Since 2007, L-EPA has equipped a total of 14 stations with NH₃ monitors, as detailed in Table 1. The classification of sites adheres to the European Directive, with six located in urban or suburban areas (BG, Co, CR-FbF, MI-PA, PV, and SnB), three in proximity to intensive livestock activities (CdC, Pi, and Be), three in rural agricultural areas (CR-GB, MV, and SKI), and one at an elevation of about 1,200 m in a grazing land (Mo). The site labeled MB-P is situated in the middle of one of the largest fenced parks in Europe, close to a small livestock farm within the park and police horse stables. The locations of these stations, which have been part of the NH₃ monitoring network over the years, are depicted in Figure 1.

This paper focuses on four case studies involving the following sites: MI-PA[UB] (an urban background station in Milan, the largest city in the middle of the Po Valley), CdC[farmRB] (a site near a swine farm where commonly elevated concentrations are detected), SKI[RB] (under investigation in the LIFE Prepair project about aerosol chemical composition and source apportionment, along with MI-PA[UB] and other Po Valley sites), and Mo[Rem] (noteworthy for its position often above the mixing layer height). Acronyms in square brackets, denoting the site types, are provided alongside the site names. Refer to the caption in Table 1 for a complete explanation of the acronyms. For the Mo station, “Rem” is used to emphasize its more remote location compared to SKI. Similarly, CdC[farmRB] indicates the proximity of this site to a swine livestock facility.

To detect ambient air concentrations of NH₃, the L-EPA network is mainly equipped with chemiluminescence-based monitors (57, 58). Because chemiluminescence is identified as the reference method for NO_x [UNI (59)], this technique is widely used in air quality networks.

However, over the years, L-EPA has improved its instrumentation with additional equipment such as cavity ring-down spectroscopy [CRDS (60, 61)]. Although every monitoring station listed in Table 1 is equipped with a chemiluminescence-based monitor, only MI-PA[UB] and MV[SB] allow a comparison between these two techniques (Section 2.3).

Minimizing the error of measurements and correctly determining the uncertainty associated with the measurement means reducing both the stochastic and deterministic components; this is guaranteed by the application of QA/QC procedures, provided by 2008/50/CE and by the Italian ministerial decree (DM) of 30th March 2017 (62). In particular, QC activities consist of a series of procedural actions implemented by the L-EPA technicians according to the indications of the aforementioned DM and help ensure the quality control of the measured data in compliance with the quality objectives of European legislation. QA procedures consist of a series of scheduled activities through blind testing and performance auditing.

L-EPA applies QC procedures also to NH₃ analyzers through scheduled preventive maintenance activities, other than regular calibrations and zero and span checks. NH₃ measurements are affected by several positive artifacts, and a major source of interference is the presence of high and variable water vapor in the ambient air (63). As reported in the literature (64, 65) and, in particular, at low concentrations, other factors identified for affecting the measurements are the inlet design, the filter material and aging, and the quality of calibration standards. Based on own experience and knowledge of literature, L-EPA has developed further internal procedures. For that, in addition to the scheduled maintenance activities, various precautions have been taken during the installation of the instruments to minimize the interferers and maximize the efficiency of the measurement. Some of them are (i) all pneumatic connections must be made with chromatography-grade (passivated) stainless steel tubing, glass, Teflon®; (ii) it is preferred to avoid filters on the sample line or, at least, to use a PTFE filter; (iii) use of a siliconized glass fiber sleeve to minimize the condensation of water; (iv) short pneumatic connections are recommended; (v) diluters with capillaries (no Mass Flow Controller) for the production of the NH₃ calibration sample are to be preferred; (vi) oscillations of the ambient working temperature have to stay within ±5°C; (vii) it is recommended to use three different lines for the zero sample and for the NO/GPT and NH₃ calibration samples (manual switching); (viii) it is necessary to wait at least 12 h for the stabilization of the NH₃ sample (rise time = 90% at 5 min). Finally, the calibration must be carried out in equipped laboratories; the CRDS also are checked in the laboratory with sample gas readings.

About QA procedures, several intercomparison campaigns have been conducted over the years to confirm the reliability of the measurements.

Using the reference method for analysis (66–68), radially symmetric diffusive samplers have been used compared to the online measurements (three in parallel for reproducibility). The use of diffusive or passive samplers for the study of air pollution in environments of work has been known since the 1970s (69, 70). Considering the high variability of the passive samplers used since 2007, the inherent variability of the method, and dependence on atmospheric conditions, the results shown in Supplementary Table S1 are to be considered acceptable. Within the same technology, i.e., chemiluminescence, tools from different brands have been also compared. In fact, despite being based on the same measurement principle, various manufacturers adopt small technological differences. These results are summarized in Supplementary Table S1.

In Supplementary Figure S1, a comparison is shown between the data collected from a CRDS monitor (Proceas AP2E) and a chemiluminescence monitor (TEI-17i) near an agricultural field. The time series, with a 10-min resolution in the graph, highlights the good agreement between the two instruments. The comparison campaign between the two technologies consists of approximately 7,900 h of measurement. The results of the linear regression are indicated in the text box of the figure.

Starting from 2013 for MI-PA[UB] and 2018 for SKI[RB], a time series for PM₁₀ chemical composition is available. Aerosol samples are collected daily on quartz fiber filters (Pall TissuQuartz®, Ø = 47 mm) and Teflon filters (Pall, Ø = 47 mm, PMP ring, 2 µm porosity) by means of gravimetric samplers [Skypost PM, TCR-TECORA, Cogliate (MB), Italy, or Lifetek PMS, Megasystem, Bareggio (MI)] equipped with a PM₁₀ sampling head (1 m³/h, 24 h). After collection, filters are stored in the darkness and at low temperature to prevent photochemical reactions and compound volatilizations. A punch of 1.5 cm² is removed from each filter, and the water-soluble compounds are extracted through ultra-pure water (Sartorius® Arium Mini, resistivity 18.2 MΩ) in a sonic bath (20 min). The resultant solution is then filtrated (Nylon or PTFE Syringe Filter—pore size 0.45 µm) and injected into an ion chromatographer (Metrohm 930 and 881) for the determination of anions (Cl⁻, NO₂⁻, Br⁻, NO₃⁻, PO₄³⁻, SO₄²⁻) and cations (Na⁺, NH₄⁺, K⁺, Mg²⁺, Ca²⁺). Another punch of 1.5 cm² is removed from each filter for the determination of the carbonaceous fraction by means of thermal–optical analysis (Sunset Laboratory Inc., Tigard, OR, USA), according to the NIOSH-like and EUSAAR-2 protocols. Teflon filters are used for elemental composition determination by dispersive x-Ray fluorescence (ED-XRF, Epsilon 4, Malvern Panalytical) for elements with atomic number Z higher than 11 (Mg, Al, Si, P, S, Cl, K, Ca, Ti, V, Cr, Mn, Fe, Ni, Cu, Zn, As, Br, Rb, Cd, Pb, Sr, Sn, Sb, Ba). Starting from 2013, in MI-PA[UB], hourly Black Carbon measurements with a Multi-Angle Absorption Photometer (MAAP, Thermo Scientific Model 5012, PM_2.5 cutoff, no dryer) that measures the aerosol absorption coefficient at a wavelength of 637 nm are also available (71), and from it, the BC is computed considering the deposit area and sampling air flow and using a mass-specific absorption coefficient of 6.6 m²/g. In addition, during the latter part of winter 2022 (from 22 February to 15 March), a higher time resolution sampling campaign was performed to observe the variability of secondary inorganic compounds and the role of gaseous NH₃ at various sites in Lombardy. For this reason, PM₁₀ samples were collected (Pall TissuQuartz®, Ø = 47 mm, 1 m³/h, 6 h) at six sites of the L-EPA air quality network and equipped with NH₃ monitors, i.e., MI-PA[UB], SKI[RB], CdC[farmRB], MV[RB], Be[RB], and SdB[UI].

A valuable overview of NH₃ concentrations is obtained by calculating the average daily concentrations (as the geometric mean) across the years for each day of the year. Figure 3 illustrates the annual cycle for the four selected monitoring sites as case studies, with dark lines representing the geometric mean. Generally, the arithmetic mean is more susceptible to spikes associated with local sources, whereas the geometric mean is more effective at representing background conditions (72).

Ammonia concentrations measured at the MI-PA[UB] station are depicted in Figure 3A, where values of the geometric mean comparable to the literature for measurements in urban areas (29, 63) are presented. Although the average annual cycle does not show significant variability throughout the year (the 10th–90th percentiles range is between 3 and 40 µg/m³, with a geometric mean value of 7.5 µg/m³), the measurements suggest three periods with an increase in NH₃ concentrations: late winter/beginning of spring, late springtime, and autumn. At the SKI[RB] station (Figure 3B), as for MI-PA[UB], the time series shows higher values in three periods, albeit slightly different compared to the previously described site: winter/early springtime, from June to August, and from mid-October to early November. The growth in July of the annual cycle is dominated by high peaks: this is due to very high hourly average concentrations (up to 740 µg/m³) observed every second year since 2018. L-EPA technicians verified that fertilization operations of nearby agricultural fields were in progress during those events. Nevertheless, the geometric annual mean for the NH₃ level amounts to 13 µg/m³.

The highest concentrations are recorded at the CdC[farmRB] station, as expected. At this site, the yearly geometric mean level reaches 37 µg/m³. Although the maximum concentration measured on an hourly basis is 708 µg/m³, comparable to that measured in SKI[RB], the CdC[farmRB] site significantly differs from SKI[RB]. This difference is evident when considering the 90th percentile value of the two sites (28 µg/m³ at SKI[RB] compared to 110 µg/m³ at CdC[farmRB]), as reported in Table 2. In addition, the annual cycle at CdC[farmRB] station shows two main periods of rising concentrations: from February to the beginning of April and from July to November, the latter one is preceded by 2 months of a modest increase in concentrations. At the Mo[Rem] site, the annual cycle shows a distinctly different behavior. Broadly speaking, NH₃ concentrations are higher during summertime and very low (sometimes even below the detection limit) during winter, and the overall shape is an upside-down “U.” However, data collected at the Mo[Rem] station from 2007 to 2020 suggest even here the influence of NH₃ in three distinct periods. Rises in NH₃ concentration pattern are observed from February to April, during June and July, and during September and October. Nevertheless, the annual mean concentration is about 1 µg/m³.

Averaged concentrations during three representative periods, i.e., from mid-February to mid-April, from mid-May to mid-June, and from mid-September to mid-November, are calculated and shown for each site. Further statistics on an hourly basis are summarized in Table 2.

In Figure 4, the atmospheric NH₃ concentrations are presented as weekly arithmetic means for the four sites previously described. The weekly time resolution allows appreciation of the variability of the shown data, avoiding excessive scatter. On the right axes, the maximum hourly data in the specific average week are also shown. Although two high values were detected in MI-PA[UB] in 2009 and 2012 (up to 180 µg/m³ as maximum hourly), NH₃ weekly average concentrations typically remain within a range between 3 and 30 µg/m³. As shown with the dashed red line in the figure, the time series indicates a negligible cycle over the years for MI-PA[UB]. As previously mentioned, SKI[RB] confirms to be characterized by low variability as well, besides rare events connected to soil fertilization (the direct link between fertilization and high levels is shown in Section 3.3). For this site, a negligible cycle is observed too, even though a slight increase of the interannual trend is shown in this case. A more appreciable reduction of NH₃ interannual trend levels is suggested by the time series of CdC[farmRB] site, also displayed by the mean annual cycle (Figure 5). The Mo[Rem] site, instead, clearly shows that values above the detection limit are mainly detected when the warmer season starts and again decrease after it.

In general, a clear interannual trend is not evident; on the whole, the NH₃ values measured by the L-EPA Air Quality Network seem quite stable, as shown in Figures 4, 5. The data, therefore, suggest not great changes in the activities emitting NH₃ into the atmosphere, over the years, with a few exceptions. As already mentioned, the reduction in annual average concentrations measured at CdC[farmRB], which affects the whole maximum hourly dataset depicted in Figure 2, can be attributed to local improvements in waste management practices and in a change in the number of pig livestock. On the other hand, the monthly average cycle (Figure 6) confirms the two main periods of rising concentrations, with the highest values in March and August to September. At SKI[RB], concentrations exhibit a slight positive pattern, although strongly influenced by high concentrations observed in July 2018, 2020, and 2022, as also highlighted in Figure 4. MI-PA[UB] is influenced by transport events that are unlikely to result in a similar increase in values as observed in locations near direct emissions. Mo[Rem] shows a completely different graph than other sites, showing a bell-trend: the highest concentrations are measured during warmer periods reaching the maximum value in June.

As demonstrated in the previous section, specific periods of the year are typically marked by higher NH₃ concentrations, particularly in agricultural areas. As suggested by emission inventory databases, these variations could indicate that some specific agricultural and zootechnical activities contribute significantly to NH₃ emissions. For this reason, starting from 2017, an examination of the impact of the agricultural sector on NH₃ concentrations has been conducted. In this context, four monitoring campaigns at different animal husbandry activities along the Lombardy region have been carried out to compare different fertilization techniques on agricultural fields. The considered techniques were (1) surficial spreading, (2) direct injection within the soil, (3) nebulization by means of a pivot 20 cm above the ground level of N-abate and micro-filtrated slurry, and (4) fertigation of N-abate and micro-filtrated slurry. Among these, the first two are the most used techniques in the Po Valley. In this section, we briefly describe the results of the first technique. The monitoring campaign was conducted in an agricultural area between Milan and Bergamo cities in September 2018. The comparison among the different techniques will be discussed in a work that is in progress.

Ammonia concentrations were measured by a chemiluminescence monitor installed on a mobile laboratory positioned near the border of the agricultural field (Figure 7A). Data were acquired with a 1-min time resolution starting from 3 days before the fertilization event, dated 26 September 2018. During this time range, observations displayed NH₃ concentrations lower than 20 µg/m³. The superficial spreading fertilization used in this area consists of a rotating disk, which spreads the fertilizer on the soil from a height approximately 3 m above ground level. Due to the emission and strong volatilization caused by the spreading technique, atmospheric NH₃ concentrations increased from the above-mentioned value up to 1,300 µg/m³. Conversely, when the monitor was upwind of the fertilized field, concentrations were lower than 50 µg/m³. In this regard, the polar plot in Figure 7B reports the probability that the measured concentrations above the 50th percentile come from the target fertilized area. Ammonia concentrations rapidly decreased during the following days, thanks to the regulations that order such fertilizers to be buried within 48 h after spreading. This led the concentrations to decrease below 100 µg/m³ on 29 September.

As already mentioned, the Po Valley is a hotspot regarding PM concentrations. Many studies focused their attention on the chemical characterization of atmospheric particles in this basin and provided averaged PM composition both on an annual (73–75) and seasonal (76–78) basis. Following the results shown in the Life PrepAIR—Interim Report (18), in Milan, the averaged composition of PM₁₀ in the last 10 years is due to SIA (34%), organic carbon (OC) (20%), crustal matter (12%), elemental carbon (EC) (4%), and other trace elements (2%). Carbon compounds remain quite constant in percentage from summer to winter, while SIA increases up to 38% in the cold season. By focusing solely on the results of chemical analyses associated with days surpassing the 50 µg/m³ limit, the influence of NH₄NO₃ becomes distinctly evident. In instances where the limit is exceeded, NH₄NO₃ constitutes 28 ± 10% of the PM₁₀, contrasting with the 17 ± 13% observed in cases below the limit. Beyond a mere distinction between below and above the limit, it is evident that NH₄NO₃ plays a crucial role in determining PM₁₀ levels in the Po Valley. Thus, the study of NH₃ concentrations in relation to its sources and meteorological phenomena is important to understand its role in secondary inorganic compounds formation. In this section, data regarding secondary inorganic compounds are presented. Figure 8 shows the results of the chromatographic analysis performed on about 4,000 PM₁₀ samples. Daily concentrations were aggregated in monthly time resolution and then used to highlight the annual cycle. The ammonium sulfate compound shows very low variability, ranging from 2.0 to 3.5 µg/m³ and no seasonality (Figure 8C). It is worth noticing that MI-PA[UB] and SKI[RB] show the same annual cycle and the same absolute value of this compound.

On the contrary, NH₄NO₃ is found to contribute up to 40% of PM₁₀ mass on a monthly basis in colder periods, which can reach up to 60% of PM₁₀ on a daily average basis (Figures 8D,E).

To identify a clearer relationship within the non-linear (20, 79) NH₃–NO_x–NH₄NO₃ system, further investigation was carried out to observe the cycle of NH₄NO₃ during one of the periods with the highest probability of elevated NH₃ concentrations, based on the levels of the previous years. The investigation has been conducted through an intensive monitoring campaign between February and March 2022 at six different sites, as explained in Section 2.4. Figure 9 displays the variability of two parameters, gaseous NH₃ and NH₄NO₃ on PM₁₀, over a 6-h sampling period. Gaseous NH₃ concentrations exhibit high variability among the monitoring sites (Figure 9A), suggesting that the emissive source and its distance from the measurement site have the greatest impact on the determination of atmospheric concentrations. Within the considered period, the CdC[farmRB] site shows the highest values (57 ± 35 µg/m³) followed by Be[RB] (47 ± 23 µg/m³), whereas the lowest ones are at MI-PA[UB] (9 ± 2 µg/m³) and SdB[UI] (6 ± 3 µg/m³) sites. On the contrary, NH₄NO₃ (Figure 9B) shows an extremely low variability. The mean concentrations are in a range from 7 ± 6 µg/m³ [at SKI(RB)] to 11 ± 7 µg/m³ [CdC(farmRB)]. Figure 10 allows us to observe the NH₃ and NH₄NO₃ cycles during the intensive campaign in CdC[farmRB]. Although the determination coefficient is very low (R²= 0.2), the data point out a common pattern in specific events. This pattern is observed across all individual measurement sites, indicating a potential direct relation between these two parameters.

The selected locations enable us to investigate how concentrations of this gaseous compound may vary depending on the proximity of the primary source identified by the emission inventory. Figure 11’s polar plots illustrate the relationship between NH₃ levels, wind speed, and direction.

SKI[RB] and CdC[farmRB] sites are surrounded by agricultural fields. Source apportionment analysis at SKI[RB] suggests that traffic and/or industrial activities collectively contribute about 10% to the air quality impact. Monitoring at CdC[farmRB] in 2014 revealed low concentrations of SO₂ and NO_x, proxies for industrial and combustion sources. The regional emission inventory (INEMAR) emphasizes that these two sources contribute less than 5% of total NH₃ emissions in the Schivenoglia and Corte de’ Cortesi administrative areas. Conversely, NH₃ emissions are primarily attributed to the agricultural sector, particularly swine husbandries, accounting for over 75%. Thus, SKI[RB] and CdC[farmRB] sites reinforce the idea that NH₃ concentration levels are strongly influenced by the proximity of the emissive source.

At CdC[farmRB], the polar plot indicates that the highest values are measured under no-wind or low-wind conditions, supporting a local source, such as animal husbandries. As wind speed increases, concentrations rapidly decline, though remaining high due to a persistent high background condition. Supplementary Figure S3 further supports this evidence.

The conditional functional plot, generated by eliminating extreme cases, enables the determination of the source of concentrations below the 80th percentile (Supplementary Figure S3A). This allows observation of the homogeneity of the probability of such concentrations across all wind directions and speeds, indicating that the calculated values are consistent with the local agricultural background in the Schivenoglia area. On the other hand, a local source is also suggested to affect the SKI[RB] site with high NH₃ levels: the polar plot (Supplementary Figure S3B) displays the highest concentrations in low-wind conditions but also highlights that transport events can increase the background values. Limiting the observation only to the data over the 90th percentile, or 27 µg/m³, confirms that the source is indeed local, but also the concentrations in these cases are indicative of a transport process with wind speeds below 4 m/s. The little variability in the direction of origin of such concentrations suggests that the monitoring station detects activities carried out at west–northwest directions.

Figures 3, 4 demonstrate that hourly NH₃ concentrations for SKI[RB] and CdC[farmRB] can reach 700 g/m³. This aligns with the findings of the campaign described in Section 3.2: soil fertilization by means of animal manure or slurry strongly affects the detected levels of NH₃. Table 2 reports the arithmetical averages calculated on an hourly time resolution data in four periods, i.e., mid-February to mid-April, mid-May to mid-July, mid-September to mid-November, and the remaining periods of the year. Based on the information about agricultural practices, the observations found an explanation referring to the detected values. These are the periods in which soil fertilization occurs, although each agricultural area has its own particularity. In addition, meteorological parameters (such as temperature) affect resulting NH₃ concentrations due to volatilization variability. Both these facts explain the lowest values detected for SKI[RB] and CdC[farmRB] during December and January (and for the entire L-EPA network): in these 2 months, spreading procedures are prohibited or heavily limited within the entire Po basin, and NH₃ volatilization is prevented by the lower temperature that, at the same time, enhances gas-to-particles partitioning in the aerosol phase.

The Mo[Rem] site, as depicted in Figure 6D, shows distinct variability compared to other sites. In particular, the highest NH₃ concentrations are measured during warmer periods and the lowest during colder ones. This observed pattern is attributed to the mixing layer height (Supplementary Figure S4): considering its location, i.e., about 1,200 m a.s.l, the site is always above the mixing layer from October to April, with the exception of some isolated events where meteorological conditions favor convective motions vertically, raising the mixing height to levels compatible with that of the site. Otherwise, during this period, the vertical stability of the atmosphere hinders compounds emitted by the Po basin from reaching altitudes above the mixing layer height. As a result, NH₃ concentrations are frequently below the instrument's quantification limit. On the contrary, the maximum monthly average is in June (Figure 6D) when the site is within the mixing layer, allowing for the measurement of concentrations of air masses transported from the Po Basin (9, 14–16, 80, 81).

MI-PA[UB] (Figure 11A) demonstrates that the main source of atmospheric NH₃ comes from the plain located ESE of the site, whereas the lowest values are detected in cases of wind reinforcement that cleans the atmosphere. The graph also indicates a slight increase in concentrations under light or gentle breezes (2–4 m/s) from the west, suggesting that these increases may originate from livestock activities on the opposite side of the city of Milan or from vehicular traffic.

It is known that in urban sites, a significant source of NH₃ is vehicular traffic. NH₃ emissions from gasoline vehicles equipped with a three-way catalyst (TWC) are an important source of NH₃ in areas with heavy traffic (29, 82, 83), since it is generated as a side product in the NO_x reduction process (84). Furthermore, the recent introduction of the selective catalytic reduction (SCR) system with the addition of urea or NH₃ in heavy-duty vehicles (HDV), and mandatory since 2016 for Euro 5 and Euro 6 vehicles, resulted in increased NH₃ emissions from traffic (85), which needs further investigation.

To verify that a fraction of the NH₃ at the MI-PA[UB] site may be influenced by this source, the concentrations measured in December and January were compared with Black Carbon values (Section 2.4; Supplementary Figure S5), a well-known marker for vehicular combustion (86). In this period, regional limitations ban spreading activities according to the Nitrates Directive [and its Italian regulatory transposition DM 5046/2016 (87)]. The resulting correlation demonstrates a good level of agreement between these two parameters (R²_adj= 0.661), confirming a relative contribution from this source. With the same hypotheses, an attempt was made to verify whether the addition of urea had repercussions on NH₃ concentrations in Milan. The historical pattern (Figure 4) does not show an increase. It is necessary to remember that the restrictions introduced during the COVID-19 pandemic have certainly had a positive effect in reducing the impact of vehicular traffic. However, similar to before, taking into consideration the months of December and January when there is a good correlation between atmospheric NH₃ and traffic, and excluding only December 2020 and January 2021, which suffered the restrictions of the second wave of COVID-19, the NH₃ concentrations were averaged over the two periods: 2007–2016 and 2017–2022 with 6.4 and 11.8 µg/m³, respectively. This result cannot be used as verification of an increase in NH₃ concentrations due to the addition of urea but suggests the possibility of further investigations.

The representative cycle of the six chosen monitoring sites, as detailed in Section 3.3, appears to confirm a qualitative correlation between gaseous NH₃ and NH₄NO₃ in the aerosol phase. However, the low coefficient of determination equally indicates that the relationship between the two compounds is not linear (Section 3.3). The intensive period spans 21 days (from late February to mid-March) and is characterized by meteorological stability, with no rainfall, an average wind speed of 2 m/s, and atmospheric pressure of 1,014 hPa. In addition, thermodynamic air parameters, including humidity and temperature, are at levels that could favor the partitioning of NH₄NO₃ into the particulate phase.

The variability of NH₄NO₃ during this period (Figure 9B) highlights three events with a significant increase in concentrations occurring on (I) 25 February, (II) 4 March, and (III) 10–11 March. Compared to an average over the period of 22% of NH₄NO₃ in PM₁₀, the contributions were 37%, 42%, and 33%, respectively. These three periods were investigated by analyzing the PM₁₀ and NO_x patterns, the mixing layer height, and relative humidity (Figure 12). The graphs present cycles as moving averages, helping to smooth out the time series curve by computing the average of all data points in a fixed-length window. It can be observed that in the first and third episodes, the increase in concentrations was followed by a significant decrease in the mixing layer height, reaching the lowest value in the campaign between 10 and 11 March. The second episode of increased NH₄NO₃ occurs instead in conjunction with an increase in the mixing layer height, which remained quite high even in the earlier time slots. The relative humidity also increases together with NH₄NO₃, while precursors decrease. These factors suggest the possible occurrence of an NH₄NO₃ formation event.

On the other hand, Figure 13 shows the average daily concentration cycle for both compounds for 1 year (2019 was chosen as an example). During the cold seasons, particularly for many peak episodes in spring and fall, it can be noticed that when NH₃ increases, NH₄NO₃ also increases. However, several events contradict a direct cause–effect connection: in warmer periods, the condensation into NH₄NO₃ is inhibited by the temperature, which favors the evaporation of the nitrates. Moreover, the formation of NH₄NO₃ is also caused by accumulation phenomena, which occur frequently in the Po Valley, and by combustion sources. Ammonium nitrate formation, and secondary inorganic aerosol formation in general, is a complex process influenced not only by the concentration of its precursors but also by thermodynamics and meteorological conditions (20, 79), as already discussed. Nenes et al. (88) and Thunis et al. (89) published two different works about the PM_2.5 response to HNO₃ or NO_x and NH₃ emissions changes. Their modeling approaches converge to similar results, i.e., a variation in one of the two gaseous precursors does not lead to a linear change in PM_2.5 concentrations and, in some cases, it could lead to the opposite result.

Nevertheless, it is worth noting that the findings presented in this study are based on offline analyses conducted on PM₁₀ samples. These methodologies can be prone to negative artifacts, which may significantly underestimate the semi-volatile component of particulate matter. Past studies by Minguillón et al. (90) and Poulain et al. (91) have effectively compared offline and online analyses, the latter utilizing data from the Aerosol Chemical Speciation Monitor (ACSM, Aerodyne Research Inc.). Importantly, within the scope of this study, it is essential to highlight that offline and online nitrate correlations exhibit slopes greater than 6 during the summer months. This suggests that offline results obtained during sampling under high-temperature conditions could consistently underestimate NH₄NO₃ concentrations, thereby reducing the observable relationship between concentrations of gaseous precursors and the resulting particulate phase.