Observing the Diurnal Variation of Atmospheric Ozone From the Geostationary Interferometric Infrared Sounder (GIIRS) Over East Asia Shangyi Liu1, Jiancong Hua1, Huidong Wang1 , Shan Han1,2, Lu Lee3, Chengli Qi3, Feng Lu3 , Yangcheng Luo4, Xiaoyi Zhao5 , Zhengqiang Li6 , Sang‐Woo Kim7, Chang Keun Song8 , Yugo Kanaya9, Arno Keppens10 , Jean‐Christopher Lambert10 , Cathy Clerbaux11 , and Zhao‐Cheng Zeng1 1School of Earth and Space Sciences, Peking University, Beijing, China, 2College of Geoscience and Surveying Engineering, China University of Mining and Technology (Beijing), Beijing, China, 3Key Laboratory of Radiometric Calibration and Validation for Environmental Satellites, Innovation Center for FengYun Meteorological Satellite, National Satellite Meteorological Center, China Meteorological Administration, Beijing, China, 4LMD/IPSL, Sorbonne Université, ENS, PSL, École Polytechnique, Institut Polytechnique de Paris, CNRS, Paris, France, 5Air Quality Research Division, Environment and Climate Change Canada, Toronto, ON, Canada, 6State Environmental Protection Key Laboratory of Satellite Remote Sensing, Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing, China, 7School of Earth and Environmental Sciences, Seoul National University, Seoul, South Korea, 8Graduate School of Carbon Neutrality, Ulsan National Institute of Science and Technology, Ulsan, South Korea, 9Institute of Arctic Climate and Environment Research, Research Institute for Global Change, Japan Agency for Marine‐Earth Science and Technology (JAMSTEC), Yokohama, Japan, 10Royal Belgian Institute for Space Aeronomy (BIRA‐IASB), Uccle, Belgium, 11LATMOS/ IPSL, Sorbonne Université, UVSQ, CNRS, Paris, France Abstract Ozone is an important atmospheric constituent, exerting a pivotal influence on atmospheric chemistry, air quality, and climate change. The monitoring of its distribution and variation is crucial for advancing our understanding of ozone development and related processes. This study presents the first spatial and temporal distributions of total ozone columns (TOC) retrieved from the Geostationary Interferometric Infrared Sounder (GIIRS), on board China's FengYun‐4B satellite (FY‐4B/GIIRS) launched in 2021. Particularly, we focus on the variations of TOCs in East Asia from diurnal to seasonal time scales. Retrievals are implemented using spectra from March, June, September, and December, representing different seasons. The results show that the degree of freedom for the signal (DOFS) typically exhibited a range of 0.8–1.4, with the vertical detection sensitivity of GIIRS peaking in the upper troposphere/lower stratosphere (UTLS) region, where the ozone variability is the highest. Collocation comparisons with the Infrared Atmospheric Sounding Interferometer (IASI) retrievals, the Ozone Monitoring Instrument (OMI) measurements, the European Center for Medium‐Range Weather Forecasts (ECMWF) Reanalysis v5 (ERA5) simulations, and in situ ozone observations show good agreement. The comparisons of TOCs between GIIRS, Pandora and ERA5 at different latitudes and different time scales demonstrate the ability of FY‐4B/GIIRS in capturing the temporal and latitudinal ozone variations, particularly at middle and high latitudes. Our work demonstrates that FY‐4B/GIIRS has good capability to track ozone variations from diurnal to seasonal in East Asia, which will contribute to the understanding of regional and global ozone variations. Plain Language Summary Ozone in the atmosphere is an important compound that affects air quality and the Earth's radiation budget. It is primarily located in the stratosphere, where it plays a beneficial role by blocking the sun's damaging ultraviolet rays, preserving life on Earth. But in the troposphere, ozone is a harmful air pollutant that poses a threat to human health and the ecosystem. Satellite‐based ozone measurements can provide broad, continuous observations that enhance our insight into its spatial and temporal dynamics worldwide. The Geostationary Interferometric Infrared Sounder (GIIRS) on board China's FengYun satellite series represents the world's first hyperspectral infrared sounder in a geostationary orbit. This work presents the first spatial and temporal distributions of total ozone column (TOC) retrieved using spectra from GIIRS on board FengYun‐4B satellite (FY‐4B/GIIRS), offering a spatial sampling of 12 km and a 2‐hourly time interval. Through comparisons with other ozone measurements, we demonstrate that FY‐4B/GIIRS can capture the TOC variations from diurnal to seasonal in East Asia. RESEARCH ARTICLE 10.1029/2024JD042292 Key Points: • We show the first ozone retrieval results from a geostationary hyperspectral infrared sounder • Retrieval sensitivity peaks in the upper troposphere/lower stratosphere region • FY‐4B/GIIRS shows great potential for monitoring the diurnal to seasonal ozone variability over East Asia Supporting Information: Supporting Information may be found in the online version of this article. Correspondence to: Z.‐C. Zeng, zczeng@pku.edu.cn Citation: Liu, S., Hua, J., Wang, H., Han, S., Lee, L., Qi, C., et al. (2025). Observing the diurnal variation of atmospheric ozone from the Geostationary Interferometric Infrared Sounder (GIIRS) over East Asia. Journal of Geophysical Research: Atmospheres, 130, e2024JD042292. https://doi.org/10. 1029/2024JD042292 Received 21 AUG 2024 Accepted 13 APR 2025 Author Contributions: Conceptualization: Zhao‐Cheng Zeng Data curation: Lu Lee, Chengli Qi, Feng Lu, Xiaoyi Zhao, Zhengqiang Li, Sang‐Woo Kim, Chang Keun Song, Yugo Kanaya, Arno Keppens, Jean‐ Christopher Lambert Formal analysis: Shangyi Liu, Jiancong Hua, Huidong Wang, Shan Han, Lu Lee, Chengli Qi, Feng Lu, Yangcheng Luo, Cathy Clerbaux, Zhao‐ Cheng Zeng Funding acquisition: Zhao‐Cheng Zeng Investigation: Zhao‐Cheng Zeng Methodology: Shangyi Liu, Zhao‐ Cheng Zeng Project administration: Zhao‐ Cheng Zeng Resources: Cathy Clerbaux Software: Shangyi Liu, Zhao‐Cheng Zeng Supervision: Shangyi Liu, Zhao‐ Cheng Zeng Validation: Shangyi Liu Visualization: Shangyi Liu Writing – original draft: Shangyi Liu © 2025. American Geophysical Union. All Rights Reserved. LIU ET AL. 1 of 27 https://orcid.org/0009-0004-5382-5003 https://orcid.org/0009-0006-6589-6719 https://orcid.org/0000-0003-4784-4502 https://orcid.org/0000-0002-7795-3630 https://orcid.org/0000-0002-8987-2176 https://orcid.org/0000-0002-9544-6392 https://orcid.org/0000-0001-7243-6848 https://orcid.org/0000-0003-0394-7200 https://orcid.org/0000-0002-0008-6508 mailto:zczeng@pku.edu.cn https://doi.org/10.1029/2024JD042292 https://doi.org/10.1029/2024JD042292 http://crossmark.crossref.org/dialog/?doi=10.1029%2F2024JD042292&domain=pdf&date_stamp=2025-04-24 1. Introduction Ozone (O3) is an important trace gas in the Earth's atmosphere, playing a pivotal role in climate change and ecological environment. It is primarily found in the stratosphere (almost 90%), where it is produced by the photolysis of molecular oxygen. Its presence serves as a protective barrier for the biosphere and human life by filtering out harmful ultraviolet radiation. In the troposphere, ozone is predominantly produced by photochemical reactions of nitrogen oxides and volatile organic compounds (Crutzen, 1974). Furthermore, its concentration is also influenced by meteorological phenomena, including the exchange between the stratosphere and troposphere (e.g., Griffiths et al., 2020). Due to its high oxidizing capacity, tropospheric ozone has the potential to cause adverse effects on human health (e.g., Cohen et al., 2017), as well as to ecosystems and agri‐food products (e.g., Fowler et al., 2009). In the free troposphere, the ozone can be transported on an intercontinental scale due to its estimated lifetime of several weeks (e.g., Monks et al., 2015). It is thus evident that the monitoring of O3 is of significant importance for the causes of and effects of climate change, the comprehension of atmospheric chemical processes, and the development of environmental policies. In order to gain a deeper comprehension of the variability and consequences of ozone, it is of paramount importance to obtain data concerning its vertical and spatio‐temporal distribution. Such data are accessible through obtained from satellite observations and in‐situ observations. Early satellite‐based monitoring of ozone began with USSR Kosmos missions in 1964–1965 (Iozenas et al., 1969), NASA's Orbiting Geophysical Ob- servatory in 1967–1969 (Anderson et al., 1969) and NASA's BUV on Nimbus 4 in 1970–1975 (Heath et al., 1973). Since then, a variety of spaceborne instruments with different spectral bands and observation modes have been launched. These include the Microwave Limb Sounder (MLS, Waters et al., 2006), Global Ozone Monitoring Experiment‐2 (GOME‐2) (Munro et al., 2016), Ozone Mapping and Profiler Suite (OMPS) (Seftor et al., 2014), Tropospheric Monitoring Instrument (TROPOMI) (Veefkind et al., 2012), the Environmental Trace Gases Monitoring Instrument (EMI) (Su et al., 2022) and Geostationary Environment Monitoring Spectrometer (GEMS) (Kim et al., 2020) and so on, which provide continuous and consistent atmospheric ozone observations. Of the various sensors available, thermal infrared hyperspectral sounders have a distinct advantage as it allows for observations to be made both in the daytime and nighttime. Currently, such ozone observations can be obtained using instruments such as the Atmospheric Infrared Sounder (AIRS) (e.g., Susskind et al., 2014), the three Infrared Atmospheric Sounding Interferometers (IASI) (e.g., Boynard et al., 2018), the two Cross‐track Infrared Sounder (CrIS) (e.g., Nalli et al., 2018), and the Thermal and Near Infrared Sensor for Carbon Observation‐ Fourier Transform Spectrometer (TANSO‐FTS) (e.g., Ohyama et al., 2012). In addition, previous studies have also discussed the effectiveness of combining ultraviolet and infrared sensors for ozone retrieval, such as the combination of GOME‐2 with IASI (Cuesta et al., 2013), Ozone Monitoring Instrument (OMI) with AIRS (Fu et al., 2018), and more recently TROPOMI with CrIS (Mettig et al., 2022). The results showed that the infor- mation content and the accuracy of the retrieval have been improved to some extent. The above mentioned thermal infrared hyperspectral instruments are deployed on polar orbits, which can only take a limited number of repetitive measurements over the same location during a single day, that is, two at low latitudes and possibly multiple at high latitudes. Using instruments on geostationary platforms with higher temporal sampling frequency to make similar measurements has the potential to offer significant advantages in monitoring total ozone column (TOC). The two Chinese Geostationary Interferometric Infrared Sounders (GIIRS) represent the world's first hyperspectral infrared sounder to be installed on geostationary meteorological satellites. These include the Fengyun‐4A (FY‐4A), launched in 2016, and the Fengyun‐4B (FY‐4B), launched in 2021 (Li, Menzel, et al., 2022; Yang et al., 2017). FY‐4B/GIIRS scans the East Asian region every 2 hr with a sampling of 12 km × 12 km. It has a good spectral resolution of 0.625 cm− 1 similar to that of current low‐Earth‐ orbit (LEO) satellites. For example, it is 0.1 cm− 1 at nadir for TES (Beer et al., 2001); 0.25 cm− 1 for IASI (Clerbaux et al., 2009); 0.625 cm− 1 for CrIS (Han et al., 2013); from 0.5 to 2 cm− 1 for AIRS (Pagano et al., 2003). FY‐4B/GIIRS enables the detection of multiple atmospheric trace gases, including carbon monoxide (Zeng, Lee, & Qi, 2023), ammonia (Zeng, Lee, Qi, Clarisse et al., 2023) and formic acid (Zeng et al., 2024). The combination of good spectral, spatial, and temporal resolutions in FY‐4B/GIIRS observations offers significant potential for obtaining accurate and timely information on atmospheric ozone. The utilization of FY‐4B/GIIRS observations has the potential to enhance the comprehension of exchange processes of ozone between the troposphere and stratosphere, as well as its regional transport. Furthermore, given the close relationship between short‐term TOC variability and weather systems, TOC is regarded as an important Writing – review & editing: Shangyi Liu, Jiancong Hua, Huidong Wang, Shan Han, Lu Lee, Chengli Qi, Feng Lu, Yangcheng Luo, Xiaoyi Zhao, Sang‐ Woo Kim, Chang Keun Song, Yugo Kanaya, Arno Keppens, Jean‐ Christopher Lambert, Cathy Clerbaux, Zhao‐Cheng Zeng Journal of Geophysical Research: Atmospheres 10.1029/2024JD042292 LIU ET AL. 2 of 27 21698996, 2025, 8, D ow nloaded from https://agupubs.onlinelibrary.w iley.com /doi/10.1029/2024JD 042292 by E nvironm ent C anada, W iley O nline L ibrary on [09/01/2026]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense tracer for meteorological processes, including tropical cyclones (Rodgers et al., 1990) and hurricanes (Joiner et al., 2006). In theory, the 2‐hourly TOCs data from FY‐4B/GIIRS can provide more detailed information on the variations of these weather processes. Additionally, high temporal resolution time‐varying TOCs can provide essential information for modeling future ozone variations. Moreover, the atmospheric ozone variation is also related to several parameters such as ultraviolet radiation (e.g., Antón et al., 2008), regional wind fields, and potential vorticity (e.g., Vaughan & Price, 1991). The assimilation of O3 data from satellites into numerical models can effectively improve the forecast accuracy of the parameters related to ozone changes (e.g., Levelt et al., 1996; Liu & Zou, 2015). Therefore, high‐frequency ozone observations from FY‐4B/GIIRS have great potential to further improve the forecast. This study applies the FY‐GeoAIR retrieval algorithm developed by Zeng, Lee, and Qi (2023), based on optimal estimation theory (e.g., Rodgers, 2000), to retrieve O3 from the observed spectrum of FY‐4B/GIIRS, with ad- aptations made to accommodate ozone, such as the construction of the forward model for ozone absorption band, the a priori ozone profile with its variability and the a priori ozone variance‐covariance matrix. To our knowledge this represents the first report in the literature of retrieving atmospheric O3 from a thermal infrared hyperspectral sounder on board a geostationary satellite. Our retrieval algorithm utilizes the strong absorption band of ozone at 9.6 μm to produce diurnally continuous ozone distributions for the East Asian region, including TOC and its associated vertical profile. Additionally, the retrieval process provides the degrees of freedom for the signal (DOFS), which quantifies the information content derived from satellite measurements. To illustrate the reli- ability of the retrieved results, ozone data from IASI and OMI instrument, the European Center for Medium‐ Range Weather Forecasts (ECMWF) Reanalysis v5 (ERA5), ozonesonde observations and Pandora measure- ments are compared to FY‐4B/GIIRS ozone retrievals. In addition, we explore the use of GIIRS to track the dynamics of TOCs from diurnal to seasonal time scales in different latitudinal zones. The rest of this paper is structured as follows. Section 2 describes the instrument and the O3 retrieval algorithm based on FY‐GeoAIR. Section 3 presents a simulated experiment aiming at evaluating the efficacy of the retrieval algorithm. Sections 4 and 5 provide the evaluation of the information content, the cross‐comparison of the GIIRS ozone products with IASI, OMI, ERA5 ozonesonde and Pandora data, and analysis of the diurnal to seasonal variations of TOC captured by GIIRS. Section 6 summarizes this study and presents an outlook. 2. Measurements and Methods 2.1. The GIIRS Instrument The FY‐4B satellite is located at an altitude of about 36,000 km above the Earth's equator at a longitude of 133°E. It observes the East Asian region with a 12‐km spatial sampling at nadir and a 2‐hr time interval, that is, a total of 12 observations per day (commencing at 1, 3, 5, …, 23hr UTC, respectively). Note that before 6th September 2022, the start times were at 0, 2, 4, …, 22 hr UTC. With an apodized spectral resolution of 0.625 cm− 1, FY‐4B/ GIIRS measures long‐wave infrared radiation of 680–1,130 cm− 1 and the mid‐wave infrared radiation of 1,650– 2,250 cm− 1. Therefore, GIIRS has the ability to detect several atmospheric trace gases and provide continuous day‐night observations. Figure 1a shows an example of the GIIRS long‐wave infrared observation spectrum, which includes the ozone absorption band (highlighted in gray). To assess the radiative performance of the GIIRS spectra, Li, Ni, et al. (2022) conducted blackbody calibration experiments in a laboratory thermal vacuum tank before launch. The results indicated that the average noise equivalent differential radiance (NedR) in the long‐wave infrared band, which covers the ozone absorption channels, is below 0.5 mW/(m2 sr cm− 1) (equivalent to ∼0.4 K for reference temperature of 280 K at the top of atmosphere) on average. Moreover, according to the post‐launch GIIRS radiation products provided by the FengYun Satellite Data Center (FengYun Satellite Data Center, 2023), the average NedR in the O3 absorption band from 1,000 to 1,100 cm− 1 is approximately 0.3 mW/(m2 sr cm− 1). Figure 1b shows the NedR values corresponding to the GIIRS observed spectrum in Figure 1a, in the ozone absorption band. The instrumental noise of FY‐4B/GIIRS is close to that of existing infrared sounders, such as IASI (∼0.2 mW/(m2 sr cm− 1) in 1,000– 1,100 cm− 1, Clerbaux et al., 2009), thereby allowing for precise ozone retrieval in the East Asia region. Journal of Geophysical Research: Atmospheres 10.1029/2024JD042292 LIU ET AL. 3 of 27 21698996, 2025, 8, D ow nloaded from https://agupubs.onlinelibrary.w iley.com /doi/10.1029/2024JD 042292 by E nvironm ent C anada, W iley O nline L ibrary on [09/01/2026]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense 2.2. Radiative Transfer Model In this work, the FY‐GeoAIR (FengYun Geostationary satellite Atmospheric Infrared Retrieval) algorithm is applied to obtain total ozone and coarse vertical profiles from large FY‐4B/GIIRS spectra. The algorithm was initially designed for the extraction of carbon monoxide (Zeng, Lee, & Qi, 2023) and has been successfully applied to ammonia and formic acid retrieval (Zeng, Lee, & Qi, 2023; Zeng et al., 2024). This section briefly introduces the FY‐GeoAIR algorithm that is adapted to O3 retrieval. For further detailed information, please refer to the source text by Zeng, Lee, and Qi (2023). An accurate forward model based on radiative transfer theory is essential for atmospheric trace gases retrieval. The forward model requires the definition of a discrete atmospheric grid, the collection of input parameters of atmospheric state, surface features and instrumental specifications, and the spectroscopic database for calculating the trace gases absorption cross‐sections. In accordance with the radiative transfer theory pertaining to thermal Figure 1. FY‐4B/GIIRS observation spectra and the a priori strategies for ozone retrieval. (a) Example of a longwave spectrum measured by FY‐4B/GIIRS from 03:00 to 04:00 UTC in March 2023, covering the O3 absorption band centered at 1,050 cm− 1 (the gray rectangle). (b) NedR of the observed spectra corresponding to the gray rectangles in panel (a). (c) A priori profile (in units of ppm, blue line) with associated variability (shaded blue, one standard deviation) for ozone and the retrieval levels in blue points. The red dashed line represents the smallest pressure of the retrieved layer in the retrieval algorithm. The inset in the lower right corner is an enlarged view of the a priori profile between 1,000 and 100 hPa. (d) A priori correlations matrix (top) and a priori variance‐covariance matrix in unitless multiplicative factor. (Bottom) built from ERA5 data. Journal of Geophysical Research: Atmospheres 10.1029/2024JD042292 LIU ET AL. 4 of 27 21698996, 2025, 8, D ow nloaded from https://agupubs.onlinelibrary.w iley.com /doi/10.1029/2024JD 042292 by E nvironm ent C anada, W iley O nline L ibrary on [09/01/2026]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense infrared radiation (Liou, 2002), under conditions of clear skies and when the scattering effects of aerosols and clouds are negligible, the upwelling radiance at the TOA comprises: (a) The upwelling radiance emitted from the Earth's surface; (b) the radiance emitted upwards from the whole atmosphere; (c) the downwards atmospheric radiance, which is reflected by the surface and re‐emitted upwards; (d) the reflected solar radiation. Notwith- standing the negligible quantity of solar radiation within the ozone absorption window, it is nevertheless included here for the sake of completeness. For the purpose of radiative transfer calculations, it is necessary to consider a discrete layered atmosphere. In our approach, the forward model uses fixed pressure vertical grids (ranging from 1,000 to 0.02 hPa, absolute pressure values provided in Table S1 in Supporting Information S1). Similar to the grid setup in Boynard et al. (2009), we separate the atmosphere into 32 layers with similar thicknesses, which are about 1 km for the bottom layer and 2 km for the other layers below 1 hPa. The atmosphere from 1 to 0.02 hPa is separated into 6 layers. The bottom layer thickness used in the calculations is adaptable in accordance with the varying surface pressure. For atmospheric parameters, the atmospheric temperature and H2O variables are sourced using the ERA5 data (Hersbach et al., 2020). The exception is that for the upper‐atmosphere grid with pressures less than 1 hPa, the atmospheric parameters are obtained from the US standard atmosphere data. The other interfering gas considered in this study within the ozone absorption spectral window is carbon dioxide (CO2) and H2O. The CO2 data are obtained from ECMWF Copernicus Atmosphere Monitoring Service (CAMS) (Copernicus Atmosphere Moni- toring Service, 2021). In addition, we cloud‐screened the GIIRS spectra using the Level‐2 cloud mask data product from the Advanced Geostationary Radiation Imager (AGRI) onboard FY‐4B. The footprints of GIIRS and AGRI are collocated, and GIIRS observations were classified as clear or near‐clear when at least 80% of the corresponding AGRI pixels are labeled as clear or probably clear. More technical details can be found in Zeng, Lee, and Qi (2023). For surface parameters, the surface temperature and pressure are from hourly ERA5 single‐ level data set (Hersbach et al., 2020). The surface emissivity data are sourced from the global infrared surface emissivity database, which is maintained at the University of Wisconsin‐Madison (UOW‐M) and documented by Seemann et al. (2008). As for the spectroscopic database, we pre‐calculate the trace gases absorption coefficients at different pressures and temperatures and store them in the look‐up table. Specifically, this look‐up table is constructed using the HITRAN database within the Line‐By‐Line Radiative Transfer Model, as noted by Clough et al. (2005) and Rothman et al. (2013). The look‐up tables are constructed for 59 levels of atmospheric pressure from 1,025 hPa down to 0.02 hPa and for 15 levels of temperature from 180 to 320 K with a step of 10 K. 2.3. The Retrieval Algorithm This paper develops an ozone retrieval methodology for FY‐4B/GIIRS spectra based on the Optimal Estimation Method (OEM) (Rodgers, 2000). The OEM has been widely utilized in the atmospheric trace gases retrieving from remote sensing observations (Zeng et al., 2017, 2021). The objective of OEM is to identify the solution minimizing the cost function as J(x) = [y − F(x,b)]TS− 1ε [y − F(x,b)] + (x − xa)TS− 1a (x − xa), (1) where y denotes the measurement vector containing the GIIRS spectral radiance ranging from 1,025 to 1,075 cm− 1, which is a commonly used window dominated by O3 lines (e.g., Boynard et al., 2009), with only a few H2O and CO2 lines interfering; F denotes the forward radiative transfer model; x is the state vector composed of the parameters to be retrieved (listed in Table 1), where ozone is retrieved as vertical profiles. In practice, we used scale factors of the a priori partial columns for each layer as the state vector elements for ozone. In the following analysis, total ozone columns are obtained by integrating ozone profiles. b represents all the other parameters that are not retrieved and have effects on the observed spectral radiance (such as satellite zenith angle, surface pressure and other relevant geophysical parameters); Sε is the measurement error covariance matrix; Sa denotes the a priori covariance matrix for x; xa is the a priori state vector. The algorithm retrieves only the ozone layers with a pressure greater than 1 hPa and utilizes the a priori partial columns for the layers above this pressure. The objective of the inverse model based on OEM is to find a solution that generates simulated spectra from the forward model to have the optimal consistency with the measurement, subject to constraints from both the ob- servations and the a priori values. The weights between the two are determined by their error covariance matrices representing their statistical variations. The a priori variance‐covariance matrix Sa is a representation of our Journal of Geophysical Research: Atmospheres 10.1029/2024JD042292 LIU ET AL. 5 of 27 21698996, 2025, 8, D ow nloaded from https://agupubs.onlinelibrary.w iley.com /doi/10.1029/2024JD 042292 by E nvironm ent C anada, W iley O nline L ibrary on [09/01/2026]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense knowledge of the expected variability of the state vector for the specific time and place of observation. Figures 1c and 1d display the ozone a priori profile and covariance matrix, which are utilized in this study. They are con- structed from ERA5 data from July 2022 to June 2023 with a focus on the representative areas within East Asia (0–60°N, 110–130°E). The a priori ozone at altitudes above the 1 hPa pressure level is taken from the US standard atmosphere data. To make it coherent with the a priori ozone profile at altitudes below 1 hPa, it is slightly shifted by less than 1 Dobson Unit (1 DU = 2.687 × 1020 molecule m− 2) in the partial column. We use a fixed a priori ozone profile to all observations at different times and locations, which has the advantage of ensuring that any changes in the spatio‐temporal distribution of O3 retrieved are a true reflection of the data captured by the FY‐4B/ GIIRS spectra. As illustrated in Figure 1c, the ozone concentration slowly increases from ∼44 ppb at the surface to ∼160 ppb at 200 hPa and reaches a maximum of ∼8.2 ppm in the middle stratosphere. The highest degree of ozone variability is observed in the upper troposphere and the lower stratosphere (UTLS) region, approximately at 10–20 km, with a variability of approximately 80%. In other altitude layers, the variability does not exceed 30%. In addition, a significant positive correlation of ozone concentrations with altitude can be observed from the correlation matrix in Figure 1d in the layer indexes of 1–5, 5–13, 14–22 and 22–26 (corresponding to about 0–9, 9–25, 25–43 and 43–51 km, respectively). This is similar to the covariance matrix employed by FORLI, the IASI retrieval software (Hurtmans et al., 2012). As for the measurement error covariance matrixSε, in this study we assume that it has a diagonal structure, which implies the absence of cross‐correlation among the distinct spectral channels of the measurements. The values of the diagonal of the matrix are the square of NedR, as a function of wavenumber, provided by the L1 spectra data product for each observation, as introduced in Section 2.1. However, extending beyond mere instrumental noise, Sε incorporates a multitude of error sources in theory, such as spectra correction errors, forward model errors, residual cloud contamination in the radiances, and uncertainties originating from atmospheric variables. Quan- tifying the errors from all of these sources is challenging. Instead, we construct Sε by scaling up the diagonal measurement noise, as in Boynard et al. (2009). For consistency, we choose a uniform multiplier for scaling since the average radiometric noise of the GIIRS spectra in the ozone absorption window is close at different times and regions. To determine the scaling multiplier, we use the reduced χ2 as an indicator to evaluate the goodness of fit of the retrieval, which is a commonly used statistic. After numerous tests at different times and locations, we found the measurement noise enlarged by 8 times, to approximately 2.3 mW/m2 sr cm− 1, to achieve the objective that the reduced χ2 values are close to or less than 1.0. Although in some cases, this may result in a reduction in the extent of information available from the observations, it also avoids the incorrect retrievals caused by treating the errors as measurement information for most observations. The reduced χ2 values of different months are shown in the Supplement. The Levenberg‐Marquardt method is used to search for optimal solutions (Rodgers, 2000). For iteration i + 1: xi+1 = xi + [(1 + γ)S− 1a +KT i Ki] − 1 {KT i S− 1ε [y − F(xi,b)] − S− 1a [xi − xa]}, (2) where K represents the Jacobian matrix, with its the rows denoting the first derivatives of F with respect to the retrieved parameters; the description of γ can be found in Rodgers (2000). Convergence is achieved when the Table 1 The State Vector Parameters Used in the Retrieval Algorithm Variables Number of variables A priori values A priori uncertainty Descriptions O3 26 Fixed See Figure 2 Derived from ERA5 reanalysis H2O 26 ERA5 20% CO2 1 CAMS 5% Only profile scaling factors are retrieved Surface skin temperature 1 ERA5 5K Atmospheric temperature profile 1 ERA5 0.05% Only a profile scaling factor is retrieved Surface emissivity slope and curvature 4 [0.0, 0.0, 0.0, 0.0] [0.1%, 0.01%, 0.001%, 0.0001%] UOW‐M Note. The gases, including O3, H2O, and CO2, in the state vector are represented by profiles of partial columns in units of molecules/cm 2. Journal of Geophysical Research: Atmospheres 10.1029/2024JD042292 LIU ET AL. 6 of 27 21698996, 2025, 8, D ow nloaded from https://agupubs.onlinelibrary.w iley.com /doi/10.1029/2024JD 042292 by E nvironm ent C anada, W iley O nline L ibrary on [09/01/2026]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense change in the state vector is small compared to the a posteriori error, similar to Zeng et al. (2021). The a posteriori error variance–covariance matrix Ŝ, can be estimated as Ŝ = (KTS− 1ε K + S− 1a ) − 1 (3) The averaging kernel (AK) matrix, which is commonly used to characterize the quality of the retrieval results, can be written as A = (KTS− 1 ε K + S− 1 a ) − 1KTS− 1 ε K, (4) where the rows of A are the derivatives of the retrieved variables with respect to the true atmospheric state. The trace of the AK matrix is referred to as the degree of freedom for signal (DOFS), representing the quantity of individual information fragments provided by the different atmospheric layers. In summary, when applying the FY‐GeoAIR retrieval algorithm to ozone retrieval, unique challenges emerge due to both spectral and vertical profile characteristics that are distinct from gases such as carbon monoxide. First, the ozone absorption band at 9.6 μm has different interfering gases (mainly water vapor and CO2) and higher spectral noise in the GIIRS instrument than in other gases (see Figure S25 in Supporting Information S1). This requires adjustments to the measurement error covariance matrix Sε. Second, ozone exhibits a unique vertical profile and variation. Ozone spans from the troposphere to the stratosphere, with a dominant stratospheric presence, unlike carbon monoxide that is concentrated near the surface or in the lower troposphere. Furthermore, ozone profile exhibits significant daily and seasonal variability, especially in the UTLS region. Therefore, it demands ozone‐ specific prior constraints for the retrieval that take into account its special vertical variability and inter‐layer correlations. 3. Simulated Experiments Using Synthetic Spectra In this section, we construct simulated synthetic spectra using the ECMWF ERA5 hourly O3 data (defined as the “truth” here) and apply the retrieval algorithm presented in the previous section. The effectiveness of the retrieval algorithm can be assessed by comparing the retrieved results with the “truth”. We use the forward radiative transfer model described in Section 2.2 to generate the simulated spectra. The assumed noise (Gaussian noise) is added according to the scaled spectral noise described in Section 2.3 (i.e., 8 times of the nominal instrumental noise). The a priori ozone profile and variance‐covariance matrix described in Figures 1c and 1d are used in the simulation experiments. Specifically, we randomly selected a day in March 2023 when the TOC reaches its yearly maximum in the northern hemisphere high latitudes (e.g., Salawitch et al., 2023). We sampled four observation cycles of FY‐4B/ GIIRS at different time periods (03–04, 09–10, 13–14, and 21–22 hr UTC) for simulations. In the experiments, we sequentially extracted 10% data points from the clear‐sky observations, and performed a total of 4,972 simula- tions. The averaged spectral fitting residuals for all simulations are in the Figure S1 in Supporting Information S1, showing a good spectral fitting with small residual values. Finally, we compared the TOC retrievals with the “truth” in Figure 2. The DOFS is also shown. The a priori TOC is also shown in the figure and varies due to the different surface pressures at different locations. The scatterplots show that the retrieved results are in very good agreement with the “truth” across different observation cycles and the retrievals are almost uncorrelated with the a priori values. The correlation coefficients between the retrieved results and the “truth” are 0.99 and the root mean square errors (RMSEs) range from 8.83 to 11.42 DU. The good agreement demonstrates the effectiveness of the retrieval algorithm. Furthermore, it can be seen that, compared to observations with low TOCs, observations with high TOCs have lower DOFS values. The agreement between the retrieved results with high TOC and the “truth” is not as good, with a slight underestimation. The distribution of the total ozone is latitudinal, with high values mainly at high‐latitude regions and low values mainly in the tropical regions, as further discussed in Sections 4.1 and 4.2. At high latitudes with high TOC values, the surface temperature is low in March, which is not favorable for thermal infrared remote sensing because lower surface temperatures result in poorer signal‐to‐noise ratios in observations. As a result, the information content that can be extracted from the measured spectra is small, corresponding to low DOFS values and high uncertainties for the ozone retrievals. Moreover, the structure of the “truth” ozone profile differs significantly from the a priori profile structure at high latitudes. The retrieved results Journal of Geophysical Research: Atmospheres 10.1029/2024JD042292 LIU ET AL. 7 of 27 21698996, 2025, 8, D ow nloaded from https://agupubs.onlinelibrary.w iley.com /doi/10.1029/2024JD 042292 by E nvironm ent C anada, W iley O nline L ibrary on [09/01/2026]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense are more biased toward the a priori profile. This reveals the reason for the slightly lower retrieved results at high latitudes. In addition, in the simulated experiments at various periods throughout the day, we found that daytime retrievals (03–04, 09–10 hr UTC) are better than that at nighttime (13–14 hr, 21–22 hr UTC), particularly in high latitude regions. There are more retrievals with low DOFS values and large deviations from the “truth” in the nighttime retrievals compared to that in the daytime. The difference is likely due to the lower surface temperature at night that decreases signal‐to‐noise ratios in the spectra observation. Therefore, it is necessary to filter the retrieved results with low DOFS properly. Overall, the simulated experiments demonstrate the effectiveness of the FY‐GeoAIR for ozone retrieval from GIIRS spectra. It is expected that the FY‐4B/GIIRS observations with a two‐hour measurement cycle can effectively quantify the diel and seasonal variations of ozone over East Asia. 4. Results and Comparisons 4.1. O3 Retrievals From GIIRS Spectral Radiance The FY‐4B/GIIRS, with its 2‐hourly observation recycle, has advantages over conventional LEO satellite sounders in measuring the distribution and variability of atmospheric ozone on hourly scales. This section spe- cifically describes the characterization of O3 retrievals from GIIRS observations. We selected all clear‐sky ob- servations of GIIRS from four months, including December 2022 and March, June and September 2023, for our experiments. The months of March and September correspond to the early stages of spring and early autumn, respectively, in the Northern Hemisphere. June and December, in turn, are associated with early summer and early winter. These selected observations are sufficiently representative in time and space to illustrate the efficacy of the retrieval algorithm. After cloud filtering, the total number of clear sky observations is about 1.46 million for the 4 months. In addition, to guarantee the good quality of the retrievals, multiple filters were used in the post‐ processing. First, any retrievals that do not converge within 10 iterations were deemed invalid and discarded. Figure 2. Comparison of the retrieved and the “truth” ozone columns obtained by the simulated experiments in 4 GIIRS observation cycles on 15 March 2023. In the experiments, the retrieval results are based on the simulated spectra generated by the “truth” data. The gray points indicate the priori values for each retrieval. The data points retrieved are color‐coded in accordance with the corresponding DOFS values. Journal of Geophysical Research: Atmospheres 10.1029/2024JD042292 LIU ET AL. 8 of 27 21698996, 2025, 8, D ow nloaded from https://agupubs.onlinelibrary.w iley.com /doi/10.1029/2024JD 042292 by E nvironm ent C anada, W iley O nline L ibrary on [09/01/2026]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense Subsequently, retrievals exhibiting a fitting residual RMSE larger than the sum of the mean and two times the standard deviation of RMSE (approximately 2.2 K) were excluded. Next, retrievals with DOFS values less than 0.8, representing low information content and thus unreliable results, were removed. Moreover, retrievals with abnormally high and low TOC values are also filtered, which is delineated as results that are more than three times the standard deviation above or below the monthly mean. Finally, observations with a difference of more than 15 K in the surface skin temperature from the retrieved data compared to ERA5 data were excluded, as such large discrepancies likely indicate cloud‐contaminated measurements. After post‐processing, approximately 1.36 million observations pass the filter, accounting for 92.8% of the clear‐sky observations. Figure 3 gives an example of retrieval from a GIIRS observation spectrum, including the spectral fit with its residuals, the retrieved ozone profiles, the corresponding AK functions and the error profiles of the retrieval. The fitted spectrum from the retrieved result is closer to the observed spectrum than the a priori spectrum, in Figure 3a. The RMSE of the spectral fit residuals decreases to 1.58 K from 6.27 K derived from the a priori spectrum. There is no evident systematic error in the fitting residuals from the retrieval. As seen in Figure 3b, the retrieval accurately captures the significant increase in the ozone partial columns in the UTLS region and is closer to the estimate from ERA5 data which are used as a reference. The total ozone column increases from 316.6 DU for the a priori profile to 412 DU for the retrieval profile. The total column from ERA5 value is 418.7 DU. The discrepancy of TOC between the retrieval and ERA5 is about 1.6%. This example illustrates the effectiveness of the O3 Figure 3. An example of retrievals using GIIRS spectra measured at 53.14°W, 145.17°E on 15 March 2023. (a) The spectral fit and associated residuals for the initial spectrum and the fitted spectrum; (b) comparison of the ozone partial column profiles from a priori, GIIRS retrievals and ERA5 data (in DU). Each dot represents the ozone partial column in each retrieval layer at a different pressure level. The comparison of the ozone profiles in volume mixing ratio (ppm) is shown in Figure S5a in Supporting Information S1. (c) The averaging kernel rows and (d) the a posteriori error of the retrieval derived from the a posteriori error matrix (left) and the variability defined for the a priori O3 profile (right). The dots in panel (c) represent diagonal elements of the AK matrix, associated with the partial columns in the retrieval layers. Journal of Geophysical Research: Atmospheres 10.1029/2024JD042292 LIU ET AL. 9 of 27 21698996, 2025, 8, D ow nloaded from https://agupubs.onlinelibrary.w iley.com /doi/10.1029/2024JD 042292 by E nvironm ent C anada, W iley O nline L ibrary on [09/01/2026]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense retrieval algorithm proposed in this paper. Furthermore, the average spectral fitting residuals for all retrievals during the 4‐month studied period after post‐processing are provided as Figure S4 in Supporting Information S1. It can be seen that most channels have residuals smaller than the predefined reference errors. The characteristics of the spectral fit residuals may be caused by the uncertainty in the spectroscopy data set or the forward model in the algorithm, which need further investigation. Figure 3c illustrates an example of the AK matrix from the retrieval, where each profile corresponds to a row of elements of the AK matrix. The vertical profile of the row sums of the AK matrix is provided in Figure S5b in Supporting Information S1. Extended results of AK matrices from additional retrievals are shown in Figure S6 in Supporting Information S1. It can be seen that the GIIRS observations are particularly sensitive to the ozone vertical distribution in the altitude from the mid‐troposphere to the lower‐stratosphere (from approximately 500 to 40 hPa). The maximal sensitivity is located from the tropopause to the lower stratosphere (about 150 hPa), where the atmospheric circulation and stratospheric intrusion of ozone mainly occur and drive the large variability (Archibald et al., 2020; Griffiths et al., 2020). This observation corresponds to a DOFS value of 0.95, a surface temperature of 269.0 K and a thermal contrast of 2.1 K. As highlighted in Clerbaux et al. (2009), DOFS values are mainly influenced by the surface temperature and thermal contrast. Better information content is expected in the lower troposphere from a greater positive or negative thermal contrast. However, ozone is mainly concentrated in the stratosphere, with only a minor proportion present in the lower troposphere. Moreover, GIIRS measurements demonstrate a relatively lower sensitivity to ozone change in the lower troposphere (>500 hPa), compared to IASI, and only a small amount of information can be obtained. One possible explanation is the relatively lower signal‐to‐noise‐ratio of the GIIRS instrument compared to IASI. Figure 3d illustrates the a posteriori error the uncertainty of the a priori O3 profile for this example retrieval. The a posteriori error is largest between 300 and 150 hPa, up to 9%, which is mainly due to the effect of variability of O3 close to the tropopause. The a posteriori error for the retrieval is significantly reduced at all altitudes compared to the a priori uncertainty, demonstrating that the GIIRS measurements provide valuable information. Figure 4 summarizes the retrieved results from GIIRS in September 2023 after post‐processing, including the TOC and the associated DOFS and the RMSE of spectral fitting residuals. All these results are monthly averaged into 0.5°× 0.5° grids. GIIRS has a total of 12 observation cycles per day with a period of two hours. Here, we only show 6 measurements at 2‐hr intervals as examples. The remaining results are shown in the Supporting Infor- mation S1. The blank areas in the figure represent data gaps, mainly due to the presence of clouds. From Figure 4, it can be seen that the TOCs range from about 220 to 380 DU with a typical latitudinal distribution in September 2023: high values are to be expected at high latitudes, rapidly decreasing at mid‐latitudes and further gradually decreasing toward the equator. This phenomenon is likely attributable to the Brewer‐Dobson circulation in the stratosphere, which transports O3 from the equatorial regions, where O3 is photochemically generated, to the lower stratosphere in the mid‐ and high‐latitude regions, where the longer residence time of ozone allows for its accumulation. The TOCs in the Tibetan Plateau region are also at low values close to the equator, which may be due to topographic factors (e.g., Bian et al., 2011). The diurnal variations of TOC are further described in Section 4.4. There is a clear spatial gradient for the DOFS values: the lower the latitude, the higher the DOFS, which changes from around 1.0 at high latitudes to about 1.2 at mid‐latitudes and approximately 1.4 at the equator. The dis- tribution is mainly driven by the surface temperature. In addition, there is also a slight variation in the DOFS values for the same area throughout the diurnal period due to the influence of the thermal contrast, although it is small. For land regions, DOFS is typically high during the daytime, peaking around midday, and low at night. For oceanic regions, DOFS is high in the morning and noon, decreases at dusk and gradually increases at night. For the nighttime observations, the DOFS values over oceans are significantly higher than those over land. This pattern is mainly caused by the fact that the sea surface temperature decreases more slowly than the land surface temperature due to a difference in heat capacity. The RMSE of spectral fitting residuals show a clear land‐ocean and latitude distribution, with large residuals in the observations over oceans and the tropical land regions, and small values in the middle and high latitudinal land regions. This characterization of the spectral fitting residual between land and ocean may be related to their difference in surface temperature which is associated with the spectral signal‐to‐noise‐ratio. We note that the quality of the spectral fit is not as good at the edges of the GIIRS observation domain. This may be due to the large zenith angles of the observation paths in these regions. Our radiative transfer model assumes a plane‐parallel atmosphere for simplicity. We filter out satellite observations with viewing zenith angles greater than 70° to reduce possible error related to slant path calculation. However, Journal of Geophysical Research: Atmospheres 10.1029/2024JD042292 LIU ET AL. 10 of 27 21698996, 2025, 8, D ow nloaded from https://agupubs.onlinelibrary.w iley.com /doi/10.1029/2024JD 042292 by E nvironm ent C anada, W iley O nline L ibrary on [09/01/2026]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense Figure 4. Overview of GIIRS O3 retrieval performances for hours 1–2, 5–6, 9–10, 13–14, 17–18, and 21–22 UTC averaged for the month of September 2023. These data have been filtered by the post‐processing filters. All the results are averaged monthly in 0.5° × 0.5° grids. These 3 panels are the TOCs retrievals (molecules/cm2), the DOFS, and the RMSE of spectral fitting residuals (K), respectively. Journal of Geophysical Research: Atmospheres 10.1029/2024JD042292 LIU ET AL. 11 of 27 21698996, 2025, 8, D ow nloaded from https://agupubs.onlinelibrary.w iley.com /doi/10.1029/2024JD 042292 by E nvironm ent C anada, W iley O nline L ibrary on [09/01/2026]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense observations close to the study area edge with larger viewing zenith angles may be more susceptible to contamination from clouds, which can introduce additional errors. 4.2. Comparisons With O3 From ERA5 Reanalysis Data Satellite retrievals are subject to observation spectral uncertainty and retrieval errors arising from the ill‐posed retrieval systems. It is, consequently, imperative that validation be conducted to evaluate the uncertainty of the satellite retrievals. This section presents a data evaluation of GIIRS O3 retrievals, conducted through a cross‐ comparison with ERA5 reanalysis data. ERA5 is the fifth generation of reanalysis data sets produced by the European Center for Medium‐Range Weather Forecasts. The ozone field in the ERA5 is generated from the assimilation of model simulations and satellite observations, which integrates satellite‐based and ground‐based ozone measurements from multi‐sources, including the OMI, GOME‐2 and Aura's Microwave Limb Sounder (MLS) instrument, and has undergone strict quality control. The ERA5 ozone data have been widely validated (e.g., Bernet et al., 2020; Marshall et al., 2022; Nerobelov et al., 2022) and utilized in various studies (e.g., Krzyścin, 2023; Orr et al., 2020; Zhang et al., 2022). First, we average the TOCs from ERA5 data and GIIRS retrievals in 0.5° × 0.5° grids, and compare the monthly mean, as shown in Figure S8 in Supporting Information S1 (and in 0.25° × 0.25° grids shown in Figure S9 in Supporting Information S1). We collect the closest spatial and temporal ERA5 data for each GIIRS retrieval. The TOCs from GIIRS and ERA5 show very similar spatial variations with season and latitude. As expected, the highest values of TOC are observed in high‐latitude regions, while the lowest values are typically found in the equatorial region for all months. This spatial pattern can be accounted for by the influence of large‐scale air circulation in the stratosphere. The process results in the gradual transport of ozone from equatorial regions to high‐latitude regions (Salawitch et al., 2023). The TOC varies significantly from month to month, especially at high latitudes, where it reaches a maximum in December and March, decreases considerably in July, and drops to a 4‐month low in September. This phenomenon is mainly caused by the fact that the ozone transport from the tropical regions to the Polar Regions is much stronger during the late autumn and winter months than it is during the summer and early autumn months. In addition, the apparently low TOC values in the Tibetan Plateau region are well observed from GIIRS during all the months studied, which corroborates the findings of Chen et al. (2017). Furthermore, we performed a spatial and temporal point‐by‐point collocation comparison of the TOCs from GIIRS and ERA5 for all months. For the comparison, it is necessary to take into account the vertical sensitivity of GIIRS retrieved results, which has been quantified by the AK matrix. First, we collect the ERA5 data closest to the locations and times of the GIIRS retrievals. Then we smoothed the ERA5 O3 profile using the corresponding AK matrix from GIIRS retrievals (Rodgers & Connor, 2003), written as: xsmooth = xa + A(xERA5 − xa) (5) where xERA5 represents the spatial interpolated ERA5 ozone profile, and xsmooth denotes the smoothed ERA5 profile by the AK matrix. Figure 5 illustrates the comparison results between the GIIRS retrievals and the smoothed ERA5 data. It also presents the correlation coefficient (r), RMSE and the fitting slope from linear regression. The comparison results show that GIIRS and ERA5 total ozone columns have high consistency, with r ranging from 0.96 to 0.99, the slopes from linear regression close to 1.0 and RMSEs ranging from 13.81 to 16.46 DU. The mean bias of TOCs from GIIRS and ERA5 is − 11.63 DU (∼− 3.9%), which has the same order of magnitude as that of IASI and GOME‐2 as shown by Boynard et al. (2009). In addition, there are a minor number of retrievals in the middle to high latitudinal zones where the TOCs are overestimated compared to the ERA5 data, especially in December and March, also shown in Figures S8 and S9 in Supporting Information S1. These overestimated retrievals are mainly located at the edges of the GIIRS observation domain. As described in Section 4.1, for these data, the viewing zenith angles are large and their spectral fittings are of low quality (see Figure 4), leading to a potentially poor quality of retrieval results. In December (spring) and March (winter), the signal‐to‐noise ratio of the GIIRS observations is low due to the low surface temperature, thus producing more anomalous points. These points could be excluded by more stringent filtering criteria. This comparison also shows an underestimation of the TOC from GIIRS compared to the ERA5 data, especially at low latitudes. Journal of Geophysical Research: Atmospheres 10.1029/2024JD042292 LIU ET AL. 12 of 27 21698996, 2025, 8, D ow nloaded from https://agupubs.onlinelibrary.w iley.com /doi/10.1029/2024JD 042292 by E nvironm ent C anada, W iley O nline L ibrary on [09/01/2026]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense To further illustrate the differences, the mean bias of TOC between GIIRS and ERA5 as a function of latitudes and months is shown in Figure 6a. The difference (GIIRS minus ERA5) shows apparently negative mean biases, with the maximum in the tropics in June and December at around − 7.5%, and the smallest in the high latitudes in June and September at about − 0.5%, close to the magnitude of the bias reported in Baek et al. (2023) based on ob- servations between GEMS and TROPOMI. A slight seasonal and latitudinal tendency for the mean bias can be Figure 5. Intercomparison of collocated GIIRS and ERA5 total ozone columns for December of 2022 and March, June, September of 2023. The correlation coefficient, RMSE, and fitting slope from linear regression (in red) are also indicated. The data points are color‐coded in accordance with the data number. The black dashed line indicates a slope of 1:1 for reference. Figure 6. The relative mean bias and the standard deviation of the bias (in percentage) for total ozone columns between (a) GIIRS and ERA5, (b) GIIRS and TROPOMI with latitude and months for December of 2022, and March, June, September of 2023. Journal of Geophysical Research: Atmospheres 10.1029/2024JD042292 LIU ET AL. 13 of 27 21698996, 2025, 8, D ow nloaded from https://agupubs.onlinelibrary.w iley.com /doi/10.1029/2024JD 042292 by E nvironm ent C anada, W iley O nline L ibrary on [09/01/2026]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense observed in the figure. In June and September, the closer to the equator, the higher the bias, which is similar to the distribution of the spectral fitting quality as shown in Figure 4; in March and December, the mean bias is at the maximum in the ranges 0–10°N and 40–50°N, and minimum in the ranges 20–30°N and 50–60°N. The systematic differences between GIIRS and ERA5 are not clear and require further investigations. Additionally, for better comparison, we introduce TROPOMI data and analyze the differences in TOC between TROPOMI and GIIRS with respect to latitude and month in Figure 6b. TROPOMI is a nadir‐viewing spectrometer that measures in the ultraviolet, visible, near‐infrared, and shortwave infrared spectral bands (Veefkind et al., 2012). It was launched aboard the Sentinel‐5 Precursor (S5P) satellite in October 2017. In this study, we use the Offline (OFFL) total ozone Level 2 data product of TROPOMI (processor version of 02.04.01 and 02.05.00) in comparisons (Lerot et al., 2024). A detailed description of the retrieval algorithm and a full validation of the TROPOMI operational ozone product can be found in (Garane et al., 2019). TROPOMI TOC are filtered to include only those with quality assurance values “qa_value” above 0.5 (recommended in the Product Readme File). The results indicate that the magnitude of the bias between GIIRS and TROPOMI is comparable to that between GIIRS and ERA5. In most cases, GIIRS tends to underestimate the values. However, in high‐latitude regions during June and September, GIIRS shows obvious higher TOC values than TROPOMI. The reasons for these discrepancies are still under investigation.We also provide the point‐by‐point comparison results between GIIRS and TROPOMI in Figure S10 in Supporting Information S1. In general, our results show that, as a new data set, the GIIRS ozone retrievals with high spatio‐temporal resolution can provide information on the continuous changes in the TOCs over East Asia. 4.3. Comparisons With IASI Ozone Retrievals In this section, the Level 2 O3 data products from IASI on board Metop‐B (IASI Level 2: Product Guide) are used to compare with the total ozone columns retrieved from GIIRS. IASI/Metop‐B, with equatorial overpassing times at 09:30 and 21:30 each day, monitors the composition of the atmosphere on a global scale with a footprint of 12 km at nadir. The current implementation of the ozone retrieval algorithm for IASI in the EUMETSAT pro- cessing facility is the FORLI‐O3 v20151001. The algorithm was originally proposed by Hurtmans et al. (2012) and has undergone a series of developments and extensive validation over time (Boynard et al., 2016). Two different comparisons are described below. First, we randomly select the retrievals from one local daytime and one local nighttime during the study period as examples for a comparative analysis of the regionally averaged TOC from GIIRS and IASI. All the available retrievals are averaged in 0.5° × 0.5° grids thus assuming the temporal variation in O3 columns is negligible. The results are presented in Figure 7. The mean total ozone columns from GIIRS and IASI exhibit a high degree of correlation, with r of 0.94 and slopes from linear regression ranging from 1.0 to 1.02. This example is also a good illustration of the difference in the observation modes between the two satellites. GIIRS provides observations in East Asia for every 2 hours, while IASI ob- servations are in narrower swaths with a twice‐daily coverage. As a result, the available observations from IASI are far fewer than GIIRS in East Asia. Furthermore, a point‐by‐point collocation comparison of the TOC retrieved from GIIRS and IASI is conducted. The observations are regarded as collocated if the distance between pixel centers is within 20 km and if the time difference between observations is less than 1 hr. Note that the footprint sizes are 12 km for both GIIRS and IASI at nadir and can get to much larger size off nadir. All the retrievals satisfy the aforementioned conditions are employed for the purpose of comparison and presented in Figure 8. The comparison demonstrates a good degree of correlation, with r in the range of 0.83–0.98, slopes between 0.96 and 1.01 and RMSEs between 13.75 and 21.71 DU for different months. The largest bias occurs in the nighttime in December, which is expected since low temperatures are not favorable for thermal infrared observations. Moreover, we can observe that at high latitudes (corresponding to high TOC values) the IASI TOCs are comparatively larger than those from GIIRS in winter (i.e., December). A comparable result is presented by Lee et al. (2019), indicating that the Advanced Himawari Imager (AHI) TOC values are obviously smaller than the IASI TOCs in January (winter) at high latitudes (50°N–70°N). Further studies are needed to comprehensively investigate the discrepancies observed in TOC retrievals between IASI and GIIRS at various locations and time points. The causes may include: (a) The differences in observation modes and geometries, indicating that distinct air masses are sounded, with observations potentially influenced by varying degrees of cloud contamination; (b) the a priori information employed by the two retrieval algorithms is distinct; (c) the two instruments have different response functions, different spectral resolution and signal to Journal of Geophysical Research: Atmospheres 10.1029/2024JD042292 LIU ET AL. 14 of 27 21698996, 2025, 8, D ow nloaded from https://agupubs.onlinelibrary.w iley.com /doi/10.1029/2024JD 042292 by E nvironm ent C anada, W iley O nline L ibrary on [09/01/2026]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense noise; (d) the spatial and temporal differences of the observation footprints. The footprint centers and the observation times of the collocated data do not exactly coincide. 4.4. Comparisons With OMI Total Ozone Measurements OMI is a near ultraviolet‐visible (UV/VIS) spectrometer on board NASA's EOS Aura satellite, which was launched in July 2004 (Levelt et al., 2006). The OMI instrument measures solar radiation backscattered by the Earth's atmosphere and surface within the wavelength range of 270–500 nm. The measurement swath width of OMI is 2,600 km, with a ground resolution of 13 km × 24 km. The local equatorial crossing time is 1:45 PM. In this study, we use the OMI TOC product OMTO3 V003, which is retrieved from the enhanced TOMS version‐8 algorithm developed by NASA (Bhartia & Wellemeyer, 2002). This algorithm is primarily based on retrievals from four wavelengths: 313, 318, 331, and 360 nm. The OMI TOC data have been extensively validated by the measurements from Brewer, Dobson, and Pandora instruments (e.g., Balis et al., 2007; Garane et al., 2019). Figure 9 shows the results of a point‐by‐point comparison of GIIRS and OMI TOC over the 4‐month study period. For this comparison, OMI TOC data are filtered to include only those with quality flag equal to 0 or 1 and cloud fraction of less than or equal to 0.3. Collocated observations are identified when the pixel center distance between the two observations is less than 20 km and the time difference is less than 1 hr. The results show that the two data sets exhibit good correlation, with r ranging from 0.91 to 0.99, linear fitting slopes between 0.97 and 1.02, and RMSEs from 12.4 to 14.1 DU across different months. Their difference can be attributed to different spectral observations and instrument characteristics. UV/VIS observations of OMI are more sensitive to the surface, while IR observations of GIIRS are more sensitive to the upper troposphere and lower stratosphere. In addition, the two instruments have different nadir observation footprints (12 km for GIIRS and 13 × 24 km for OMI), and observation geometries that lead to different sampling air masses. Figure 7. Intercomparison of gridded total ozone columns between GIIRS and IASI on two specific times: Daytime on 02 December 2022 (a) and nighttime on 15 June 2023 (b). The observation hours are 0–6 hr UTC for (a), and 11–17 hr UTC for (b), illustrated in Figure S11 in Supporting Information S1. The retrievals are averaged into 0.5° × 0.5° grids. The scatter diagrams illustrate the comparison of averaged column data between GIIRS and IASI. The black dashed line indicates a slope of 1 for reference. Journal of Geophysical Research: Atmospheres 10.1029/2024JD042292 LIU ET AL. 15 of 27 21698996, 2025, 8, D ow nloaded from https://agupubs.onlinelibrary.w iley.com /doi/10.1029/2024JD 042292 by E nvironm ent C anada, W iley O nline L ibrary on [09/01/2026]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense Figure 8. Intercomparison of collocated TOCs between GIIRS and IASI for December of 2022, and March, June, September of 2023. If the difference between the observation centers is less than 20 km and the difference between the observation hours is less than 1 hr, they are considered to be collocated. The linear regression and data color‐coding are similar to Figure 5. Journal of Geophysical Research: Atmospheres 10.1029/2024JD042292 LIU ET AL. 16 of 27 21698996, 2025, 8, D ow nloaded from https://agupubs.onlinelibrary.w iley.com /doi/10.1029/2024JD 042292 by E nvironm ent C anada, W iley O nline L ibrary on [09/01/2026]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense 4.5. Comparisons With Ozonesonde Measurements To further analyze the GIIRS O3 retrievals, high resolution O3 vertical profiles, obtained through ozonesondes, have been used. Ozonesonde observations are obtained from the World Ozone and Ultraviolet Radiation Data Center (WOUDC) archive. Unfortunately, only two stations have complete data during the study period and can be used in the comparison, with their locations shown in Figure 10a. The coincidence criteria for the comparison are of a 100 km search radius and ±2h, with the best matched GIIRS observations being selected. The result is a total of 20 validations of GIIRS retrievals using sonde measurements. All these sonde measurements are using the electrochemical concentration cell (ECC) method, which has a generally good accuracy (±3%∼5%) (Deshler et al., 2008). Figure 10b shows results of the comparison. As the ozonesonde provides measurements only up to about 30–35 km, we only calculate the ozone columns from surface to about 10 hPa for both GIIRS and ozo- nesonde (also referred to as “TOC” in the figure). The two data are of good agreement, with r of 0.97, slope of 1.01 and RMSE of 11.18 DU. Furthermore, examples of intercomparison of the O3 profiles obtained by the GIIRS instrument and the ozonesonde (Tsukuba) over a 4‐month period are provided in Figure 10c. For meaningful comparison, the partial ozone columns from ozonesonde measurements are integrated corresponding to the GIIRS vertical pressure layers. It can be observed that there is a high degree of correlation between the retrieval profiles and the sonde profiles. The ozone variations are nicely captured by GIIRS (seen from the profiles in March and September, the sonde profiles are far from the a priori profile), particularly in the UTLS region, also mentioned in Figure 3. The results also show the retrieval performance is not good in the lower troposphere. This is to be expected because GIIRS observations to ozone are not sensitive in that region. Other comparisons of ozone profiles between GIIRS and the two stations are shown in Figures S12 and S13 in Supporting Information S1. The comparison results that the AK matrix is considered are in the Supplement. Figure 9. Intercomparison of collocated TOCs between GIIRS and OMI in December of 2022, and March, June, September of 2023. The collocation criteria, the linear regression and data color‐coding are similar to Figure 8. Journal of Geophysical Research: Atmospheres 10.1029/2024JD042292 LIU ET AL. 17 of 27 21698996, 2025, 8, D ow nloaded from https://agupubs.onlinelibrary.w iley.com /doi/10.1029/2024JD 042292 by E nvironm ent C anada, W iley O nline L ibrary on [09/01/2026]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense Figure 10. Intercomparison of ozone data between GIIRS and ozonesonde. (a) Geographic locations of the two ozonesonde stations utilized for validation. (b) Scatter plots for the GIIRS and ozonesonde total ozone columns. The total ozone columns here have been computed for the region from surface to about 10 hPa. The blue dots represent observations from King's Park station and the orange dots represent those from Tsukuba. If measurements are within 100 km and the measurement hours are within 2 hr, they are regarded as collocated. The correlation coefficient, RMSE, and fitting slopes from linear regression are also indicated. (c) Examples of intercomparison of the ozone profiles from GIIRS retrievals, a priori, Tsukuba measurements, and ERA5 data for December of 2022, and March, June, September of 2023. Journal of Geophysical Research: Atmospheres 10.1029/2024JD042292 LIU ET AL. 18 of 27 21698996, 2025, 8, D ow nloaded from https://agupubs.onlinelibrary.w iley.com /doi/10.1029/2024JD 042292 by E nvironm ent C anada, W iley O nline L ibrary on [09/01/2026]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense 5. The Diurnal to Seasonal Variations of Ozone Columns From GIIRS Retrievals FY‐4B/GIIRS provides continuous diurnal monitoring of the total ozone column over East Asia with an un- precedented 2‐hr temporal resolution, which will help improve our understanding of ozone. In this section, we explore the potential of GIIRS observations to track the changes in the total ozone column over different temporal scales from diurnal to seasonal, which is achieved by comparing the temporal and latitudinal characteristics of the total ozone column from GIIRS, Pandora, and ERA5 data. Pandora is a ground‐based spectrometer system that measures the direct sunlight from 280 to 525 nm and provides total columns of ozone and other trace gases (Herman et al., 2009; Tzortziou et al., 2012). The Pandora TOC retrieval algorithm uses a modified Differential Optical Absorption Spectroscopy (DOAS) method with spectra in the 305–328.6 nm wavelength range. Technical details of the algorithm are given in Cede (2021). The original Pandora total column ozone data (processing version rout0) has good high precision and accuracy, but it shows a seasonal dependency and a bias when compared to observations from Brewer spectrophotometers, primarily due to its sensitivity to stratospheric temperature (Zhao et al., 2016). The latest Pandora data set (rout2) demonstrates improved agreement with Brewer measurements, with the temperature dependency issue largely resolved. In this study, we collect observations from six Pandora sites at different latitudes and longitudes within the GIIRS coverage during the study period. The TOC observations were processed by using the Blick Software Suite version 1.8. The site information is given in Table S2 in Supporting Information S1. We exclude the cloud‐ and aerosol‐contaminated Pandora measurements and use data that met the following conditions: Normalized root mean square of the weighted spectral fitting residuals <0.05%; the estimated error in TOC < 2 DU; the solar zenith angle<75° and L2 data quality flag= 0 or 1 or 10 or 11 (indicating high or medium quality data). Figure 11 and Figure S17 in Supporting Information S1 show a comparison of the time series of total ozone columns from GIIRS, Pandora, and ERA5 data at the six Pandora sites over the study period. For this analysis, only GIIRS observations within 50 km from the Pandora sites are included; for each GIIRS observation, the temporally and spatially closest ERA5 data are selected; and the Pandora observations are averaged on an hourly basis. It should be noted that no high‐quality data are available at the Yokosuka station during the experimental period. The results indicate that, overall, there is a good consistency in TOC across different sites (corresponding to various latitudes and longitudes) and seasons among the different data sets. FY‐4B/GIIRS effectively captures the dy- namics of the total ozone column within seasons as well as inter‐seasonally (also seen in Figures S8 and S9 in Supporting Information S1). The results from different stations show that the temporal variation of TOC has significant seasonal and latitudinal effects. The variation of TOC increases with latitude and is greater in spring and winter than that in summer and autumn. Notably, GIIRS TOC is overestimated compared to the Pandora observations at some stations in winter, indicating a certain systematic bias. This discrepancy may be attributed to the fact that thermal infrared observations can be affected by surface temperatures. The lower surface temper- atures during winter can lead to poorer signal‐to‐noise ratios in satellite observations, thereby reducing the ac- curacy of the retrievals. In addition, we compare the diurnal variations of TOC observed by GIIRS and Pandora to assess the ability of GIIRS observations in tracking the diurnal variations of TOC. This comparison may be limited because Pandora only provides daytime measurements. The processed Pandora data have hourly resolution, whereas the GIIRS observations are of every 2 hours. To properly compare the diurnal trends in TOC between the two data, we select only those cases where GIIRS and Pandora observations overlapped at least four times during a day. The results are shown in the Figures S18–S23 in Supporting Information S1. The data show that the maximum diurnal variation of TOC occurs at Dakazadfad and Beijing stations (the two sites with the highest latitude) during spring and winter (up to 50 DU), as shown in Figures S18 and S19 in Supporting Information S1. Over these two sites, the diurnal variations of TOC shown by the GIIRS and Pandora observations are highly consistent on most of the dates. This consistency is also seen in most of the results during spring and winter in Figures S20–S22 in Sup- porting Information S1. When the diurnal variation of TOC is relatively large (e.g., >20 DU), GIIRS captures the diurnal variations relatively well. However, there are also some instances (e.g., Figure S23 in Supporting In- formation S1 for the Dhaka site), where the diurnal variations of TOC are small (less than 15 DU), GIIRS and Pandora do not match well. This discrepancy may be due to uncertainties in the satellite observations, as well as the differences in the detection sensitivity based on ultraviolet and infrared observation methods. The infrared observing mode is less sensitive to lower troposphere ozone. In particular, the information from GIIRS Journal of Geophysical Research: Atmospheres 10.1029/2024JD042292 LIU ET AL. 19 of 27 21698996, 2025, 8, D ow nloaded from https://agupubs.onlinelibrary.w iley.com /doi/10.1029/2024JD 042292 by E nvironm ent C anada, W iley O nline L ibrary on [09/01/2026]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense Figure 11. Time series of TOC from GIIRS retrievals, ERA5 data and Pandora measurements in December of 2022 and March, June, September of 2023 at three sites: (a) Dakazadfad; (b) Beijing; (c) Seoul‐SNU. Pandora measurements are hourly averaged. The GIIRS data within 50 km from the sites are used. The closest ERA5 data in space and time to the GIIRS data are used. Journal of Geophysical Research: Atmospheres 10.1029/2024JD042292 LIU ET AL. 20 of 27 21698996, 2025, 8, D ow nloaded from https://agupubs.onlinelibrary.w iley.com /doi/10.1029/2024JD 042292 by E nvironm ent C anada, W iley O nline L ibrary on [09/01/2026]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense Figure 12. (a) Hourly ozone profiles and PV isopleths from ERA5 reanalysis data over the Beijing site in March 2023. PV values of 2 are shown with black line and PV values of 6 with red line. (b) Time‐series of TOC from ERA5 data and Pandora measurements at Beijing site. (c) The temporal variation of the ozone partial column for different air pressure layers, with the data at 0:00 on March 1 as a reference, based on ERA5 data. Journal of Geophysical Research: Atmospheres 10.1029/2024JD042292 LIU ET AL. 21 of 27 21698996, 2025, 8, D ow nloaded from https://agupubs.onlinelibrary.w iley.com /doi/10.1029/2024JD 042292 by E nvironm ent C anada, W iley O nline L ibrary on [09/01/2026]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense observations is concentrated only in the region approximately from 500 to 40 hPa, making it difficult to detect small variations in ozone beyond this altitude range. Further, we analyze the causes of the diurnal variations of TOC captured by GIIRS. Figure 12a based on the ERA5 data over the Beijing site in March 2023 illustrates the multiple exchange processes between ozone‐rich strato- spheric air and ozone‐poor tropospheric air. As a parameter to characterize the boundary between the troposphere and the stratosphere, the isopleths of the potential vortex (PV) are also plotted,where the black line indicates PV= 2 and the red line indicates PV= 6. Figure 12b shows the temporal variation of TOC fromPandora andERA5over the Beijing site, indicating a good agreement. Figure 12c displays the temporal variation of the ozone partial column for different air pressure layers, based on the ERA5 data. It reveals the contribution of ozone variations in the vertical direction to the variation of TOC. Previous research has described the effect of air mixing between the stratosphere and troposphere on the TOC variability (e.g., Stohl et al., 2003). Atmospheric transport between stratospheric regions with high and low ozone levels can also cause a redistribution of ozone molecules (e.g., Salby & Call- aghan, 1993). For example, around March 13, we can see that the TOC shows dramatic successive increases and decreases (Figure 12b), corresponding to the significant changes in the ozonemixing ratio and the altitude of the PV isopleths (Figure 12a). Specifically, on March 12, the isopleth of PV = 2 reaches to as low as 500 hPa, with high‐ concentration ozone flowing into the troposphere, leading to a rapid increase in TOC (Figures 12b and 12c). On March 18 as shown in Figure 12c, an increment of ozone values was observed at 300 hPa and even lower regions, indicating the intrusion of stratospheric air. At the same time, a significant decrease in ozone values is observed in the region of 100–40 hPa. The combined effect results in a relatively small change in TOC value. In addition, Weather systems such as tropical cyclones and typhoons can also induce dramatic changes in total ozone columns (e.g., Midya et al., 2012; Zou &Wu, 2005), along with photochemical factors that have been studied extensively (e.g., Chapman, 1930; Crutzen, 1970; Johnston, 1971; Stolarski & Cicerone, 1974). Many studies have discussed the diurnal variation of stratospheric ozone (e.g., Sakazaki et al., 2013; Schanz et al., 2014, 2021). In particular, Sakazaki et al. (2013) confirmed through observations and simulations that the variations in the lower stratosphere (20–30 km) are largely driven by atmospheric dynamics, while the variations at 30–45 km are largely caused by photochemical factors. However, as described in Section 4.1, the observation sensitivity of GIIRS to the ozone peaks in the region from the mid‐troposphere to the lower stratosphere (approximately 500–40 hPa, <30 km), GIIRS may be difficult to capture ozone variations in higher altitude regions. As a result, the diurnal variation of TOC observed by GIIRS is dominated by atmospheric dynamics. In addition to the research at specific sites, we also calculate the variations in TOC from daily to seasonal at different latitudes in the whole GIIRS coverage, as shown in Figure 13 and Figure S24 in Supporting Infor- mation S1. The results indicate that the dynamic changes in the daily average TOC displayed by GIIRS are highly consistent with those of ERA5. Similar to diurnal variations, the maximum variability of the daily mean TOC also occurs during spring at high latitudes (50°N–60°N), reaching approximately 50 DU. In this region, we can clearly see the trends of daily mean TOC values throughout different seasons. The total ozone columns slightly increase in December, peak in March and decline significantly in late March. Ozone then slowly decreases in June and reaches the lowest values in September and remains basically stable. It's worth mentioning that the seasonal variations in TOC in the tropics are the opposite of those at high latitudes. In the tropical regions, ozone column levels are high in northern summer and autumn months, but exhibit a decline during the northern winter and spring. These features are the result of a combination of seasonal variations in solar radiation and large‐scale ozone transport (Chen & Nunez, 1998; Tung & Yang, 1988). High latitude regions experience significant sea- sonal variations in solar radiation and ozone transport, while tropical regions are relatively stable with smaller seasonal variations. This leads to strong fluctuations in the total ozone column during the spring in high latitude. This has also been discussed in Section 4.2. Overall, our results demonstrate the ability of GIIRS in capturing diurnal to seasonal variations in the ozone columns. The GIIRS ozone retrievals with high spatiotemporal resolution provide an invaluable source of ob- servations to contribute to a better understanding of ozone dynamics in East Asia. 6. Conclusions In this paper, we develop an ozone retrieval algorithm for FY‐4B/GIIRS in East Asia, which achieves high spatio‐ temporal resolution monitoring of TOC (and ozone vertical profile, to some extent) over East Asia, and dem- onstrates the complex dynamics of ozone from diurnal to seasonal variations. Our algorithm uses the ozone Journal of Geophysical Research: Atmospheres 10.1029/2024JD042292 LIU ET AL. 22 of 27 21698996, 2025, 8, D ow nloaded from https://agupubs.onlinelibrary.w iley.com /doi/10.1029/2024JD 042292 by E nvironm ent C anada, W iley O nline L ibrary on [09/01/2026]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense absorption micro‐window (1,025–1,075 cm− 1), selects four months for retrieval experiments (December 2022, and March, June and September 2023, representing different seasons), and obtains ozone maps of the East Asian region at a resolution of 12 km × 12 km, every 2 hr. Detailed comparisons with IASI and OMI satellite mea- surements, ERA5 reanalysis data, ozonesonde and Pandora measurements confirm the accuracy and reliability of the GIIRS ozone products, with results showing high correlation coefficients and low RMSEs. Most DOFS values in observations range from 0.8 to 1.4 and vary with latitude. The detection sensitivity of GIIRS peaks in the UTLS region and is not sensitive near the surface and above the middle stratosphere. By comparing the TOCs from GIIRS, Pandora and ERA5, we confirm that FY‐4B/GIIRS can capture the ozone variations from diurnal to seasonal timescales well, especially in the middle and high latitudes. Due to the sensitivity of GIIRS, we infer that the ozone variation observed by GIIRS is dominated by atmospheric dynamics. In summary, this study successfully demonstrates the potential application of GIIRS in ozone monitoring and provides a solid scientific foundation and technical support for the future utilization of geostationary satellites in atmospheric component monitoring. Our research results not only provide a new perspective for understanding regional atmospheric chemical processes, but also contribute valuable data resources to global climate change research. This paper represents a significant initial step toward the monitoring of atmospheric ozone from a constellation of geostationary thermal infrared sounders. These include the existing and forthcoming GIIRS on board the FY‐4 series over Asia, the forthcoming European geostationary infrared sounder (IRS) on board Meteosat Third Generation (MTG) over Europe and other planned missions. In addition, the integration of GIIRS with other existing and planned geostationary missions enables a global satellite constellation. These missions include the Figure 13. The daily mean TOC values for each day in different months and latitude zones. The TOCs are from GIIRS, ERA5 and the smoothed ERA5 data based on the averaging kernel matrix from GIIRS. The latitudinal bands are (a) 50°N–60°N, (b) 40°N–50°N, and (c) 30°N–40°N, respectively. Journal of Geophysical Research: Atmospheres 10.1029/2024JD042292 LIU ET AL. 23 of 27 21698996, 2025, 8, D ow nloaded from https://agupubs.onlinelibrary.w iley.com /doi/10.1029/2024JD 042292 by E nvironm ent C anada, W iley O nline L ibrary on [09/01/2026]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense South Korean Geostationary Environment Monitoring Spectrometer (GEMS; Kim et al., 2020), the European Space Agency's upcoming Sentinel‐4 mission, and NASA's Tropospheric Emissions: Monitoring of Pollution (TEMPO; Zoogman et al., 2017). Distinct geostationary field‐of‐views are connected together by the constel- lation of polar orbiting satellites used as traveling standards. This global constellation is expected to provide richer and more accurate information for research in atmospheric chemistry, air quality, and climate change on a global scale. Data Availability Statement The ozone retrieval data from FY‐4B/GIIRS in this study are available on Zenodo (Zeng & Liu, 2025). The FY‐ 4B/GIIRS Level 1 data are available to the general public from the FengYun Satellite Data Centre (National Satellite Meteorological Center, 2024). The surface emissivity data sets can be downloaded from the Global Infrared Land Surface Emissivity: UW‐Madison Baseline Fit Emissivity Database (Seemann et al., 2008). The ECMWF ERA5 reanalysis data sets are available from the Copernicus Climate Data Store (Hersbach et al., 2020). The ECMWF atmospheric composition data sets are available from the Copernicus Atmosphere Data Store (Copernicus Atmosphere Monitoring Service, 2021). IASI is a joint mission of EUMETSAT and the Centre National d’Etudes Spatiales (CNES, France). The IASI ozone products are available from the Aeris website (Aeris, 2024). The ozonesonde data are available from the World Ozone and Ultraviolet Data Centre (WOUDC, 2024). Pandora data are available at Pandora Global Network (PGN, 2024). OMI OMTO3 data are available from the GES DISC (Bhartia, 2005). TROPOMI ozone products are available via the Copernicus Data Space Ecosystem (Copernicus Sentinel‐5P, 2024). References Aeris. (2024). The IASI O3 products processed with FORLI‐O3 [Dataset]. IASI. Retrieved from http://iasi.aeris‐data.fr/O3/ Anderson, G. P., Barth, C. A., Cayla, F., & London, J. (1969). Satellite observations of the vertical ozone distribution in the upper stratosphere. Annales Geophysicae, 25, 341–345. Antón, M., Serrano, A., Cancillo, M. L., & García, J. A. (2008). Total ozone and solar erythemal irradiance in Southwestern Spain: Day‐to‐day variability and extreme episodes. Geophysical Research Letters, 35(20), 2008GL035290. https://doi.org/10.1029/2008GL035290 Archibald, A. T., Neu, J. L., Elshorbany, Y. F., Cooper, O. R., Young, P. J., Akiyoshi, H., et al. (2020). Tropospheric ozone assessment report. Elementa: Science of the Anthropocene, 8(1), 034. https://doi.org/10.1525/elementa.2020.034 Baek, K., Kim, J. H., Bak, J., Haffner, D. P., Kang, M., & Hong, H. (2023). Evaluation of total ozone measurements from Geostationary Environmental Monitoring Spectrometer (GEMS). Atmospheric Measurement Techniques, 16(22), 5461–5478. https://doi.org/10.5194/amt‐ 16‐5461‐2023 Balis, D., Kroon, M., Koukouli, M. E., Brinksma, E. J., Labow, G., Veefkind, J. P., & McPeters, R. D. (2007). Validation of ozone monitoring instrument total ozone column measurements using brewer and Dobson spectrophotometer ground‐based observations. Journal of Geophysical Research, 112(D24), 2007JD008796. https://doi.org/10.1029/2007JD008796 Beer, R., Glavich, T. A., & Rider, D. M. (2001). Tropospheric emission spectrometer for the Earth observing system’s aura satellite. Applied Optics, 40(15), 2356–2367. https://doi.org/10.1364/AO.40.002356 Bernet, L., Boyd, I., Nedoluha, G., Querel, R., Swart, D., & Hocke, K. (2020). Validation and trend analysis of stratospheric ozone data from ground‐based observations at Lauder, New Zealand. Remote Sensing, 13(1), 109. https://doi.org/10.3390/rs13010109 Bhartia, P. K. (2005). OMI/Aura Ozone(O3) total column 1‐orbit L2 swath 13x24 km V003 [Dataset]. Goddard Earth Sciences Data and In- formation Services Center (GES DISC). https://doi.org/10.5067/AURA/OMI/DATA2024 Bhartia, P. K., & Wellemeyer, C. (2002). OMI TOMS‐V8 total O3 algorithm, algorithm theoretical baseline document: OMI ozone products. NASA Goddard Space Flight Center. Bian, J., Yan, R., Chen, H., Lü, D., & Massie, S. T. (2011). Formation of the summertime ozone valley over the Tibetan Plateau: The Asian summer monsoon and air column variations. Advances in Atmospheric Sciences, 28(6), 1318–1325. https://doi.org/10.1007/s00376‐011‐ 0174‐9 Boynard, A., Clerbaux, C., Coheur, P.‐F., Hurtmans, D., Turquety, S., George, M., et al. (2009). Measurements of total and tropospheric ozone from IASI: Comparison with correlative satellite, ground‐based and ozonesonde observations. Atmospheric Chemistry and Physics, 9(16), 6255–6271. https://doi.org/10.5194/acp‐9‐6255‐2009 Boynard, A., Hurtmans, D., Garane, K., Goutail, F., Hadji‐Lazaro, J., Koukouli, M. E., et al. (2018). Validation of the IASI FORLI/EUMETSAT ozone products using satellite (GOME‐2), ground‐based (Brewer–Dobson, SAOZ, FTIR) and ozonesonde measurements. Atmospheric Measurement Techniques, 11(9), 5125–5152. https://doi.org/10.5194/amt‐11‐5125‐2018 Boynard, A., Hurtmans, D., Koukouli, M. E., Goutail, F., Bureau, J., Safieddine, S., et al. (2016). Seven years of IASI ozone retrievals from FORLI: Validation with independenttotal column and vertical profile measurements. Atmospheric Measurement Techniques, 9(9), 4327–4353. https://doi.org/10.5194/amt‐9‐4327‐2016 Cede, A. (2021). Manual for Blick software suite version 1.8. Retrieved from https://www.pandonia‐global‐network.org/wp‐content/uploads/ 2021/05/BlickSoftwareSuite‐Manual_v1‐8‐3.pdf Chapman, S. (1930). A theory of upper atmospheric ozone (Vol. 3, pp. 103–125). Memoirs of the Royal Meteorological Society. Chen, D., & Nunez, M. (1998). Temporal and spatial variability of total ozone in Southwest Sweden revealed by two ground‐based instruments. International Journal of Climatology, 18(11), 1237–1246. https://doi.org/10.1002/(SICI)1097‐0088(199809)18:11<1237::AID‐JOC285>3.0. CO;2‐6 Chen, S. B., Zhao, L., & Tao, Y. L. (2017). Stratospheric ozone change over the Tibetan Plateau. Atmospheric Pollution Research, 8(3), 528–534. https://doi.org/10.1016/j.apr.2016.11.007 Acknowledgments We would like to thank Dr. Masoud Ghahremanloo and Dr. Wuhu Feng for stimulating discussions on our work. We are grateful for the financial support from the National Natural Science Foundation of China (Grant 42275142 and 12292981) and the National Key R&D Program of China (Grant 2022YFA1003801). This work was supported by the High‐ performance Computing Platform of Peking University. The research conducted at the National Satellite Meteorological Center (NSMC) was supported by NSMC of China Meteorological Administration (CMA) under the program of Calibration Technology Development and Level‐1 Data Production for the Hyperspectral Imaging and Sounding Instruments onboard FY‐3E and FY‐4B Satellites (FY‐ APP‐2021.0507). We thank the PI(s), support staff and funding for establishing and maintaining the Dakazadfad, Beijing‐ RADI, Seoul‐SNU, Ulsan, Yokosuka and Dhaka sites of the PGN data and the King's Park and Tsukuba sites of the ozonesonde data used in this investigation. The PGN is a bilateral project supported with funding from NASA and ESA. Journal of Geophysical Research: Atmospheres 10.1029/2024JD042292 LIU ET AL. 24 of 27 21698996, 2025, 8, D ow nloaded from https://agupubs.onlinelibrary.w iley.com /doi/10.1029/2024JD 042292 by E nvironm ent C anada, W iley O nline L ibrary on [09/01/2026]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense http://iasi.aeris-data.fr/O3/ https://doi.org/10.1029/2008GL035290 https://doi.org/10.1525/elementa.2020.034 https://doi.org/10.5194/amt-16-5461-2023 https://doi.org/10.5194/amt-16-5461-2023 https://doi.org/10.1029/2007JD008796 https://doi.org/10.1364/AO.40.002356 https://doi.org/10.3390/rs13010109 https://doi.org/10.5067/AURA/OMI/DATA2024 https://doi.org/10.1007/s00376-011-0174-9 https://doi.org/10.1007/s00376-011-0174-9 https://doi.org/10.5194/acp-9-6255-2009 https://doi.org/10.5194/amt-11-5125-2018 https://doi.org/10.5194/amt-9-4327-2016 https://www.pandonia-global-network.org/wp-content/uploads/2021/05/BlickSoftwareSuite-Manual_v1-8-3.pdf https://www.pandonia-global-network.org/wp-content/uploads/2021/05/BlickSoftwareSuite-Manual_v1-8-3.pdf https://doi.org/10.1002/(SICI)1097-0088(199809)18:11%3C1237::AID-JOC285%3E3.0.CO;2-6 https://doi.org/10.1002/(SICI)1097-0088(199809)18:11%3C1237::AID-JOC285%3E3.0.CO;2-6 https://doi.org/10.1016/j.apr.2016.11.007 Clerbaux, C., Boynard, A., Clarisse, L., George, M., Hadji‐Lazaro, J., Herbin, H., et al. (2009). Monitoring of atmospheric composition using the thermal infrared IASI/MetOp sounder. Atmospheric Chemistry and Physics, 9(16), 6041–6054. https://doi.org/10.5194/acp‐9‐6041‐2009 Clough, S. A., Shephard, M. W., Mlawer, E. J., Delamere, J. S., Iacono, M. J., Cady‐Pereira, K., et al. (2005). Atmospheric radiative transfer modeling: A summary of the AER codes. Journal of Quantitative Spectroscopy and Radiative Transfer, 91(2), 233–244. https://doi.org/10. 1016/j.jqsrt.2004.05.058 Cohen, A. J., Brauer, M., Burnett, R., Anderson, H. R., Frostad, J., Estep, K., et al. (2017). Estimates and 25‐year trends of the global burden of disease attributable to ambient air pollution: An analysis of data from the Global Burden of Diseases Study 2015. The Lancet, 389(10082), 1907–1918. https://doi.org/10.1016/S0140‐6736(17)30505‐6 Copernicus Atmosphere Monitoring Service. (2021). CAMS global atmospheric composition forecasts [Dataset]. Copernicus Atmosphere Monitoring Service (CAMS) Atmosphere Data Store. https://doi.org/10.24381/04a0b097 Copernicus Sentinel‐5P. (2024). TROPOMI Level 2 ozone total column products Version 2.x [Dataset]. European Space Agency. https://doi.org/ 10.5270/S5P‐ft13p57 Crutzen, P. J. (1970). The influence of nitrogen oxides on the atmospheric ozone content. Quarterly Journal of the Royal Meteorological Society, 96(408), 320–325. https://doi.org/10.1002/qj.49709640815 Crutzen, P. J. (1974). Photochemical reactions initiated by and influencing ozone in unpolluted tropospheric air. Tellus, 26(1–2), 47–57. https:// doi.org/10.1111/j.2153‐3490.1974.tb01951.x Cuesta, J., Eremenko, M., Liu, X., Dufour, G., Cai, Z., Höpfner, M., et al. (2013). Satellite observation of lowermost tropospheric ozone by multispectral synergism of IASI thermal infrared and GOME‐2 ultraviolet measurements over Europe. Atmospheric Chemistry and Physics, 13(19), 9675–9693. https://doi.org/10.5194/acp‐13‐9675‐2013 Deshler, T., Mercer, J. L., Smit, H. G. J., Stubi, R., Levrat, G., Johnson, B. J., et al. (2008). Atmospheric comparison of electrochemical cell ozonesondes from different manufacturers, and with different cathode solution strengths: The Balloon Experiment on Standards for Ozone- sondes. Journal of Geophysical Research, 113(D4), 2007JD008975. https://doi.org/10.1029/2007JD008975 Fowler, D., Pilegaard, K., Sutton, M. A., Ambus, P., Raivonen, M., Duyzer, J., et al. (2009). Atmospheric composition change: Ecosystems– Atmosphere interactions. Atmospheric Environment, 43(33), 5193–5267. https://doi.org/10.1016/j.atmosenv.2009.07.068 Fu, D., Kulawik, S. S., Miyazaki, K., Bowman, K.W., Worden, J. R., Eldering, A., et al. (2018). Retrievals of tropospheric ozone profiles from the synergism of AIRS and OMI: Methodology and validation. Atmospheric Measurement Techniques, 11(10), 5587–5605. https://doi.org/10. 5194/amt‐11‐5587‐2018 Garane, K., Koukouli, M.‐E., Verhoelst, T., Lerot, C., Heue, K.‐P., Fioletov, V., et al. (2019). TROPOMI/S5P total ozone column data: Global ground‐based validation and consistency with other satellite missions. Atmospheric Measurement Techniques, 12(10), 5263–5287. https://doi. org/10.5194/amt‐12‐5263‐2019 Griffiths, P. T., Keeble, J., Shin, Y. M., Abraham, N. L., Archibald, A. T., & Pyle, J. A. (2020). On the changing role of the stratosphere on the tropospheric ozone budget: 1979–2010. Geophysical Research Letters, 47(10), e2019GL086901. https://doi.org/10.1029/2019GL086901 Han, Y., Revercomb, H., Cro