Measurement report: Nitrogen isotope (δ¹⁵N) signatures of ammonia emissions from livestock farming: implications for source apportionment of haze pollution

Ammonia emissions from agriculture are the primary source of atmospheric reactive nitrogen, significantly impacting air pollution, soil acidification, eutrophication of water bodies, and human health. Accurate quantification of ammonia from different sources is crucial for effective mitigation. In this study, the air extraction method was employed to collect gases from livestock farms, and the δ¹⁵N values of volatilized ammonia (NH₃) from the animal husbandry industry in the southern Huang – Huai – Hai Plain of China were analyzed using stable nitrogen isotopes. The results show that isotopic signatures differ significantly among livestock types: dairy cows (−20.6 ‰ ± 0.8 ‰), laying hens (−27.4 ‰ ± 1.0 ‰), and pigs (−38.4 ‰ ± 1.7 ‰). These livestock-derived signatures are distinct from those associated with combustion sources (−7.0 ‰ ± 2.1 ‰) and traffic emissions (6.6 ‰ ± 2.1 ‰), and they exhibit considerably lower variability than fertilizer-derived signatures. Overall, this work provides high-precision isotopic source signatures for livestock operations, offering essential parameters for regional atmospheric ammonia source apportionment and highlighting the need for locally tailored mitigation strategies.

1 Introduction

Ammonia (NH₃) is a highly reactive and abundant nitrogenous gas in the atmosphere. It is classified as a major alkaline species and readily reacts with sulfuric acid and nitric acid to produce ammonium sulfate ((NH₄)₂SO₄) and ammonium nitrate (NH₄NO₃) (Kawashima et al., 2023; Kirkby et al., 2011). These compounds can form particulate ammonium salts or interact with organic aerosols to generate secondary aerosols. In moderately polluted environments, the mass fraction of these ammonium-containing particles within PM_2.5 is relatively low (Huang et al., 2014; Yang et al., 2011). Under severe pollution conditions, however, ammonium sulfate, ammonium nitrate, and other ammonium salts can account for up to approximately 50 % of the total PM_2.5 mass (Battye, 2003; Beusen et al., 2008; Goebes et al., 2003). As a key precursor of secondary inorganic aerosols, NH₃ is a primary contributor to haze formation and constitutes a substantial component of PM_2.5 in polluted atmospheres (Wu et al., 2024; Xiang et al., 2022). Excessive ammonia emissions also drive a range of environmental problems, including soil acidification, climate perturbation, reduced atmospheric visibility, and eutrophication of aquatic ecosystems (Huang et al., 2012; Jiang et al., 2021). Consequently, reducing NH₃ emissions has recently been proposed as a strategy to mitigate smog pollution in China (Liu et al., 2019).

Over the past few decades, substantial changes in air quality have been observed across many countries worldwide (Boyle, 2017; Warner et al., 2017). Notably, China has consistently ranked first in global ammonia (NH₃) emissions (Liu et al., 2013). Current NH₃ emission inventories identify the principal sources as agricultural activities-including fertilizer application and livestock and poultry farming-and non-agricultural sources, such as combustion processes and vehicular emissions (Bouwman et al., 1997; Schlesinger and Hartley, 1992; Streets et al., 2003). It is widely recognized that agriculture represents the predominant source of atmospheric NH₃, contributing over 70 % of total emissions (Meng et al., 2017; Xu et al., 2024), accounting for more than 70 % of the total (Ma et al., 2021; Ti et al., 2019), with livestock and poultry farming alone accounting for 50 % to 60 % of agricultural NH₃ emission (Huang et al., 2012; Wang et al., 2018). Despite this, substantial uncertainty remains regarding the contribution of livestock-derived NH₃ to nitrogen deposition (Elliott et al., 2019), and estimating these contributions using satellite remote sensing and livestock emission inventories remains challenging (Beusen et al., 2008; Li et al., 2023a; Van Damme et al., 2018). These conventional approaches typically rely on fixed emission factors, such as unit animal excretion coefficients, which are limited by temporal lags and insufficient spatial resolution, thereby hindering the capture of real-time variations in NH₃ emissions and the resulting nitrogen deposition at the farm scale. In contrast, nitrogen stable isotope analysis (δ¹⁵N) provides a direct and highly effective approach for tracing the sources of NH₃ and NH${}_{\mathrm{4}}^{+}$ (Bhattarai et al., 2020; Xiao et al., 2020). This methodology relies on the principle that distinct emission sources and environmental processes generally exhibit unique isotopic fingerprints (Elliott et al., 2019; Li et al., 2024; Sui et al., 2020), defined by the ratio of heavy (¹⁵N) to light (¹⁴N) nitrogen isotopes in collected samples (Song et al., 2021).

Numerous studies have employed stable nitrogen isotope (δ¹⁵N) techniques to quantify the contributions of combustion, transportation, and agricultural activities to atmospheric NH₃ and NH${}_{\mathrm{4}}^{+}$ (Xiang et al., 2022; Xie et al., 2008). For example, during the corn growing season in Northeast China, δ¹⁵N values of NH₃ volatilized from farmland exhibited a wide range, from −38.0 ‰ to −0.2 ‰. Notably, δ¹⁵N emission rates were considerably lower during the early stages of corn growth compared to later stages, indicating clear seasonal variation (Song et al., 2024). Under different fertilization regimes, significant differences in δ¹⁵N-NH₃ emissions were observed, with values fluctuating between −46.0 ‰ and −4.7 ‰ throughout the volatilization period (Ti et al., 2021). Previous studies report that δ¹⁵N-NH₃ and δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ emissions from combustion sources (−7.6 ‰ to +16.2 ‰) predominate in winter, contributing up to 51.6 % of total ammonia emissions (Xiao et al., 2022, 2025; Zhou et al., 2021). In contrast, NH₃ emissions from vehicle exhaust exhibit relatively high δ¹⁵N values (13.7 ± 3.7 ‰) (Savard et al., 2017; Xi et al., 2023). However, these emissions are primarily localized in urban environments.

Currently, limited studies have reported the δ¹⁵N characteristics of ammonia from livestock and poultry farming. Existing data mostly rely on passive sampling methods (Berner and David Felix, 2020; Chang et al., 2016; Ti et al., 2018), which assess δ¹⁵N changes by collecting wet deposition samples surrounding farms (pig farms: −35.1 ‰ to −10.5 ‰; cattle farms: −24.7 ‰ to −11.3 ‰). Additional research has quantified δ¹⁵N variability in livestock and poultry (−31.0 ‰ to −15.0 ‰) through simulated ammonia emissions during manure management processes (Hristov et al., 2009). It is noteworthy that δ¹⁵N-NH₃ fluctuations in livestock and poultry operations may also depend on animal growth stages and reproductive status

The Bayesian stable isotope mixing model MixSAIR is primarily used to allocate contributions of atmospheric emission sources through isotope analysis. MixSIAR is a stable isotope mixing model based on the Bayesian statistical framework, designed to quantitatively analyze the relative contributions of multiple potential sources to the isotopic composition of observed mixtures. Its fundamental assumption posits that the isotopic signature of a mixture can be expressed as a linear combination of the isotopic characteristics of each source weighted by their proportional contributions, while explicitly accounting for source variability, measurement errors, and isotopic fractionation. (Chang et al., 2016; Walters et al., 2022). However, there is no universally fixed δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ value for each emission source. As a result, substantial variations in reported δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values for the same source have been documented across different studies. To date, no research has validated changes in δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ resulting specifically from livestock and poultry farm emissions, nor has the relationship between δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ from different sources and regional variations been examined. To obtain more accurate assessments of δ¹⁵N-NH₃ variations associated with ammonia emissions from livestock and poultry farming, and to achieve reliable atmospheric NH₃ source apportionment, it is essential to characterize the correlation between δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ from different sources and regional differences. In this study, active dynamic sampling methods were used to collect ammonia emissions from intensive pig farms, dairy farms, and laying hen farms located in the southern region of the Huang-Huai-Hai Plain. Meta-analysis techniques were employed to analyze the δ¹⁵N signatures of different ammonia emission sources. The specific objectives of this research are: (1) to determine the δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values of emissions from livestock and poultry housing at various growth stages; and (2) to investigate the relationship between δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ from different sources and regional variations.

2 Materials and methods

2.1 Sampling points in the study area and sample collection and processing

The sampling experiment at the farm was conducted from 9 May to 6 December 2024. No samples were collected in July and August due to the absence of livestock or poultry during these months. The collected samples covered the entire breeding period of fattening pigs and the period from chicks to peak egg production in laying hens. Throughout the trial period, six batches of samples were obtained, amounting to a total of 120 samples for measuring ammonia emissions from livestock and poultry housing. On days when samples were collected during hazy weather, the air pollution level was classified as severe, whereas samples collected under clean atmospheric conditions corresponded to air quality classified as excellent. The sampling principle is based on active air sampling combined with aqueous absorption. Ambient air was continuously drawn through an impinger containing deionized water, in which gaseous NH₃ was absorbed and converted to dissolved NH${}_{\mathrm{4}}^{+}$. After sampling, the absorption solution was quantitatively recovered for subsequent laboratory analysis. Under the applied sampling flow rate and duration, the method detection limit for atmospheric NH₃ was on the order of 3 ppb, which is adequate for resolving ambient concentration variations during the observation period. Potential interferences include the co-collection of particulate NH${}_{\mathrm{4}}^{+}$ and the absorption of other water-soluble alkaline gases. These effects were minimized by controlled sampling duration, appropriate flow rates, and blank correction procedures, and are considered to have a negligible influence on the measured NH₃ concentrations. Samples were collected using atmospheric samplers (Beijing Ke'an Labor Protection Company) at a flow rate of 0.1 to 2 L min⁻¹, with each sample collected over a duration of 60 min (Ferm, 1979; Harrison and Kitto, 1990; Heaton, 1986). All NH₃ samples were collected at a constant and identical flow rate throughout the study period, with all flow rates complying with the National Standard GB 3095-2012.

The intensive fattening pig farm is located in Luoyang City, Henan Province (112.71° E, 34.52° N), with no other livestock operations in the surrounding area. The sampled fattening pig farm houses 2600 pigs distributed across four fully enclosed pig houses. One of these houses was selected as the target sampling site. The sampling procedure was as follows: an atmospheric sampler was positioned 2.0 m from the exhaust vent of the livestock and poultry house at a height of 1.6 m, corresponding to the central height of the exhaust outlet. The sampling duration was set to 60 min, with the gas flow rate maintained at 2 L min⁻¹ using a flow meter. According to the sample requirements for mass spectrometry pretreatment, we used deionized water as the absorption solution. A bubbler absorption bottle filled with absorption solution was used to collect NH₃. Three atmospheric samplers were operated simultaneously during each sampling event. Figure 1 marks the sampling points of the intensive pig farms with green pentagrams. Sampling was conducted in a typical commercial intensive laying-hen house with a conventional cage-based rearing system, representative of large-scale laying-hen farms in northern China. We collected gases from the exhaust vents of chicken houses and pig houses. These types of animal housing have centralized air inlets and outlets, so collecting from the exhaust vents can represent the ammonia emissions from these two types of housing into the atmosphere. For cattle farms, since the barns are open, we selected cattle sheds located in the middle of the farm to more effectively collect ammonia gas.

Figure 1 Sampling sites of livestock farms, haze weather, and clear weather in this study, extracted from the main research sampling locations. Yellow dots represent the main global research sampling sites, pink triangles represent sampling sites during haze and clear weather, dark blue pentagons represent cattle farms, light blue pentagons represent layer farms, and green pentagons represent fattening pig farms.

In the case of intensive laying hens farms, each building houses approximately 15 000 laying hens and is fully enclosed, with a total of 300 000 laying hens being raised. The sampling site is located in Zhengzhou City, Henan Province (114.03° E, 34.59° N). One building was selected as the target sampling point, with the sampling method mirroring that used for the fattening pig farms. As shown in Fig. 1, the light blue pentagons represent the sampling points of intensive layer farms.

The intensive dairy farm operates with an open-style barn design, housing 400 dairy cows per barn, with a total of 4000 dairy cows being raised. Four atmospheric samplers were installed in the passageways of the dairy barns, with each sampler spaced 10 m apart and positioned at a height of 1.6 m. The dairy farm is located in Zhengzhou City, Henan Province (114.11° E, 34.81° N). The sampling time and method remained consistent with those described above. In Fig. 1, the dark blue pentagons represent the sampling points of intensive dairy farms.

To investigate the variations in δ¹⁵N levels associated with differing degrees of air pollution, samples collected for δ¹⁵N measurement during periods of severe smog and when air quality was pristine. The sampling location was situated on a spacious lawn within the campus of Henan Agricultural University, devoid of tall buildings or traffic. The sampling point is illustrated in Fig. 1, where the pink triangle represents the sampling site for both haze and clean air (Longitude 113.82° E, Latitude 34.80° N). Each sampling event utilized three atmospheric samplers, positioned at a height of 1.6 m, with the duration of sampling aligned with that of the livestock farm.

The collected sample solution is transferred into a centrifuge tube and returned to the laboratory, where the concentration of NH₃ is measured using a UV spectrophotometer. The detection method adheres to the guidelines outlined in “Determination of Ammonia Nitrogen in Water by Salicylic Acid Spectrophotometry” (HJ 536-2009), and the calculation method is presented in Eq. (1):

$$\begin{array}{}\text{(1)}& {\mathit{\rho }}_N={\displaystyle \frac{A_{\mathrm{s}}-A_{\mathrm{b}}-a}{b\times V}}\times D\end{array}$$

where, ρ_N represents the mass concentration of ammonia nitrogen in the water sample (expressed as N), in mg L⁻¹. The variables are defined as follows: A_s denotes the absorbance of the sample, while A_b indicates the absorbance of the blank experiment, which is prepared from the same batch as the sample. The parameters a and b correspond to the intercept and slope of the calibration curve, respectively. Additionally, V refers to the volume of the water sample taken, measured in mL, and D signifies the dilution factor of the water sample.

The analytical method for N isotope determination employs the hypobromite-hydroxylamine hydrochloride chemical method (Song et al., 2024) (Soler-Jofra et al., 2016; Zhang et al., 2007). Initially, a potassium bromate-potassium bromide solution reacts under acidic conditions to produce bromine, which subsequently reacts in a strongly alkaline environment to generate bromate, a potent oxidizing agent capable of oxidizing NH${}_{\mathrm{4}}^{+}$ to NO${}_{\mathrm{2}}^{-}$. In the following step, hydroxylamine hydrochloride reduces NO${}_{\mathrm{2}}^{-}$ in an acidic environment to form N₂O. The resultant N₂O is then analyzed using a stable isotope ratio mass spectrometer, along with a multi-purpose online gas preparation device, and an automatic sampler, to determine the δ¹⁵N value. For each sample analysis, four international standard materials for NH${}_{\mathrm{4}}^{+}$ (IAEA-N-1, USGS-25, IAEA-N-2, and USGS-26, with δ¹⁵N concentrations of 0.4 ‰, −30.41 ‰, 20.3 ‰, and 53.75 ‰, respectively) are processed simultaneously. NH₃ concentrations and δ¹⁵N values are presented as mean ± standard error (SE). Differences in δ¹⁵N values among livestock categories were evaluated using one-way analysis of variance (ANOVA). When data did not meet the assumptions of normality or homogeneity of variance, non-parametric tests were applied. Statistical significance was defined at p<0.05. All statistical analyses were conducted using standard statistical software.

2.2 Data collection and processing

We screened articles published between January 2000 and January 2025 regarding the sources of δ¹⁵N-NH₃ and δ¹⁵N-NH${}_{\mathrm{4}}^{+}$. Specifically, we utilized ISI Web of Science, Google Scholar, and PubMed, employing the search terms “δ¹⁵N,” “NH₃,” “ammonia emissions,” and “isotopes” to identify relevant literature. Studies included in our analysis were required to meet the following criteria: (1) Samples must be measured for either δ¹⁵N-NH₃ or δ¹⁵N-NH${}_{\mathrm{4}}^{+}$; (2) Experiments must encompass at least one of the following: combustion, fertilization, agriculture, transportation, or livestock farming; (3) The number of experimental replicates and sampling events must be explicitly reported; (4) Samples must primarily consist of atmospheric NH₃ or PM_2.5, and detection must employ chemical methods. A total of 37 documents were included in the analysis. This dataset comprehensively encompasses multiple meta-analyses and original studies, detailing changes in δ¹⁵N-NH₃ and δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ from combustion sources, transportation sources, agricultural sources, and livestock farming sources; the proportion of δ¹⁵N values in the atmosphere; geographical location (latitude and longitude); and the GDP of each city where samples were collected. If the data in the literature was presented solely in chart form, we utilized WebPlotDigitizer-4.7 (https://apps.automeris.io/wpd4/, last access: 20 May 2025) to extract the data. We categorized the collected data into five distinct groups: combustion, transportation, farmland, livestock farming, and PM_2.5. To ensure reproducibility, literature-derived δ¹⁵N-NH₃ values were synthesized following a consistent aggregation protocol. When multiple isotopic values for the same source category were reported within a single study, a sample-size-weighted mean was calculated if the number of samples (n) was explicitly provided. In cases where sample size information was unavailable, simple arithmetic means were used, and the resulting uncertainty was reflected by expanding the reported end-member range. No additional weighting based on study duration or subjective data quality scores was applied, in order to avoid introducing implicit bias across studies. Differences between sampling methodologies were explicitly considered. Active sampling studies, including the present work, were prioritized for constraining source end-member values. Passive sampling data were used only for qualitative comparison, as previous studies have demonstrated systematic low biases in δ¹⁵N–NH₃ derived from passive samplers relative to active methods. Consequently, passive sampling results were not directly incorporated into end-member mean calculations used for isotope mixing analyses.

A total of 126 samples were collected, and 41 literature references were gathered. Data analysis was performed using Excel, SPSS, and Python version 3.11.

3 Result and discussion

3.1 Temporal Variations in Ammonia Emissions and δ¹⁵N Signatures from Livestock Farms

During the sampling period from May to December, ammonia emissions varied significantly among the three farm types: 4.9 to 6.7 mg m⁻³ for fattening pigs (Fig. 2a), 1.7 and 2.5 mg m⁻³ for dairy cows (Fig. 2b), and 3.8 to 7.1 mg m⁻³ for laying hens (Fig. 2c), with the latter exhibiting substantial temporal fluctuations. NH₃ emissions from fattening pigs peaked when the pigs reached 130 kg per head (Fig. 2a). For laying hens, NH₃ concentrations initially increased and subsequently declined in response to temperature variations, reflecting enhanced urease activity within the housing environment, which accelerates urea hydrolysis and promotes NH₃ volatilization. δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ levels at the livestock farms showed significant temporal variation (p<0.05) (Groot Koerkamp et al., 1998; Rosa et al., 2020). From May to June, the δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ increased from −31.0 ‰ to −25.2 ‰ in fattening pig farms and from −26.4 ‰ to −24.6 ‰ in laying hen farms. In September, δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values from fattening pig farms (−13.3 ± 1.3 ‰) were significantly higher than those from laying hen and dairy cow farms (−13.9 ± 0.9 ‰), which were comparable. Over the following three months, δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ levels decreased significantly across both farm types. Although relatively large variability was observed within each livestock category, the differences in mean δ¹⁵N values among groups were statistically significant (one-way ANOVA, p<0.05). The large within-group variability reflects realistic operational and environmental heterogeneity and does not negate the statistically significant differences observed among livestock categories. As illustrated in Fig. 2, the highest NH₃ concentration at the dairy farm (2.5±0.3 mg m⁻³) occurred in October, coinciding with the lowest δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values. while laying hen farms also recorded minimum δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ during this period of elevated NH₃. Conversely, the lowest δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ at fattening pig farms was observed in December, despite peak NH₃ concentrations. NH₃ concentrations differed significantly between hazy and clear weather in December (Fig. 2d), with δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values being significantly higher under clear conditions (1.9 ± 0.8 ‰) than under hazy conditions (1.6 ± 0.2 ‰; p<0.05).

Figure 2 Changes in NH₃ emissions and δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values outside the livestock farms among different months. (a) Fatting pig farm; (b) Dairy cow farm; (c) Laying hens farm; (d) Comparison of Haze and clean air samples. Statistical difference was calculated by t-test, p<0.05, n=3.

As illustrated in Fig. 3, throughout the entire monitoring period, ammonia (NH₃) sources form the farms exhibited nitrogen depletion, indicated by negative δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values. Overall, δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values exhibited significant fluctuations in dairy and fattening pig farms, while variations were comparatively moderate in laying hens farms. Notably, the δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values at dairy cattle farms displayed substantially greater overall changes during the monitoring period compared to those in laying hens and fattening pig farms. The arithmetic mean value at fattening pig farms was −30.8 ± 1.6 ‰, the lowest among the three types of farms, whereas the δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values in laying hens manure remained at an intermediate level throughout the entire period. From October to December, the δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values at livestock and poultry farms were generally lower than those observed in the first half of the monitoring period (Fig. 3). However, when comparing hazy and clear weather conditions, the δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values for all three types of farms consistently remained at a relatively low level during this timeframe (Fig. 3). High temperatures enhance enzyme activity and volatilization, thereby intensifying the isotopic fractionation effect during summer; whereas low temperatures inhibit these processes and reduce isotopic deviations. The nitrogen isotopic signature of livestock-derived ammonia is influenced by various biogeochemical processes, including urea hydrolysis during manure storage, microbial ammonification, and ammonia volatilization (Bhattarai and Wang, 2023; Huang et al., 2012; Li et al., 2023a).

Figure 3 Changes of δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ abandance at intensive livestock farms during the sampling period. Hazy and clean air were also sampled at December. The air sample of laying hens in December was missed, because of death of chicken by avian influenza.

3.2 Comparison with Literature and Implications for Local Sources

During the monitoring period, the δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values ranged from −50.0 ‰ to −10.0 ‰ (Fig. 4a). For fattening pigs, δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values averaged −38.4 ‰ ± 1.8 ‰ between October and December, which was significantly lower than the previously reported range of −27.10 ‰ to −31.7 ‰ (Chang et al., 2016). Notably, the overall variation remained within the δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ emission ranges report for fattening pigs in other studies (Bhattarai and Wang, 2023; Wang et al., 2022). Furthermore, due to differences in livestock management practices and nitrogen content in feed, the δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values from dairy farms in this study, averaging −29.4 ‰ ± 13.9 ‰, were substantially lower than those reported by Martine et al. (20.5 ‰ ± 34.5 ‰) (Savard et al., 2017).

Figure 4 Comparison of δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values within different livestock farms and historical reported data. (a) Comparison of the δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values among different livestock farms; (b) Comparison of the δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values from present study with previously reported data. Boxes represent the interquartile range, the horizontal line within each box denotes the median value, and whiskers indicate the minimum and maximum values excluding outliers. Individual points outside the whiskers represent statistical outliers.

Comparison with δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values measured in dairy farms in Akita, Japan, were −22.5 ‰ ± −14.6 ‰ (Kawashima, 2019), no significant difference was observed relative to the values obtained in this study. However, these values exceeded those reported by Felix et al. (2014), which ranged from −37.9 ‰ to −22.9 ‰ based on passive sampling techniques. Previous research has shown that active sampling generally yields higher δ¹⁵N values than passive sampling (Kawashima and Ono, 2019; Pan et al., 2020). This discrepancy arises from the diffusion-driven nature of passive samplers, in which lighter NH₃ molecules are preferentially adsorbed. Consequently, passive sampling typically produces δ¹⁵N values that deviate by approximately 15 ‰ from those obtained by active sampling (Bhattarai and Wang, 2023; Skinner et al., 2006). Variations in δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values are known to occur among different livestock species. During the monitoring period, δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values from laying hen farms were consistently lower than those from dairy farms but higher than those from fattening pig farms, consistent with previously reported trends (Liu et al., 2025; Ryu et al., 2021). This pattern suggests that δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ variations in emitted NH₃ are not primarily driven by animal body weight but are instead strongly modulated by environmental conditions (Choi et al., 2017; Qu and Zhang, 2021). In agreement with earlier studies, δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ emissions from fattening pig and laying hen farms differed significantly from previously documented values, whereas no significant difference was observed for dairy cattle farms. Furthermore, the magnitude of δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ fluctuations across the three farm types was smaller than that reported in earlier literature. Comparison with major atmospheric NH₃ sources further demonstrated that the δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values measured in this study diverged substantially from those associated with combustion (−7.0 ‰ ± 2.1 ‰), fertilization application (−38.0 ‰ ± 0.2 ‰), and transportation (6.6 ‰ ± 2.1 ‰). Based on δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ signatures measured under both hazy and clear weather conditions, it can therefore be inferred that agricultural and livestock emissions are not the dominant contributors to atmospheric NH₃ in Zhengzhou. Instead, traffic exhaust and combustion sources appear to constitute the primary contributors. We conducted source apportionment for haze and clean weather using the MixSIAR model. The results showed that combustion and traffic were the main contributing sources, with combustion accounting for 29.0 %, traffic for 38.0 %, agriculture for 15.1 %, and livestock for 17.8 % (Stock and Semmens, 2016; Walters et al., 2022; Wong et al., 2022).

The selected pig, dairy, and laying hen facilities are typical of intensive livestock production systems in the southern Huang–Huai–Hai Plain, where feeding strategies, manure management, and ventilation designs are relatively standardized due to regional regulations and industrial practices. Previous studies have shown that while such operational differences can induce secondary variability in δ¹⁵N signatures, their influence is generally smaller than the systematic isotopic contrasts observed among different livestock species.

Importantly, the objective of this study is not to characterize farm-to-farm variability, but to constrain representative isotopic end-member ranges for major livestock categories that can be applied in regional source apportionment frameworks. Within this context, the internally consistent sampling protocol and the clear separation of δ¹⁵N–NH${}_{\mathrm{4}}^{+}$ values among livestock types suggest that the derived signatures are robust for intensive livestock systems operating under comparable management conditions (Choi et al., 2017; Parnell et al., 2010).

Extrapolation of these δ¹⁵N signatures beyond the studied region or to non-standardized, small-scale, or pasture-based livestock systems should be undertaken with caution. Future work incorporating multiple facilities per livestock type and explicit characterization of feed and ventilation parameters would further refine the regional representativeness of livestock-derived ammonia isotope signatures.

Figure 5 Changes of δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values among different GDP cities and years. (a) The relationship between GDP and δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values (p<0.001); (b) Changes of δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values reported between 2008 to 2021 (p<0.001).

3.3 Global Variability of NH₃ Source Signatures and Challenges for Source Apportionment

Ammonia emissions that contribute to urban smog primarily arise from combustion activities, vehicle exhaust, agriculture fertilization, and livestock production. As national economies expand, the frequency and severity of smog events have intensified. Figure 5a (slope: 0.026, intercept: 1.6323, R²: 0.0963) shows that from 2000 to 2025, when GDP remains below USD 70 billion, atmospheric δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ signatures predominantly reflect fertilizer-derived emissions from agricultural regions and NH₃ volatilization from livestock operations (Kawashima et al., 2022; Kawashima and Kurahashi, 2011). This pattern indicates that lower-income regions rely heavily on agriculture and animal husbandry as the foundational components of their economies (Leng et al., 2018).

When GDP increases to between USD 80 and 300 billion, the contribution of combustion-related and vehicular sources to δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ becomes increasingly prominent. Notably, vehicle exhaust remains the dominant contributor within this GDP interval, suggesting that transportation serves as a key economic driver during mid-stage development. In densely populated and economically advanced cities, rapid vehicle growth further amplifies the influence of transportation-related δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ signatures (Lim et al., 2022; Pan et al., 2018; Stratton et al., 2019). Throughout the entire dataset, vehicle exhaust and combustion together account for nearly 70 % of ammonia emissions (Wu et al., 2019). Once GDP surpasses 300 billion USD, δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ from combustion becomes the dominant atmospheric source, while the relative contribution from vehicle exhaust begins to decline and emissions from agricultural fertilization and livestock farming become negligible (Li et al., 2023b). It is important to note that sampling sites in the present study were located near power plants (Lim et al., 2019; Zou et al., 2022), whereas comparison data from previous studies were collected in urban cores. This spatial difference further supports the conclusion that in highly developed cities, shifts in economic structure lead to combustion sources emerging as the principal contributors to atmospheric NH₃ under both hazy and clear meteorological conditions. As illustrated in Fig. 5b, the proportion of δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ attributed to combustion and vehicular sources has increased over time. This temporal trend suggests that, with economic growth, agricultural and livestock emissions no longer represent the dominant contributors to atmospheric ammonia.

The extracted dataset was classified into four major emission categories-livestock farming, combustion, farmland fertilization, and vehicle exhaust-and subsequently subjected to statistical evaluation. As illustrated in Fig. 6, δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values associated with combustion sources showed strong consistency with previously reported ranges (Chang et al., 2021). Although traffic exhaust and livestock-related δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ values exhibited moderate dispersion, both sources remained within relatively well-defined isotopic ranges. In sharp contrast, δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ signatures following farmland fertilization displayed pronounced heterogeneity, covering nearly the entire isotopic spectrum reported for combustion, livestock, and vehicular emissions. This extensive variability highlights substantial regional differences in agricultural ammonia emission processes (Felix et al., 2014; Li et al., 2023b). Consequently, accurate source apportionment of atmospheric NH₃ requires distinguishing dominant local emission pathways rather than relying solely on generalized isotopic patterns (Chen et al., 2022; Zhang et al., 2023).

Figure 6 Statistical analysis of extracted data categorized by source: combustion sources, livestock and poultry farming sources, agricultural sources, and transportation exhaust sources. Boxes represent the interquartile range, the horizontal line within each box denotes the median value, and whiskers indicate the minimum and maximum values excluding outliers. Individual points outside the whiskers represent statistical outliers.

4 Summary

This study establishes high-precision δ¹⁵N signatures for ammonia emissions from three dominant intensive livestock systems in the Huang-Huai-Hai Plain. Distinct isotopic fingerprints were identified for dairy operations (−20.6 ‰ ± 0.8 ‰), laying hen facilities (−27.4 ‰ ± 1.0 ‰), and fattening pig farms (−38.4 ‰ ± 1.7 ‰), underscoring clear differences among livestock categories. Our results further demonstrate that isotopic signatures vary dynamically with NH₃ volatilization intensity, highlighting the need to incorporate volatilization-driven fractionation effects into isotope-based source apportionment frameworks. When compared with ambient δ¹⁵N-NH${}_{\mathrm{4}}^{+}$ measurements in Zhengzhou, the newly constrained source end-members indicate that non-agricultural sources-particularly vehicular emissions and combustion-are likely major contributors to urban atmospheric ammonia. This interpretation, however, requires validation through comprehensive isotopic mixing and dispersion modeling. Moreover, global-scale evaluation reveals that the exceptional variability of δ¹⁵N associated with fertilized soils continues to pose a substantial challenge for accurate identification of agricultural contributions. Collectively, the findings presented here provide critical isotopic constraints that can enhance regional atmospheric chemistry models and support the design of more precise and effective ammonia emission control policies.