The Hefei station (31.91° N, 117.17° E) is located in the northwestern rural area of Hefei city, China. The surrounding region is part of a typical Jianghuai agricultural region, where wetlands and croplands are interspersed in a distinctive landscape pattern. Since July 2015, a high-resolution ground-based Fourier transform infrared spectrometry (FTIR) has been operating at the station to record mid-infrared (MIR) solar absorption spectra (700–4000 cm-1) for the remote sensing of greenhouse gases and trace species (Wang et al., 2017). A previous study has demonstrated that FTIR measurements at Hefei effectively capture the spatiotemporal variability of NH3 columns, and the retrievals are broadly consistent with IASI satellite data (Wang et al., 2022). In this study, FTIR-derived NH3 column data from 2017 to 2024 were averaged into hourly means and used for the validation and cross-comparison of satellite-based NH3 retrievals. It should be noted that the data primarily covers the period from 08:00 to 17:00 LST, with limited available observations between 17:00 and 18:00 LST.

The MIXv2 Asian emission inventory was developed under the framework of the Model Inter-Comparison Study for Asia Phase IV (MICS-Asia IV) project by assembling a mosaic of up-to-date regional and national emission inventories (Li et al., 2024). It provides monthly estimates of NH3 emissions for 2010–2017 at a spatial resolution of 0.1°×0.1° grid across seven sectors. The inventory incorporated nine regional and two global emission inventories, covering 23 countries and regions in East, Southeast and South Asia. Anthropogenic NH3 emissions were based on the best available emission inventories, including a process-based NH3 emission inventory developed by Peking University (PKU-NH3) for China, the official Japan emission inventory (JPN), and the Clean Air Policy Support System (CAPSS) emissions for the Republic of Korea, and were gap-filled with REAS version 3 for Asia (REASv3). Studies have suggested that the spatial distribution of NH3 emissions in the inventory is generally reliable, with differences compared to other emission inventories and satellite-constrained NH3 emissions typically within 20 %–50 % (Luo et al., 2022; Zhang et al., 2019; Kang et al., 2016).

The Goddard Earth Observing System composition forecast (GEOS-CF) is a high-resolution global constituent prediction model developed by NASA's GMAO. It expands on the GEOS weather and aerosol modeling system by introducing the GEOS-Chem chemistry module, providing near-real-time, three-dimensional gridded information on atmospheric composition (Keller et al., 2021). The emission inputs of anthropogenic NH3 emissions used by GEOS-CF are mainly from the Hemispheric Transport of Air Pollution emissions inventory (HTAP v2.2), which provides monthly data at a 0.1° spatial resolution. The model runs at hourly intervals with 0.25°×0.25° horizontal resolution and 72 hybrid-eta levels from the surface to 0.01 hPa. In this study, NH3 vertical profiles were obtained from 1 d replay simulations (referred to as “hindcast”) constrained by pre-computed meteorological analysis fields, and NH3 total columns were calculated by combining the model's vertical pressure levels and surface pressures from the fifth-generation ECMWF reanalysis (ERA5) (Hersbach et al., 2023).

Figure 2 illustrates the three-year averaged seasonal NH3 column concentrations retrieved from FY-4B/GIIRS observations over East Asia from July 2022 to June 2025, presented on a 0.5°×0.5° grid at two-hour intervals. The observations are aggregated according to local solar time (LST) to facilitate the analysis of NH3 diurnal variations and consistent comparisons. The study region spans multiple time zones and data availability during the 07:00–09:00 and 17:00–19:00 LST periods is substantially reduced due to retrieval filtering, resulting in noticeable boundaries across longitudes. Elevated NH3 columns, significantly above background levels, are observed over the Indo-Gangetic Plain, the North China Plain, the Sichuan Basin, and the Northeast China Plain (as shown by the black boxes in Fig. 1). The daytime NH3 columns reach values exceeding 4×1016 moleccm-2, primarily driven by intensive agricultural activities and supplemented by non-negligible contributions from urban and industrial emissions. Localized weaker enhancements of NH3 column concentrations are evident over agricultural core areas in China (including the Ningxia Irrigation Plain, the Wei River Plain, the Jianghan Plain, and oasis agriculture in the arid regions of Xinjiang), as well as the central Deccan Plateau in India, the Mekong Delta in Vietnam, the Chao Phraya River Plain in Thailand, and the Fergana Valley in Uzbekistan (as shown by the dashed boxes in Fig. S4). Most of these regions exhibit higher NH3 columns in summer and considerable daytime variability, with large differences between morning and afternoon observations. A detailed analysis of diurnal patterns over major source regions is provided in Sect. 3.3.

Comparisons with the spatial distribution of NH3 columns observed by polar-orbiting infrared sounders of IASI and CrIS show reasonable agreement, especially in regions with high NH3 columns above 3×1016 moleccm-2 (Figs. 3 and 4). The seasonal increases in atmospheric NH3 are most pronounced in spring and summer. Despite generally similar seasonal and spatial patterns, inter-satellite NH3 observations exhibit noticeable differences. Relative to the IASI observations, GIIRS typically shows very small systematic differences with fitted slopes close to unity. In contrast, GIIRS observations tend to show higher NH3 columns than CrIS in most high-concentration regions, with mean differences of approximately 25 %–50 % relative to the corresponding GIIRS values in summer. The data point scattering in the comparison plots primarily results from spectral noise. Overall, there is good agreement between GIIRS-derived NH3 columns and those from IASI and CrIS across different seasons.

To account for the effects of differing vertical sensitivities, as reflected by the AVKs, and a priori profiles among these satellites (see Fig. S1 for example), we conducted experiments using model profile data and satellite AVKs over the North-Northeast China Plain (30–43° N, 110–125° E) in June 2024 (Sect. S1 in the Supplement). Model simulations of NH3 profiles were convolved with different satellite AVKs to generate AVK-smoothed NH3 columns, representing the NH3 columns that a certain satellite would retrieve if the model simulations were the “truth”. We then compared AVK-smoothed NH3 column datasets generated using different satellite AVKs to evaluate the effects of different satellite vertical sensitivities. The comparison results (Fig. S5 in the Supplement) show high correlations (R>0.9) and slopes close to unity. This indicates consistency in the detection of NH3 across different sensors, and that the effects of different AVKs and a priori profiles are small in terms of total column retrieval when comparing different satellite retrievals.

Figure 5 illustrates the months corresponding to the maximum NH3 concentrations derived from GIIRS and IASI on a 0.5° grid, while the result from CrIS, which is highly scattered at this scale, is shown on a 1° grid in Fig. S6. The spatial pattern shows a clear north-south gradient in the timing of NH3 peaks. In the northern regions of East Asia, NH3 columns generally peak in June and July, which could be attributed to enhanced volatilization from agricultural soils, livestock operations, and other sources (e.g., composting facilities and wastewater treatment plants) under high-temperature conditions (Li et al., 2024). Irrigated croplands in the North and Northeast China Plains also exhibit a secondary spring maximum (Fig. S7) associated with extensive fertilizer use and animal feeding operations. In contrast, the regions of southern China, India, and Southeast Asia, characterized by a tropical monsoon climate, exhibit earlier NH3 peaks, generally occurring from March to May. The peak period coincides with their first cropping season, which involves intensive nitrogen fertilizer application. The warm and humid conditions in these regions further promote NH3 volatilization from soils and livestock waste. Following the onset of the rainy season in June–July, atmospheric NH3 concentrations decline sharply (such as R7, R8, R13, R14 in Fig. 6), primarily due to enhanced wet scavenging processes. This seasonal pattern aligns with findings from multiple studies reporting a distinct “late dry season-early wet season” NH3 maximum in tropical regions (e.g., Chen et al., 2020; Liu et al., 2019). Spring is a period of intense fire activity in the Indochina Peninsula, which is largely driven by residue burning, land-clearing, and naturally flammable vegetation during the dry season (Chang et al., 2021; Vadrevu et al., 2019). The NH3 maxima mainly occur in March–April, and the consistent spatial distribution of these NH3 peaks observed by GIIRS and IASI over Southeast Asia suggests the influence of biomass burning associated with fires on its seasonal variations. In addition, bimodal local maxima, with peaks of comparable magnitude, are also observed in regions such as central India (from May to September), Guangdong in southern China (typically in April and August), and the transboundary region between the Mekong Delta in Vietnam and Yunnan in China (March–April and June–July).

In general, the spatial and temporal distributions of NH3 peaks largely agree with the previous studies of Van Damme et al. (2015) and Shephard et al. (2020). Minor discrepancies between GIIRS and IASI result from differences in observational accuracy and retrieval algorithms, as discussed above. Satellite-observed NH3 seasonal cycles are driven by the interplay of agricultural activities and climatic conditions, leading to deviations from seasonal patterns of anthropogenic emissions (Fig. S8). Several studies have reported that NH3 emissions inferred from satellite observations exhibit distinct seasonal variations that can differ from those of the NH3 columns (Li et al., 2026; Kumar et al., 2025). Therefore, we focus on the spatial distribution of NH3 enhancements, relative to regional background levels, as captured by different satellites.

After removing monthly backgrounds derived from predefined regions (Fig. S9), NH3 enhancements (Fig. 7) reveal a spatial pattern of anthropogenic emissions that is highly consistent with the MIX emission inventory (Fig. 1). Apart from the major agricultural emission regions mentioned above, some regions (e.g. Hyderabad and Vijayawada in India) also show elevated NH3 levels compared to the background during autumn and winter, likely influenced by dense urban and industrial activities. Due to data discrepancies in local overpass times and sensor sensitivities among the polar-orbiting and geostationary satellites (Shephard et al., 2025), the enhancement values detected by CrIS are 0.50 ± 0.64 and 0.74 ± 0.90×1016 moleccm-2 lower than those detected by GIIRS and IASI, respectively. These inter-satellite differences can lead to inconsistencies in the observed NH3 enhancements across various source regions, particularly over the Sichuan Basin, the Ningxia Irrigation Plain, and the urban and biomass-burning regions of Southeast Asia. Compared with IASI and CrIS observations at different local times, GIIRS provides consistent and reliable measurements throughout the daytime, enabling accurate characterization of spatial and temporal variations in NH3 concentrations across distinct emission source regions.

To evaluate the data accuracy of GIIRS-derived NH3 columns and assess the consistency of temporal variations with polar-orbiting satellite measurements, we compared the satellite-based observations with ground-based FTIR measurements at the Hefei station. Satellite data within a 0.25° latitude/longitude radius around the Hefei station (Fig. S10) were selected and averaged on an hourly scale. A linear regression through the origin was applied to evaluate the proportional agreement between the datasets.

Figure 8 shows scatter plots comparing NH3 columns retrieved from satellite observations and FTIR measurements during their overlapping periods. Overall, the satellite-derived NH3 columns show good agreement with the FTIR data, with Pearson correlation coefficients (R) exceeding 0.70, although satellite values are generally slightly higher. Among the satellites, GIIRS agrees best with FTIR in terms of the highest correlation and the low root mean square error (RMSE). However, GIIRS exhibits a relative systematic bias toward overestimating NH3 columns, as indicated by its regression slope (1.12) being notably larger than those of IASI and CrIS. When measurement uncertainties were taken into account, regression analysis using the orthogonal distance regression (ODR) method yielded an increased slope of 1.24 (Fig. S11), indicating that GIIRS tends to produce systematically higher NH3 columns than IASI and CrIS.

Although the overestimation of GIIRS becomes more pronounced in afternoon observations (Fig. S12), the corresponding data conversely show improved consistency, which is characterized by higher R2 and lower RMSE values. This is related to the fact that the inversion accuracy of GIIRS observations is affected by TC and boundary layer conditions. High NH3 columns and significant TC represent favorable conditions for retrieving NH3 from satellite observations (Zeng et al., 2023a; Clarisse et al., 2010). As illustrated in Fig. S13, large positive TC between 11:00 and 15:00 LST enhances measurement sensitivity, leading to lower retrieval errors. GIIRS NH3 column retrievals are most reliable when TC exceeds 5 K, a condition predominantly observed during summer daytime hours (07:00–19:00 LST).

The Hefei station and its surrounding regions predominantly practice a rice-wheat rotation, with rice and winter wheat typically sown in spring (around May) and autumn (around October), respectively. According to the MIX emission inventory, agricultural ammonia emissions in this region exhibit two primary peaks in June and August, which are associated with local agricultural practices including livestock waste management, nitrogen fertilizer application, and irrigation activities during the rice growing season (Hou and Yu, 2020). Smaller peaks occur in October–November and April, corresponding to the rice harvest/wheat sowing period and the wheat regreening fertilization stage, respectively. Driven jointly by agricultural emission intensity and meteorological conditions such as high temperatures, the atmospheric NH3 concentrations observed by satellites and ground-based FTIR show consistent seasonal cycles, peaking in summer (June–August) and reaching a minimum in winter (November–January). Figure 9 shows the monthly NH3 time series observed by satellites and FTIR at different local overpass times. GIIRS and IASI show closely matched NH3 peaks in June. The systematic overestimation of peak values by GIIRS relative to CrIS and FTIR is consistent with an inherent data bias as shown in Fig. 8.

Notably, GIIRS-derived NH3 columns in June 2024 were significantly higher than those in the same month of 2023 and 2025, with increases of 1.46 and 1.67 times during morning overpass time (∼ 09:30 LST), and 1.35 and 2.15 times during afternoon overpass time (∼ 13:30 LST), respectively. Such pronounced variability is likely linked to extreme weather events reported in the region, such as persistent heatwaves and abrupt drought-flood transitions (Zhou et al., 2025; Ding et al., 2024). Morning observations from IASI align with GIIRS data, showing that NH3 concentrations in June 2024 reached 1.43 to 2.62 times the levels recorded in the years 2017–2025 (specifically, 1.59 and 1.53 times those of 2023 and 2025, respectively). Similar interannual changes were observed by CrIS and FTIR in June 2024, with NH3 levels increasing by factors of 1.17–1.53 and 1.40–2.28 compared with previous years, respectively, despite a data gap in June 2023 for FTIR.

When comparing GIIRS-based NH3 columns at different local times, afternoon concentrations are generally higher than those in the morning, with the largest differences observed in summer (July–August) and the smallest in autumn and winter. To more accurately characterize daytime NH3 variations, we further applied strict data filtering criteria by retaining only retrievals with TC>5 K to ensure high sensitivity to near-surface NH3. Over the three-year period, 37 high-quality observation days were identified, each providing six consecutive two-hour interval observations from 07:00 to 19:00 LST (Fig. S14). The maximum diurnal amplitude occurs in summer, reaching over 1.5×1016 moleccm-2. Except for autumn, all three other seasons exhibit relatively high NH3 concentrations around 07:00 and 17:00 LST, a pattern that aligns well with the diurnal variations reported by Wang et al. (2022) based on FTIR measurements (Fig. 10). However, NH3 retrievals from FTIR lack sufficient valid observations within a single day, making it difficult to consistently capture the intra-day variability in NH3 across different dates.

Moreover, the FTIR station is situated in a transitional zone between urban and suburban areas. A spatial domain with a radius of 0.25° used in this study encompasses GIIRS sampling points from both urban and agricultural regions (Fig. S10). To assess spatial variability of NH3, GIIRS retrievals with TC>5 K within this spatial domain were classified into urban and non-urban categories, and their daytime variations were statistically analyzed across seasons over three years. As shown in Fig. S15, NH3 concentrations in non-urban areas are consistently and significantly higher than those in urban areas during spring and summer, indicating the strong contribution of agricultural activities to ambient NH3 levels. Particularly in spring, statistically significant differences are most pronounced during 07:00–09:00 and 17:00–19:00 LST, corresponding to time windows of fertilizer application and livestock activity. During winter, higher NH3 levels in the morning and late afternoon are mainly observed in urban areas, likely reflecting the impact of traffic emissions. These findings highlight the importance of high-frequency satellite observations for dynamic monitoring of NH3 variations, and the contribution of agricultural activity in shaping the diurnal NH3 variability.

We focused on eight major agricultural source areas located in the North China Plain and the Northeast China Plain, the Sichuan Basin, and the Indo-Gangetic Plain, to further investigate daytime variations in NH3 columns. These areas are characterized by high NH3 column concentrations and emissions, generally flat terrain, and a large proportion of irrigated cropland (as shown in Figs. S16 and S17). To ensure sufficient measurement sensitivity to near-surface NH3, only GIIRS observations with TC greater than 5 K were retained, thereby excluding cases with low information content. During autumn and winter, available observations are sparse in the early morning and late afternoon. Here, GIIRS-observed daytime variations of NH3 columns across different seasons (Fig. 11) and the corresponding spatial distributions for each study region (Figs. 12–14 and S18–S20) are presented, with comparisons to anthropogenic NH3 emissions (Fig. S17) and GEOS-CF model simulations (Figs. S21–S24). Despite significant differences between model-simulated and satellite-based NH3 columns, we focused on analyzing the daytime variations and their spatial distributions in representative months of May–June (NH3 peak months) in 2025.

The North China Plain is characterized by extensive irrigated croplands with winter wheat-maize rotation and widespread small-scale livestock farming concentrated in east-central provinces, including Henan, Shandong, and Hebei. Due to intensive agricultural and livestock activities, dense industrial emissions and heavy urban traffic, the region ranks among the most polluted regions in China. The spatial distribution of local NH3 high values across different seasons shows considerable variations, especially during the late afternoon in the spring and summer, which may indicate a change in emission hotspots (Figs. 12 and S18). Across the three selected areas (A1–A3), GIIRS observations reveal a consistent temporal pattern of NH3 columns, with values increasing from early morning to late afternoon in summer but generally declining after 15:00 LST in other seasons. In summer, the mean variations in GIIRS NH3 column amplitudes for the afternoon and morning are 22 %, 27 %, and 29 %, respectively, while in May and June 2025, these variations increased to 56 %, 38 %, and 32 %, respectively.

The Northeast China Plain is a key grain production region, particularly for maize, soybeans, and rice. Agriculture is highly mechanized, and livestock farming is carried out on a large scale. Because agricultural emission sources are spatially concentrated and geographically stable, the spatial pattern of NH3 enhancements shows little variation at different local times (Fig. 12). The B1 area is located in the estuarine delta, with large cities such as Shenyang and Anshan to the east. Due to low population density and limited industrial activity, NH3 emissions in this area are primarily associated with fertilizer application and animal husbandry, resulting in relatively lower levels than those in other areas. GIIRS observations show that NH3 columns are at markedly reduced levels, approximately 52 % of those in the North China Plain, but follow a similar diurnal cycle, with the variation amplitude in summer being comparable (25 % for 3 years, 43 % for May–June 2025). The difference is that in winter, the diurnal variation also shows higher NH3 columns in the morning (07:00–09:00 LST), which may be related to traffic emissions and the effects of meteorological conditions (such as temperature and humidity changes) in the early morning. Ground-based studies have reported that the peak atmospheric NH3 concentration typically occurs in the morning, as observed in the urban area of Beijing (Gu et al., 2022) and in the rural areas of Xianghe (He et al., 2020), Xinxiang (Teng et al., 2017), Shanghai (Wang et al., 2015), and Hefei (Wang et al., 2022).

The Sichuan Basin, located in southwestern China, is an important agricultural region encompassing the Chengdu-Chongqing urban agglomeration. Ammonia emissions exhibit a clear east-west gradient, reflecting regional differences in agricultural practices and livestock management (Li et al., 2021; Zhang et al., 2018; Kang et al., 2016). Studies based on IASI observations indicate that NH3 hotspots in the Sichuan Basin are primarily concentrated in the Chengdu Plain, the southern urban clusters, and northwestern Sichuan, where agricultural sources dominate ammonia emissions (Yang et al., 2024; Dammers et al., 2019; Van Damme et al., 2018). GIIRS observations show that NH3 columns are significantly enhanced in the western regions, particularly along the Mianyang-Deyang-Chengdu-Meishan-Leshan corridor (Figs. 7 and 13), which is associated with intensive agricultural activity and urban-industrial emissions. The basin's unique topography and meteorological conditions such as frequent temperature inversions result in very few valid NH3 observations after 17:00 LST. Across different seasons, NH3 columns generally exhibit an increasing trend from morning to afternoon. In spring and summer, seasonal mean NH3 concentrations show relatively small differences because agricultural sources dominate and vertical mixing is enhanced by changes in boundary layer height, with daytime NH3 variations of 27 % and 23 %, respectively. The largest daytime variation occurs in winter, when afternoon values are 1.46 times higher than those in the morning.

The Indo-Gangetic Plain, South Asia's most extensive alluvial plain, spans Pakistan, northern India, and Bangladesh. The region is densely populated and intensively farmed, producing rice, wheat, sugarcane, and vegetables, often under irrigated and mechanized conditions. Fertilizer application, livestock excretion, industrial activities, and high population density lead to significant NH3 and reactive nitrogen emissions, reaching their maximum over the Punjab region in Pakistan. We selected three areas that are located in Punjab, Uttar Pradesh, and Bangladesh, respectively (Fig. 14). Differences in NH3 emission sources and climatic conditions across these regions lead to pronounced seasonal variations in NH3 concentrations. The daytime variation amplitudes of NH3 columns are highest in spring, which are 17 %, 32 %, and 21 % for D1, D2, and D3, respectively. However, under the coupled conditions of abundant precipitation and strong solar radiation in summer, NH3 columns in the D2 and D3 areas reach their peak around 13:00 LST.

The diurnal variation of atmospheric NH3 column concentrations is primarily influenced by surface emissions, boundary layer dynamics, and gas-particle partitioning associated with meteorological conditions (Lan et al., 2024; Behera et al., 2013). In these main agricultural source areas, NH3 concentrations generally rise from morning to afternoon, driven by diurnal variations in temperature and humidity that shift the gas-particle equilibrium toward the gaseous phase. The daytime NH3 variations simulated by the GEOS-CF model for May–June 2025 range from 7 % to 36 %, slightly lower than the variations observed by GIIRS (10 % to 56 %). However, the GEOS-CF model was unable to accurately reproduce the spatial pattern of NH3 enhancements in the Sichuan Basin, which can be largely attributed to the bottom-up emission inventory and the influence of complex topography and local climatic conditions. After considering the vertical sensitivities of satellite retrievals, the AVK-smoothed model data resulted in an overall reduction of 12 % ± 38 % in the model NH3 columns over the Sichuan Basin. However, considerable discrepancies in the spatial distribution remain relative to the satellite observations. Figure 15 presents a cross-comparison between the satellite-derived NH3 columns and the AVK-smoothed model data in May–June 2025. Over the North-Northeast China Plain, the model underestimates NH3 columns by 62 % on average, with a standard deviation of 45 %, whereas over the Sichuan Basin, the underestimation reaches 65 %, with a standard deviation of 30 %. In contrast, the model generally overestimates NH3 columns over the Indo-Gangetic Plain. Regarding the spatial and temporal patterns of diurnal variations (Figs. S25–S28), the largest discrepancies are observed in the areas of C1 in the Sichuan Basin and D1 in the Indo-Gangetic Plain, suggesting that the model has limited capability to simulate NH3 in these areas and may require improvements through satellite-observed constraints to improve simulation accuracy.