI, with Pearson's correlation coefficients (R) of 0.99 for both. The root mean square errors (RMSEs) between EMI TOC and OMI and TROPOMI TOCs, calculated using ∑(VCDOMI−VCDEMI) large. As mentioned in Section 3.1, larger biases are likely in the AMF calculation when the SZA is large, and we attribute the large biases of EMI TOC in some high-latitude areas to this. Furthermore, the distribution of snow and ice may influence the retrieval of TOC [12], and we will consider these parameters in future research. Additional results for global TOCs from EMI, OMI, and TROPOMI and their relative differences are provided in the Appendix (Figures A1 and A2). Our results show that the EMI TOC is highly consistent with the OMI and TROPOMI TOCs. The linear fitting of global TOC between EMI and OMI and TROPOMI in October 2018 are shown in Figure 6a,b. Regression analysis shows that the EMI TOC has a good correlation with both OMI and TROPOMI, with Pearson's correlation coefficients (R) of 0.99 for both. The root mean square errors (RMSEs) between EMI TOC and OMI and TRO-POMI TOCs, calculated using ∑ and ∑ , were 9.9 and 8.4 DU, respectively. The root mean square relative errors (RMSREs), calculated using 100% × ∑ and 100% × ∑ , were 3.6 and 3.2%, respectively.

EMI Versus GB Measurements
Seven GB measurements from the WOUDC database were used to validate the EMI TOC, and a summary of these stations is shown in Table 5. Daily average TOCs with error bars from the EMI and GB measurements are shown in Figure 7. The EMI TOC used for comparison with GB measurements was the daily average value within the ground pixel (1° × 1°) of the corresponding GB station. The average absolute difference (at a 95% confidence interval) between the EMI and GB measurements at these stations was less than 5.5% (Table 5). Figure 7 shows that the daily average EMI TOC agrees well with the GB measurements; however, the EMI TOC is higher than the GB measurements when TOC was located at a high level. The regression analyses of EMI TOC with measurements from the GB stations are shown in Figure 8. The R is greater than 0.92 at all stations.

EMI Versus GB Measurements
Seven GB measurements from the WOUDC database were used to validate the EMI TOC, and a summary of these stations is shown in Table 5. Daily average TOCs with error bars from the EMI and GB measurements are shown in Figure 7. The EMI TOC used for comparison with GB measurements was the daily average value within the ground pixel (1 • × 1 • ) of the corresponding GB station. The average absolute difference (at a 95% confidence interval) between the EMI and GB measurements at these stations was less than 5.5% (Table 5). Figure 7 shows that the daily average EMI TOC agrees well with the GB measurements; however, the EMI TOC is higher than the GB measurements when TOC was located at a high level. The regression analyses of EMI TOC with measurements from the GB stations are shown in Figure 8. The R is greater than 0.92 at all stations.

Application Case
We compared daily TOCs from EMI observations with those from ground-based ZSL-DOAS measurements from the polar vortex edge region (Chinese Great Wall station; 62.22° S, 58.96° W) of Antarctica, where rapid changes and large fluctuations in TOC have been detected [36]. Daily average EMI TOCs and ZSL-DOAS TOCs from November 2018 to November 2019 are shown in Figure 9a. The blue line located at 220 DU (1 DU = 2.69 × 10 16 molecule/cm 2 ) denotes the threshold for the ozone hole [37]. The average absolute difference (at a 95% confidence interval) was 5.33%. The ozone hole above the Great Wall Station was monitored simultaneously using both EMI and ZSL-DOAS measurements during the observation period. The daily averaged TOC from EMI is consistent with ZSL-DOAS, with an R of 0.90.

Application Case
We compared daily TOCs from EMI observations with those from ground-based ZSL-DOAS measurements from the polar vortex edge region (Chinese Great Wall station; 62.22 • S, 58.96 • W) of Antarctica, where rapid changes and large fluctuations in TOC have been detected [36]. Daily average EMI TOCs and ZSL-DOAS TOCs from November 2018 to November 2019 are shown in Figure 9a. The blue line located at 220 DU (1 DU = 2.69 × 10 16 molecule/cm 2 ) denotes the threshold for the ozone hole [37]. The average absolute difference (at a 95% confidence interval) was 5.33%. The ozone hole above the Great Wall Station was monitored simultaneously using both EMI and ZSL-DOAS measurements during the observation period. The daily averaged TOC from EMI is consistent with ZSL-DOAS, with an R of 0.90. In 2019, a sudden stratospheric warming (SSW) event in the Southern Hemisphere increased the average October TOC by 28% compared to the October average over the previous 11 years [38]. Monthly averaged Antarctic TOCs from EMI in October 2018 and 2019 are shown in Figure 10. The ozone hole is smaller in 2019 compared to 2018, and the monthly average October 2019 TOC is 45% higher compared to October 2018 in the inland area (60-90° S), corresponding with the 2019 SSW event. These two examples show that EMI has the capacity to monitor long-term trends in regional TOC and the Antarctic ozone hole.

Conclusions and Discussions
In this study, we presented some of the first EMI TOC results using the DOAS method and compared them with TOCs derived from OMI and TROPOMI at a resampled spatial resolution of 0.25° × 0.5° (lat × lon). The EMI TOC showed a similar spatial distri- In 2019, a sudden stratospheric warming (SSW) event in the Southern Hemisphere increased the average October TOC by 28% compared to the October average over the previous 11 years [38]. Monthly averaged Antarctic TOCs from EMI in October 2018 and 2019 are shown in Figure 10. The ozone hole is smaller in 2019 compared to 2018, and the monthly average October 2019 TOC is 45% higher compared to October 2018 in the inland area (60-90 • S), corresponding with the 2019 SSW event. These two examples show that EMI has the capacity to monitor long-term trends in regional TOC and the Antarctic ozone hole. In 2019, a sudden stratospheric warming (SSW) event in the Southern Hemisphere increased the average October TOC by 28% compared to the October average over the previous 11 years [38]. Monthly averaged Antarctic TOCs from EMI in October 2018 and 2019 are shown in Figure 10. The ozone hole is smaller in 2019 compared to 2018, and the monthly average October 2019 TOC is 45% higher compared to October 2018 in the inland area (60-90° S), corresponding with the 2019 SSW event. These two examples show that EMI has the capacity to monitor long-term trends in regional TOC and the Antarctic ozone hole.

Conclusions and Discussions
In this study, we presented some of the first EMI TOC results using the DOAS method and compared them with TOCs derived from OMI and TROPOMI at a resampled spatial resolution of 0.25° × 0.5° (lat × lon). The EMI TOC showed a similar spatial distribution to both OMI and TROPOMI, with an R of 0.99 for both. Daily GB measurements of TOC from seven monitoring stations were also used to validate the EMI TOC and stations

Conclusions and Discussions
In this study, we presented some of the first EMI TOC results using the DOAS method and compared them with TOCs derived from OMI and TROPOMI at a resampled spatial resolution of 0.25 • × 0.5 • (lat × lon). The EMI TOC showed a similar spatial distribution to both OMI and TROPOMI, with an R of 0.99 for both. Daily GB measurements of TOC from seven monitoring stations were also used to validate the EMI TOC and stations located in Germany, France, USA, Australia (Brisbane and Melbourne), Argentina, and Antarctica, and had R of 0.92, 0.93, 0.95, 0.94, 0.95, 0.95, and 0.97, respectively. The average absolute difference between GB measurements and EMI observations was less than 5.5%, indicating that TOC retrieved from EMI irradiance using the DOAS method is reliable.
Evaluation of the Antarctic ozone hole and ozone depletion, as one of the most important missions of the EMI, will be conducted in future work. Here, we mainly focus on the accuracy of the EMI TOC retrieval. Daily average TOC from EMI observations on the polar vortex edge region (Chinese Great Wall Station) of Antarctica, where rapid changes and large fluctuations in TOC can be detected, were compared with ground-based ZSL-DOAS measurements. The EMI TOC showed similar fluctuations compared with ZSL-DOAS measurements during the observation period, with an average absolute difference of 5.33% and an R of 0.90. The ozone hole above the Great Wall Station was simultaneously monitored using EMI and ZSL-DOAS measurements during the observation period, and the impact of an SSW event in October 2019 on the ozone hole was analyzed. These results highlight the potential for EMI observations to be used to monitor the ozone hole.
It should be noted that the EMI TOC results of this study are preliminary. More validations with global GB measurements should be conducted in future, and the EMI TOC retrieval algorithm needs further improvement. For example, the EMI cloud product should be updated in the future. The cloud information used in this study was obtained from TROPOMI cloud products, which limits the accuracy of the EMI TOC retrieval algorithm due to the different overpass times of EMI and TROPOMI. Furthermore, the surface albedo algorithm (especially in high-latitude areas) should be improved in the future, and the distribution of ice or snow should be considered when calculating the AMF in future work. Finally, only a single solar irradiance measured on 12 June 2018 in this mission was used as the reference spectrum, and more solar irradiance measurements should be conducted in future missions. The retrieval algorithm of EMI TOC evaluated in this study is valuable for the future development of EMI TOC products.

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author. Figure A1. Spatial distribution of (a) daily average EMI TOC, (b) EMI relative daily average SCD error, (c) daily average TROPOMI TOC, (d) average relative difference between EMI and TROPOMI, (e) daily average OMI TOC, and (f) average relative difference between EMI and OMI on 27 October 2018. Appendix A Figure A1. Spatial distribution of (a) daily average EMI TOC, (b) EMI relative daily average SCD error, (c) daily average TROPOMI TOC, (d) average relative difference between EMI and TROPOMI, (e) daily average OMI TOC, and (f) average relative difference between EMI and OMI on 27 October 2018. Figure A2. Spatial distribution of (a) daily average EMI TOC, (b) EMI relative daily average SCD error, (c) daily average TROPOMI TOC, (d) average relative difference between EMI and TROPOMI, (e) daily average OMI TOC, and (f) average relative difference between EMI and OMI on 16 October 2019.