This study utilizes NH3 total columns retrieved from solar absorption spectra measured with ground-based FTIR spectrometers at two sites in and around the MCMA. The urban FTIR station is located at the south of Mexico City within the campus of the Universidad Nacional Autónoma de Mexico on the rooftop of the Instituto de Ciencias de la Atmósfera y Cambio Climático (UNAM, 19.33∘ N, 99.18∘ W, 2280 m a.s.l.). A custom-built solar tracker directs solar radiation to the entrance of a FTIR spectrometer (Bruker Optik GmbH model Vertex 80) that has a maximum unapodized resolution of 0.06 cm-1. The instrument is equipped with a KBr beamsplitter and two detectors (HgCdTe and InGaAs). The HgCdTe detector is cooled with liquid nitrogen. For more details about this system, see Bezanilla et al. (2014). Measurements from the remote FTIR site were made at the Altzomoni Atmospheric Observatory (ALTZ, 19.12∘ N, 98.66∘ W, 3985 m.a.s.l.), a high-altitude station located 60 km from the UNAM urban site and within the Izta-Popo National Park surrounded by nature. This site is part of the Network for the Detection of Atmospheric Composition Change (NDACC) (De Mazière et al., 2018) contributing data from a high-resolution FTIR spectrometer since 2012. This instrument (Bruker Optik GmbH model IFS 120/125 HR) can record solar spectra with a maximum spectral resolution of 0.0035 cm-1, and is equipped with KBr and CaF2 beamsplitters and HgCdTe, InSb, and InGaAs detectors. For details of the site and the instrument, see Baylon et al. (2017). The first study presenting and validating the combined usage of trace gas products obtained from dedicated retrievals with FTIR spectra measured at UNAM and ALTZ was that of Plaza-Medina et al. (2017).

The FTIR spectra used for the retrieval of NH3 were collected with the HgCdTe detector at both sites with a spectral resolution of 0.005 cm-1 at Altzomoni, and 0.1 cm-1 (prior to 2014) and 0.075 cm-1 at UNAM, using an optical band pass filter to enhance the region between 700 and 1400 cm-1. In this study, we extend the UNAM time series used in a previous 2012–2015 comparison between IASI and FTIR (Dammers et al., 2016) with improvements to the retrieval, and we present for the first time Altzomoni FTIR-NH3 retrievals for the period between April 2012 and May 2020. The analysis presented here focuses on the region around the MCMA, also using the IASI satellite product and a back-trajectory evaluation, as described below.

For both sites, the solar FTIR spectra were analysed using PROFFIT version 9.6 for the retrievals (Hase et al., 2004) to obtain the NH3 total columns. A retrieval strategy based on that reported by Dammers et al. (2015) was used, comprising two microwindows (929.1–931.8 and 963.7–970.0 cm-1) to cover the NH3 absorption lines from the υ2 vibrational band in the mid-IR region. The trace gases H2O, CO2, O3, CO, and N2O were taken into account as interfering species for the retrieval for both stations, and the temperature and pressure profiles were obtained from the US National Centers for Environmental Prediction (NCEP). Spectroscopic parameters were obtained from the high-resolution transmission molecular absorption database HITRAN 2008 Rothman et al., 2009), and the a priori profile information about the interfering gases was obtained from 40-year averages of simulations from the Whole Atmosphere Community Climate Model (WACCM) (Eyring et al., 2007; Marsh et al., 2013). Since NH3 profiles from WACCM are only representative of remote regions, scaled a priori profiles derived from 5 years of averaged NH3 simulations from the global chemical transport model GEOS-Chem v11 were used instead. Constructed scaled a priori profiles have been used previously from GEOS-Chem simulations for the retrieval of trace gases (Shephard et al., 2011; Shephard and Cady-Pereira, 2015; Bader et al., 2017). A total of 7992 NH3 columns were retrieved for the UNAM station and 4031 for ALTZ. The resulting uncertainties obtained with PROFFIT averaged over the entire time series in molecules cm-2 at UNAM were 1.25×1015 (random), 8.40×1014 (systematic), and 1.52×1015 or 11.50 % (total); and 3.09×1014 (random), 3.14×1014 (systematic), and 4.42×1014 or 51.37 % (total) at ALTZ. The average degrees of freedom for signal (DOFS) averaged over the entire time series were 2.03 for UNAM and 1.04 for ALTZ.

IASI measures the infrared thermal radiation emitted by the Earth's surface and the atmosphere from a Sun-synchronous orbit on board the MetOp platform. Spectra are recorded in the 645–2760 cm-1 spectral range at a spectral resolution of 0.5 cm-1 (Clerbaux et al., 2009). The instrument crosses the Equator at mean local solar times of 09:30 LT and 21:30 LT providing global coverage of the Earth twice a day. IASI has a field-of-view composed of four circular footprints each with a diameter of 12 km at the nadir view and up to 20km×39km elliptical pixels outside the nadir depending on the satellite viewing angle, complemented by scanning along a swath width of 2200 km off-nadir perpendicular to the ground track (Clarisse et al., 2009; Van Damme et al., 2014, 2015). The IASI-NH3 retrieval products are based on artificial neural networks (ANNs) that link the hyperspectral range index (HRI), a calculated dimensionless index that represents the amount of NH3 in the column, to other input parameters such as temperature, pressure, water vapour profiles, and parameterized vertical profiles of NH3 to derive the NH3 total column. The algorithm maps the HRI to the NH3 total column using a trained neural network; the uncertainty of each NH3 column can be estimated by the propagation of the input parameters' uncertainties. However, large HRI values (more than 3σ) are associated to a confident detection of NH3 (Van Damme et al., 2014, 2017 Whitburn et al., 2016). The current spectral range for the retrieval process is set to 812–1126 cm-1 to increase the sensitivity of NH3 and reduce interferences (Van Damme et al., 2021). The retrieval scheme does not produce averaging kernels, however previous studies comparing the IASI-NH3 product with ground-based FTIR measurements have demonstrated good agreement (e.g. Dammers et al., 2016, 2017; Lutsch et al., 2019; Tournadre et al., 2020; Yamanouchi et al., 2021). In addition, under conditions of high NH3 and when the thermal contrast is large, IASI has maximum sensitivity to NH3 in the boundary layer (Clarisse et al., 2010). An error estimate is provided with each individual IASI observation; for this work IASI observations with errors less than 100 % were used. IASI's average detection limit for NH3 under large thermal contrast is about 3 ppbv, and can be as low as 1 ppbv under conditions of well-mixed NH3 throughout a thick boundary layer (Clarisse et al., 2010).

For this study, 11 years of the IASI-A NH3 total columns (ANNs for IASI (ANNI)-NH3-v3) between 2008–2018 were used; details of this version 3 can be found in Appendix A of Van Damme et al. (2021), and was also used by Yamanouchi et al. (2021) and Viatte et al. (2022). The spatial distribution of NH3 over Mexico City was obtained by averaging all the IASI-A morning observations between January 2008 and December 2018 over this region. The FTIR-NH3 total columns at UNAM were compared against the IASI-NH3 total columns over Mexico City to assess the agreement between both data sets. Due to the high spatiotemporal variability of NH3, the temporal and spatial coincidence criteria were tested and assessed using the correlations (both R and slope). In addition, as suggested in Dammers et al. (2016), an elevation filter (FTIR station altitude minus IASI observation <300 m) was applied. The criteria resulting in the best correlations were elevation filter <300 m, spatial sampling difference <20 km, maximum temporal sampling difference ±40 min, and maximum IASI-NH3 retrieval error of 100 %. The seasonal variability comparison and annual averages were performed using only FTIR-NH3 retrievals between 09:00 LT and 10:59 LT, corresponding to the IASI overpass time over Mexico City, and the <20 km spatial criterion for IASI-NH3 total columns. Altzomoni correlation plots with IASI-NH3 data were not included due to the few coincidences between the FTIR and IASI.

To determine the primary sources of NH3 measured at the UNAM station and to assess the dominant atmospheric transport pathways during the events with the largest hourly means of the NH3 columns in the time series, trajectory cluster analysis (Reizer and Orza, 2018) was applied. Back-trajectories of 8 h were selected to capture the air masses passing over the MCMA. Using the UNAM station as the receptor, back-trajectories were calculated using the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model (Stein et al., 2015; Draxler et al., 1997) at different altitudes above the UNAM station level (2280 m a.s.l.). The HYSPLIT model can be run online at the following link https://www.ready.noaa.gov/HYSPLIT.php (last access: 21 October 2022). The wind data used for the back-trajectories were derived from the NCEP North American Mesoscale (NAM) analysis product at 12 km and 1 h of spatial and temporal resolution, respectively (NCEPT NAM, 2015). The cluster analysis is an embedded routine in HYSPLIT and is based on the Ward's agglomerative hierarchical clustering algorithm (Ward, 1963). Finally, the total spatial variance (TSV) method (Draxler et al., 2021), included in HYSPLIT, was used to fit the number of clusters that represent the data.

The NH3 total column time series retrieved at both FTIR stations are shown in Fig. 2. The urban UNAM columns, shown in the top panel, are about one order of magnitude larger than the high-altitude Altzomoni columns. The average NH3 total columns for the entire period (1.46±0.64×1016 at UNAM and 1.87±2.40×1015 molecules cm-2 at Altzomoni) are listed and compared to values reported for stations in other parts of the world in Table 1. The Mexico City NH3 total columns are comparable with those reported in Bremen, while they are about twice as large as those measured at Toronto (Canada), Paris (France), and Lauder (New Zealand). Jungfraujoch, a remote high-altitude station in Switzerland with similar characteristics to Altzomoni, presents a significantly lower average NH3 column and also has much lower variability. The reason for this might be that Altzomoni is impacted more frequently by biomass burning events in the dry season and also by the regional boundary layer, receiving polluted air from Mexico City and other large urban centres in the afternoon (Baumgardner et al., 2009).

Figure 2 also shows the fit of a Fourier series (Baylon et al., 2017) to reproduce seasonality in both stations. A clear cycle is seen at both stations, with a maximum between mid- and late April, and a minimum in late December. However, the difference between the minimum and the maximum is larger at Altzomoni, with UNAM having greater NH3 background concentrations. The average annual increase in the NH3 columns obtained from the Fourier fit are 92±3.9×1013 at UNAM (from 2012 to 2019) and 8.4±1.4×1013 molecules cm-2 yr-1 at Altzomoni (from 2012 to 2020). Van Damme et al. (2021) reported a trend from 2008 to 2018 of 2.5±1.5×1013 molecules cm-2 yr-1 for all of Mexico, which is closer to the Altzomoni value. The difference in magnitude can be attributed to the datasets and methodology as the present study uses ground-based FTIR measurements from two sites with higher values in 2019 and 2020, while Van Damme et al. (2021) used IASI satellite data over a wider region between 2008 to 2018.

The average diurnal variability of the FTIR NH3 total columns at both stations is displayed in Fig. 3. The largest average NH3 columns at the urban station are on the order of 1.50×1016 molecules cm-2 and were observed during the morning and the evening. Although the diurnal pattern is not as evident as in other cities where motor vehicles have been found to be a dominant source of urban NH3 (e.g., Osada et al., 2019; Kotnala et al., 2020), traffic emissions in Mexico City still might play a role in conjunction with other urban sources of atmospheric NH3. The average NH3 columns at UNAM have a minimum of 1.35×1016 molecules cm-2 at 13 h, which can be attributed to the conversion to ammonium, as was observed by Moya et al. (2004) when describing the evolution of the surface gas phase NH3 and PM NH4+ at an urban site in Mexico City.

The diurnal cycle at the remote station (Fig. 3a) is noticeably different, with NH3 columns that increase systematically as the day progresses, with the largest variability in the afternoon hours. Altzomoni is located within a natural protected area with few local sources of NH3, and even less so during the morning hours when values are around 0.10×1016 molecules cm-2, when cooler temperatures do not favour the volatilization of NH3. The columns increase throughout the day, having the largest average values of 0.76×1016 molecules cm-2 in the evening, probably transported from lower altitudes by the dynamics of the regional boundary layer (Baumgardner et al., 2009). This is supported by the large variability observed in the afternoon, since the probability that NH3 is transported >1700 m a.g.l. (above ground level) up to this site strongly depends on the meteorological conditions, which vary from day to day. Comparisons between daily average NH3 columns and the daily averages of some meteorological variables from the RUOA Network (Red Universitaria de Observatorios Atmosféricos), such as temperature, relative humidity (RH), precipitation, and solar radiation, resulted in weak correlations. These correlations were positive with temperature and solar radiation (r between 0.1 and 0.3), and negative with RH and precipitation (r between -0.1 and -0.3). The average wind speed for Altzomoni is 4.5 m s-1, with dominant winds from the east-southeast and west-northwest, while the average wind speed for UNAM is 1.6 m s-1, with dominant winds from the north and the north-northwest.

The seasonal variability of the FTIR NH3 total columns is shown in Fig. 4. In general, the pattern is similar at both stations, showing the NH3 temperature dependence with larger NH3 columns in the months of March, April, and May which correspond to the warm–dry season; this season usually has days with clear skies, weak winds, high pressure systems, and biomass burning events (Molina et al., 2020), and also corresponds to the most critical part of the fire season in Mexico City (CENAPRED, 2021; Yokelson et al., 2007). In addition, there are two agricultural seasons in the country, the first one from April to September and the second one from October to March. The fertilizer application combined with meteorological conditions could favour NH3 volatilization from the agricultural sources contributing to the higher NH3 spring columns observed in Fig. 4. Smaller columns are clearly observed during the wet season (June to October), due to the increase in wet deposition, and during the cold–dry season from November to February, due to less favourable conditions for NH3 volatilization. This is in agreement with Viatte et al. (2022). The annual cycle of NH3 columns at the urban UNAM station is similar to that observed at the background Altzomoni station but has a larger amplitude. A study by Sun et al. (2017) observed that the growing efficiency of three-way catalysts in motor vehicles is responsible for large NH3 emissions detected in urban locations in the USA and China. These emissions are strongly dependent on traffic volume and thus should not have a strong seasonality. On the other hand, emissions originating from agricultural activity usually have a distinct seasonality that depends on the fertilizer application and temperature (Sun et al., 2017; Van Damme et al., 2015, Viatte et al., 2022). In the current study, the annual cycles follow the pattern of temperature (Fig. 4c), indicating that emissions from other sources, such as fires, waste treatment, human or pets emissions, may be contributing significantly to the NH3 detected over the MCMA.

There are more features to note in Fig. 4. While Altzomoni and UNAM NH3 columns have similar annual cycles, those at Altzomoni have greater variability throughout the different years during the warm–dry season than during the rest of the year. This might be due to the strong relationship between pollutants reaching the high-altitude station and the boundary layer dynamics and wind conditions, which are more variable during the warm–dry season. However, another contribution to this variability may be biomass burning activity, which has a maximum during the warm–dry months as has been shown by Cady-Pereira et al. (2017). The 2013 pollution events presented by Cady-Pereira et al. (2017) are seen on 23 April, 9 May, and 25 May in Fig. 2, with 9 May having the largest enhancements at Altzomoni; unfortunately, there are no coincident measurements for UNAM. At Altzomoni, the average NH3 column for 9 May 2013 was 2.39±0.53×1015 molecules cm-2, which is 28 % greater than the average column for the entire period (Table 1). However, in 2013, the largest NH3 column measured at Altzomoni was on the evening of 27 February with an average value of 17.4±0.58×1015 molecules cm-2, almost 10 times higher than the average column presented in Table 1. This enhancement on 27 February 2013 seems to be local and of short duration, most likely due to a nearby biomass burning event. This is supported by the detection of active fires northwest of the site on that date by the MODIS instrument on Aqua as shown in Fig. 5. With its high altitude and few local sources, Altzomoni seems to be more sensitive for the detection of pollution events than UNAM. Even if the fires are not occurring nearby, the increased lifetimes of emitted pollutants at these altitudes may favour transport over longer distances to Altzomoni.

The correlation between IASI-NH3 and the ground-based FTIR-NH3 total columns at UNAM from November 2013 to December 2018 is shown in Fig. 6. The coincidence criteria are described in Sect. 2.2, and are based on values used in previous validations of IASI products using ground-based FTIR data (Dammers et al., 2016), including an elevation correction using the Space Shuttle Radar Topography Mission Global Product at 3 arcsec resolution over Mexico City (SRTMGL3, Farr et al., 2007). A total of 64 coincident data pairs were found, from which a correlation coefficient R=0.72 and a mean relative difference (MRD) of -32.2±27.5 % were obtained (±1σ). These results are consistent with R=0.64 and MRD =-30.8±43.9 % reported by Dammers et al. (2016) for this region using an older version of the IASI-NH3 product. The correlation is also similar to that of Tournadre et al. (2020), who obtained R=0.79, when comparing IASI-NH3 to FTIR-NH3 columns using a similar instrument (Vertex 80) in Paris, and to the R=0.80 and MRD =-32.4±56.3 % reported for 547 coincidences from several ground-based FTIR stations and IASI-NH3 (Dammers et al., 2016). The ANNI-NH3-v3 product is thus in agreement with the ground-based data and even presents an improved correlation compared to the previous result. However, an underestimation in the IASI-NH3 total columns of approximately 32 % over Mexico City persists. Dammers et al. (2017), using an older version of the ANNI-NH3 product, attributed these differences to a combination of more randomly distributed error sources and large systematic errors, however these reasons need to be investigated further.

The spatial distribution of NH3 over MCMA as observed by IASI is presented in Fig. 7. The distribution shows a clear NH3 enhancement in the northeast section of Mexico City and in part of the Estado de México region. There are several potentially important sources located in this area: Mexico City International Airport, an area of continuous traffic emissions; the Bordo Poniente compost plant, which treats around 1500 t of daily organic waste from the city; wastewater discharge and treatment bodies with nearby bird colonies such as the regulation lagoon Cola de Pato; and the agricultural area at Texcoco. The combination of these factors, along with the high population density of this area, are likely to be cause of the NH3 enhancements observed in this part of the city. This enhancement is partially in agreement with the location of the larger NH3 emissions reported in the Mexico City Emissions Inventory (SEDEMA, 2018, 2021) shown in Fig. 1 in the northern part of the city, which is mainly associated with population activities and domestic animals' excreta. Figure 7d will be discussed in Sect. 3.3.

Figure 7a–c shows the variations of NH3 over the year, with the largest columns measured during the warm–dry season and to the northeast of UNAM. In contrast, the NH3 columns are reduced in the wet season when wet deposition can occur, and are smallest during the cold–dry season, when there are lower temperatures and less NH3 volatilization. To investigate the influence of local topography on the NH3 distribution, Fig. 8 compares the average IASI-NH3 total column spatial distribution (a) with altitude (b). The figure illustrates that the highest columns are located at the lowest altitudes while the lowest columns are at higher altitudes, reflecting the source locations and the boundary layer dynamics. The figure shows that the main NH3 sources in MCMA are located in the most urbanized areas in Mexico City and Estado de Mexico at an altitude of around 2250 m. These urban emissions agree with the statement of Li et al. (2020) that human NH3 emissions contribute significantly to the total NH3 emissions in hot and highly populated urban areas such as Mexico City. A rough estimation using 25 ∘C as an average diurnal temperature for Mexico City and the 0.4 mg of NH3 per hour at 25 ∘C from Li et al. (2020) resulted in an estimate of 34 t of NH3 per year, a contribution of the same order of magnitude as all the “point sources” and “urban waste” combined according to the Mexico City Emissions Inventory (SEDEMA, 2018, 2021).

Comparisons between the seasonal and temporal variability of NH3 over the UNAM station in Mexico City were performed using morning (9–11 h) FTIR-NH3 columns and IASI-NH3 columns with a spatial criterion of <20 km from the UNAM station (Fig. 9a). The NH3 seasonal variability over Mexico City in Fig. 9a is similar for both the IASI and ground-based FTIR NH3 columns and is in agreement with Fig. 4. However, IASI-NH3 shows a consistent negative bias. The evolution with time is represented by the IASI-NH3 and FTIR-NH3 annual averages in Fig. 9b. The datasets suggest an increasing trend in the annual averages of the NH3 total columns, with larger columns observed in the most recent years, even in Altzomoni, except for 2013 which was affected by the 27 February event as discussed previously. The average annual increase of the IASI-NH3 columns between 2008 and 2018 (using the <20 km spatial criterion and the same Fourier fit as for the FTIR data) is 38±7.6×1013 molecules cm-2 yr-1 for Mexico City. This is a larger positive trend than that of 2.5±1.5×1013 molecules cm2 yr-1 reported by Van Damme et al. (2021) for all of Mexico over the same period, but lies between the values obtained for the remote and urban FTIR measurements.

A cluster analysis was applied using 8 h back-trajectories 100 m above UNAM station to identify the main transport pathways for air masses arriving at this station that correspond to the highest average hourly NH3 total columns (Fig. 7d). The 100 m cluster was considered the most representative because NH3 is mostly concentrated near the surface. The TSV method was able to represent the primary trajectories at 100 m above UNAM with only three clusters. It was found that 68 % of the trajectories originate from the north (red line with black dots), 27 % from the west-southwest (green line), and 5 % from the south (blue line). However, the individual back-trajectories that comprise the red cluster (the thin black lines in Fig. 7d) indicate that most of the NH3 detected at UNAM comes from a variety of local sources and does not originate exclusively from NH3-enriched air masses transported from the enhancement region to the northeast observed in Fig. 7a–c. This is in agreement with Viatte et al. (2022). The relationship between the back-trajectories and measured NH3 columns can be explained by the fact that Mexico City is located in a basin; the wind fields are constricted in this basin and in general they are breeze winds (6 km h-1). Under these conditions, small locally distributed NH3 urban emissions seem to be the main cause of the high column values of this pollutant measured at the UNAM station, this agrees with Fig. 8 which shows that the main NH3 sources in MCMA seem to be urban.