Underestimated ammonia vehicular emissions in metropolitan city revealed by on-road mobile measurement

Atmospheric ammonia (NH3) plays a significant role in the nitrogen cycle, and can have impacts on air quality, ecological balance and climate change. While NH3 associated with natural and agricultural processes has long been considered the primary source, the contribution of combustion-related NH3, particularly from vehicular emissions, keeps on the rise. We found that high on-road NH3 concentrations occurred in a metropolitan city based on mobile measurement, and inferred that urban vehicular NH3 emission was likely underestimated in the past. NH3 emission factors (EFs) were derived from ring roads and tunnels, showing levels 74% and 20% higher than the latest standard proposed by Euro VII, respectively. To quantify the underestimation, two methods based on car ownership and traffic flow were used to estimate the annual vehicular NH3 emission in Shanghai as 2.59 and 1.76 Gg, respectively, substantially surpassing the predicted results by the Dynamic Projection model for Emissions in China. Given these discoveries, we recommend that it is urgent and imperative to establish relevant national standards and limits aiming at regulation on vehicular NH3 emissions. And more representative EFs measurements should be adopted to improve the accuracy of inventory estimation.


Introduction
Ammonia (NH 3 ) is a pivotal component within atmospheric nitrogen cycle.It is involved in the regulation of soil acidification processes, pH of water and ecosystem balance.As the primary alkaline gas in the atmosphere, NH 3 participates in the neutralization of sulfuric and nitric acid to form ammonium (NH 4 + ), which constitutes the significant component of fine particles (PM 2.5 ).During haze episodes, the mass of total ammoniated aerosol can comprise over 60% of PM 2.5 (Sun et al 2014, Pan et al 2016).Moreover, NH 3 plays a key role in the new particles formation by significantly increasing the nucleation rate of H 2 SO 4 -H 2 O system (Zhang et al 2012, Yao et al 2018), and further has impacts on human health, radiation balance and climate change.
The sources of NH 3 are primarily divided into two categories (Chen et al 2022).One is the NH 3 volatilized from NH 4 + -containing substances formed through natural pathway (release of livestock waste, soil and water) and agricultural activities (fertilizer application).Recent inventories suggested that this source accounted for approximately 80%-90% in China (Kang et al 2016, Xu et al 2016, Li et al 2021a).The other source is combustion-related emissions generated from vehicle exhausts, biomass and coal burning.Accurately estimating the contribution of this category, particularly with regards to vehicular emissions, presents challenges (Yu et al 2020).Combustion principle determines the difference of NH 3 in the vehicle exhausts.Light gasoline vehicles (LGVs) are typically equipped with threeway catalytic converters to mitigate the emissions of CO, NO x and HC.Nonetheless, NH 3 is produced as a by-product from this process.Heavy-duty diesel vehicles (HDVs) utilize the selective catalytic reduction system to control NO x emissions by adding urea [CO(NH 2 ) 2 ].However, excessive use or inappropriate temperature conditions of unstable urea will be accompanied by the emission of NH 3 .Nevertheless, the previous assessments of NH 3 emissions have typically regarded vehicular ammonia as insignificant when compared to fertilizer application or livestock waste emissions, with a proportion of at most Consequently, the measurements of EFs by the above methods are insufficient to accurately estimate the vehicular NH 3 emission of an entire city.Farren et al (2020) conducted on-road mobile measurements to calibrate the EFs, ultimately concluding that the inaccurate EFs lead to the underestimation of emissions.Recently, a new approach based on differential optical absorption spectroscopy (DOAS) technique has been applied for on-road mobile measurements of NH 3 (Dai et al 2023).
In this study, the on-road NH 3 concentration along different ring roads and tunnels were obtained by utilizing DOAS equipment during the mobile measurements in Shanghai, China.The observed high concentration suggests that estimates of ammonia emissions from vehicles were significantly underestimated in the past.Whereupon, the vehicular NH 3 EFs were subsequently calculated and compared with the latest proposal limits in Euro VII standard.Based on the on-road EFs of rings roads, tunnels and highways, we further estimated the annual vehicular NH 3 emissions of Shanghai so as to quantify the underestimation compared to previous studies and prediction.

Materials and methods
The mobile measurement was carried out in Shanghai between 27 October and 6 November 2022.Shanghai, as a representative of developed city, leads the country in vehicle ownership per capita and has stringent policies on vehicular emissions, which is meaningful to conduct on-road vehicular emissions research here for formulating pioneer policies.
The experimental vehicle was equipped with our newly designed ground-based remote sensing device based on DOAS principle (figure S1(a)).With this instrument, on-road vehicles emissions concentration of NH 3 and NO were measured.The design principle and further details have been described in previous work (Dai et al 2023).Besides, seven-wavelength aethalometer (AE33) was employed to measure black carbon (BC) at 880 nm, refer to Text S1 for calculation details.The GMP343 (VAISALA) was used to measure carbon dioxide (CO 2 ).The longitude, latitude and speed of the vehicle were determined using an onboard GPS module.The measurement route comprises four ring roads: a part of Inner Ring and Elevated Road in Puxi area (IR), Middle Ring (MR), Outer Ring (OR) and Suburban Ring (SR), as well as five tunnels, namely Tunnel of Yan'andong Rive, ShangZhong Road, JunGong Road, Outer Ring and Chongming (figure S1(b)).
Since most automotive motor fuels are discharged from the exhaust in the form of main CO 2 after combustion, the concentration ratio of NH 3 , NO (in ppbv) or BC (in µg m −3 ) to CO 2 (in ppmv) can be used as the emission ratio (ER) to characterize the component proportion in the plume.The ER could be described as follows: where P represents the concentration of NH 3 , NO or BC; ∆P and ∆CO 2 refer to the difference between the measured concentration and the ambient background concentration of the day.Due to the DOAS measurement principle, the background concentration of NH 3 and NO was eliminated during the spectral inversion process.As for BC and CO 2 , the 5% quantile concentration of the day was considered as background concentration in order to eliminate the day-to-day variation in background (Brantley et al 2014, Pu et al 2023).
In order to enable a direct comparison between fuel consumption and pollutant emission, emission indices (EIs) are defined as the mass (gram) of NH 3 , NO or BC emitted by per kilogram of fuel consumed, calculated as: Where M P is the molecular weight of NH 3 or NO; M CO2 and M C are the molecular weight of CO 2 and the atomic weight of C, respectively.ω C is the carbon weight fraction of gasoline (0.85) or diesel (0.87) (Kirchstetter et al 1999).For BC, ER × MP MCO 2 is substituted by ∆BC(µg×m −3 ) ∆CO2(µg×m −3 ) to calculate.To facilitate comparison with national standards limitation, the EF is further calculated: Here, ρ f is the density of gasoline (720-775 kg m −3 ) or diesel (810-850 kg m −3 ) (GB 17 930-2016 andGB 19 147-2016); Q s is average fuel consumption of cars per 100 km (Sullivan et al 2004, Othman et al 2022).
The COPERT and MOVES designed by different algorithm were utilized for estimating the vehicular ammonia emission.
The COPERT relies on different types of car ownership and annual average travel, expressed as follows: where i represents the target year for estimation; c denotes the vehicle subcategory; E i is the NH 3 emission in year i; O i,c is the ownership amount of subcategory c vehicles in year i; VKT c represents the annual average vehicle kilometers traveled for subcategory c; EF i,c is the NH 3 EF for subcategory c vehicles.
The MOVES estimates vehicular NH 3 emissions by considering the length of road segments and traffic flow on different road types.The method could be described: where i is the target month;

Pollutants concentration along the road
The experimental vehicle was driven multiple times along the ring roads, and the typical spatial distribution of pollutants was displayed in figure S2.
In the IR and MR, except for the tunnel areas, the concentrations of NH 3 and NO were both relatively stable, and at a similar level showing difference within 5 ppbv between each other.Occasionally, simultaneous high concentrations of three pollutants, NH 3 , NO and BC, exceeding 200 ppbv, 2000 ppbv, and 50 µg m −3 , respectively, were observed in the eastern and north-western regions of the SR, which can be attributed to the substantial traffic flow in corresponding segments of the ring roads.Detailed data statistics was summarized in table S1.
The distribution frequency of pollutant concentrations and vehicle speeds in four ring roads was depicted in figure 1.The measured NH 3 concentrations (averaged) followed the sequence of OR (27.4 ppbv) > IR (24.5 ppbv) > SR (21.3 ppbv) > MR (17.3 ppbv).While the sequence for NO was OR (272 ppbv) > SR (155 ppbv) > IR (29.0 ppbv) > MR (23.6 ppbv), same as BC.CO 2 was primarily influenced by the background concentration on each day.The ratio of NH 3 to NO can effectively represent vehicle types, reaching 1.19, 1.15, 0.20, 0.30 in four ring roads, respectively.Furthermore, as vehicles approached the city center, the average driving speed decreased due to heavier traffic flow.This may lead to higher emissions and greater differences in pollutants concentrations among the ring roads.In accordance with the motor vehicle restriction regulations of Shanghai, only LGVs are permitted to drive on the IR and MR, whereas HDVs and mixed-type dominate in the OR and the SR, respectively.NO and BC emissions are mainly generated through engine combustion, and higher proportion of HDVs in the OR and SR led to higher emissions compared to the IR and MR.
During the ring roads measurement, the experimental vehicle passed through several tunnels.To capture the change in pollutant concentrations from inlet to outlet of the tunnels, the time series data were resampled to nine segments, which facilitated the comparison between tunnels with different lengths (figure S3).The peak concentration of NH 3 in the tunnel of IR, MR, OR and SR were 212.6, 107.1, 250.8 and 126.0 ppbv, usually appearing in the 7th segment, which were 3.1, 12.3, 4.8 and 10.2 times that of their entrance, respectively.This stepwise increase of concentration towards the exit can be explained by the continuous accumulation of pollutants within the enclosed tunnel space under stable driving conditions.However, the accumulation degree of pollutants, i.e. the increment of concentration, existed difference in each tunnel.Traffic conditions and flow stability during peak hours in the MR tunnels brought a slower accumulation of pollutants (figures S3(c)-(f)).In comparison, road congestion of IR tunnel (figure S3(a)), complex vehicle types

Emission indexes of ring roads and tunnels
Figure 2(a) displayed the ERs of different ring roads calculated based on the concentration of CO 2 .The average ERs of NH 3 for IR, MR, OR and SR were 0.51 ± 0.33, 0.54 ± 0.27, 0.52 ± 0.21 and 0.63 ± 0.44 ppbv ppmv −1 , respectively.Developed cities broadly have lower ERs than developing cities (Sun et al 2017).In comparison, our results were similar to the measurements in Baoding and Shijiazhuang (Sun et al 2017), but higher than those reported in Beijing and significantly higher than those in the US and Sydney (Phillips et al 2019).The difference of ERs among developed cities highlighted the possibility of significant vehicular NH 3 emissions in Shanghai.Additionally, almost same level of ERs observed in IR, MR and OR implied that the contribution of HDVs to atmospheric ammonia cannot be neglected and may even be comparable to that of LGVs.Moreover, highways typically exhibit higher NH 3 ERs compared to urban-roads (Huang et al 2018).The ER measured in SR, known as the urban bypass highway, revealed a 20% increase compared to other ring roads, also indicating the impact of road type.The mean ERs of NO in the four ring roads were 0.63, 0.74, 4.19 and 4.20 ppbv ppmv −1 , respectively, while those for BC were 0.03, 0.03, 0.07 and 0.07 µg m −3 ppmv −1 .The huge difference between the IR/MR group and OR/SR group, exceeding 5 times for NO and 2 times for BC, could be largely attributed to the variations in fleet composition.
Corresponding EIs frequency distribution of pollutants is presented in figure S4.The NH 3 EI exhibited a proportion of 70% within 0.3-0.8g kg −1 , and the mean and median EI were 0.67 and 0.60 g kg −1 , respectively, which fell within the range of previous studies of Zhang et al (2021) and Bishop et al (2016).EIs are significantly influenced by different fuel formulation of vehicle (Suarez-Bertoa et al 2020) and engine operation conditions (Gentner et al 2017).For exploring the relationship between EIs and fleet composition, LGVs and HDVs from six road sections during the experiment were counted specifically and categorized into three scenarios: LGVs > HDVs, LGVs ≈ HDVs and LGVs < HDVs (figure 2(b)).With the decrease in the proportion of LGVs to HDVs, NH 3 EIs exhibited a significant downward trend, and the average EI dropped from 0.92 g kg −1 to 0.44 g kg −1 , with a decrease of more than 50%.Meanwhile, figure 2(c) illustrated the relationship between driving speed and NH 3 EI.The calculated EI at low-speed (<15 km h −1 ) was higher than that at moderate-speed (15-45 km h −1 ), and reached the same level as medium-high speed (45-60 km h −1 ).The stop-and-go traffic condition at low-speed led to a sharp surge in exhaust emissions.With the risen of driving speed, EI varied from 0.52 to 0.73 g kg −1 , representing an increase of about 40%.It further highlighted the dependency of emissions on engine conditions.
In contrast to the ring roads, the ERs of individual pollutant had a wider range in the tunnels (figure S5).The average ERs of NH 3 , NO and BC ranged from 0.26 to 0.46 ppbv ppmv −1 , 0.38 to 8.14 ppbv ppmv −1 and 0.02 to 0.12 µg m −3 ppmv −1 , respectively.The average NH 3 ERs in all measured tunnels were 0.35 ± 0.12 ppbv ppmv −1 .Overall, the ER of NH 3 in the tunnels showed a distribution of MR > OR > IR > SR, which were lower than those in the ring roads.Unlike the broader range of sampling areas covered by the ring roads, the tunnels' stable traffic flow at constant speed resulted in less variation amplitude of ERs, better embodying the emission characteristics of individual vehicles in close proximity to the sampling vehicle.Figure S6 provided the driving vehicle conditions and approximate distribution of vehicle types at the inlet, inside and outlet of the tunnel.The vehicle type distribution remained unchanged throughout the tunnel and was consistent with the ring roads where the tunnel was located.S2 for details.Our measured EFs in the roads and tunnels fall within the range of previous results, reaching 34.3 ± 20.5 mg km −1 and 22.5 ± 7.6 mg km −1 , respectively.The EFs in ring roads were more than 50% higher than those in tunnels.Additionally, a more substantial disparity, nearly double, was found between on-road measurements and lab-based theoretical appraisals.The act of long-distance driving on roads under different engine conditions and complex traffic conditions provides a more representative approach to measuring real-world emissions compared to short-travel tunnels or indoor laboratory environments.If only the EFs measured in tunnels or lab was used to estimation the emissions of regional NH 3 , it may be significant underestimated.flow on major roads was considered, the estimated vehicular NH 3 emissions might be incomplete and underestimated.

Implications to ammonia emissions
Currently, approximate 70% of global LDVs and HDVs sales are subject to either the Euro VI standard or a previous version, the rest may not even meet the Euro I standard (Mulholland et al 2022).The implementation of the standard brought effective control of on-road pollutant emissions.More recently, the Euro VII proposal has proposed even stricter standards for emission limitation, including new requirements for NH 3 emissions.Actually, in most cities, including Shanghai, the NH 3 EFs on roads and in tunnels were found to exceed the limits set by the new proposal standard (red dot line in figure 3(b)), with an exceedance by 74% and 20% in our work, respectively.Considering the ongoing urban development, average vehicular NH 3 emissions will rise continuously if no relevant policies are formulated to control EFs (figure S9).Once the new standard is implemented, annual NO x emissions could be reduced by 93% before the middle of this century (Mulholland et al 2022).Meanwhile, it is also expected to achieve the continuous decline of primary vehicular NH 3 emissions (Li et al 2020, Hopke and Querol 2022) and co-environmental benefits (Hartikainen et al 2023).The coordinated benefits in reduction of NH 3 and NO x emissions could even potentially double the effectiveness of PM 2.5 mitigation (Wang et al 2023).Therefore, it is imperative to include NH 3 in national vehicle-related emission standards, at least as a pilot in metropolitan cities such as Shanghai and Beijing.

Conclusions
A mobile on-road measurement based on DOAS was conducted in Shanghai, China, covering both the ring roads and tunnels.The high level of NH 3 suggested that current estimates of vehicular NH 3 emissions may be subject to underestimation.To accurately quantify the underestimation, ammonia emission indexes of different road types were derived.The results revealed that the EFs on roads were considerably higher than those measured in tunnels, reaching 34.3 ± 20.5 mg km −1 and 22.5 ± 7.6 mg km −1 , respectively.
Compared with the latest proposed Euro VII standards, the NH 3 EFs for roads and tunnels were found to be 74% and 20% higher, respectively.With the measured EFs, the COPRET and MOVES were used to estimate the annual vehicular NH 3 emissions as 2.59 Gg and 1.76 Gg, respectively.The results, notably higher than the predicted levels of DPEC, emphasized the substantial role of vehicular sources to urban NH 3 emissions.
Overall, this study underscored the importance of updating more representative vehicular NH 3 EFs in emission estimation, and informed latest on-road measurements aimed at estimating its impact on air quality and policy making regarding the limitation of vehicular NH 3 emissions.The refined measurement method of EFs in different road types should be adopted to other regions significantly affected by vehicular primary NH 3 and secondary aerosols emissions for improving the accuracy of inventory estimation, providing reliable support for subsequent predictions of national and even global ammonia emission trends.

Figure 1 .
Figure 1.Frequency distribution characteristics of pollutants concentration and vehicle speed in four ring roads.The subplots in (b) and (e) are the local zoomed in version of (b) and (e), respectively.

Figure 2 .
Figure 2. (a) Emission ratios of NH3, NO and BC to CO2 in four Ring Roads.(I: Inner Ring; M: Middle Ring; O: Outer Ring; S: Suburban Ring); On-road NH3 emission indices under (b) different proportions of light gasoline vehicles and heavy-duty diesel vehicles and (c) different driving speed ranges.

Figure 4
Figure 4 displayed the vehicular NH 3 emissions and their proportion in Shanghai based on reported inventories (An et al 2021, Li et al 2017) and previous studies (Chang et al 2016, Chang 2014, Huang et al 2012, Li et al 2021a, Wang et al 2023, Yu et al 2020).The proportion of transportation in NH 3 emissions (circle) estimated by Multi-resolution Emission Inventory model for Climate and air pollution research (MEIC) (http://meicmodel.org.cn)increased annually from 2013 to 2017.During this period, vehicular NH 3 emission in Shanghai rose by 21%, from 0.60 to 0.73 Gg yr −1 (Zheng et al 2014), and its proportion was less than 3%.Whereas, MEIC still had the significant inclination of underestimation.From 2006 to 2017, the reported vehicular NH 3 emissions based on COREPT in Shanghai (rhombus) ranged from 1.3 to 1.8 Gg and accounted for 3%-5% of total NH 3 emission assessed by respective study, presenting a stable level.By utilizing the estimation principles of COREPT and MOVES and measured EFs, the annual vehicular NH 3 emissions for Shanghai in 2021 (pentacle) were calculated as 2.59 Gg and 1.76 Gg, respectively, refer to Text S2 for calculation details.Average vehicular NH 3 proportion obtained by two methods could account for over 7% of the total NH 3 emission predicted by Dynamic Projection model for Emissions in China (DPEC) for the year 2020 (Tong et al 2020, Cheng et al 2021).Obviously, the result of COPERT were significantly higher than previous studies as well as the predictions by DPEC, indicating an underestimation of vehicular NH 3 emission in reported studies.Fixed and unrepresentative EFs measured by single location may be one of the reasons resulting in the underestimation.With the steady increase of civilian car ownership (figure S7), the vehicular NH 3 emission should have an upward trend and will almost rise by nearly 55% in 2030.As for MOVES, NH 3 emission from different road sections could be estimated based on traffic flow (figure S8).However, since only traffic

Figure 3 .
Figure 3. (a) Emission factors of NH3 measured in previous studies by method of road, tunnel and lab dynamometer.Open and solid circles represent the results of previous studies and this study, respectively.(b) Same as a but all studies were conducted in Shanghai.The red dotted line presents the emission limitation set by the Euro VII proposal.

Figure 4 .
Figure 4. Vehicular NH3 emissions and its proportion in Shanghai.The left side of the red dotted line reflects estimates from previous studies or inventories.The right side of the red dotted line corresponds to estimates from this study and predictions by MEIC-DEPC.
3% (Zhou et al 2015, Wu et al 2017, Wang et al 2018a, Yu et al 2020, Li et al 2021a).However, recent studies conducted at various scales, including urban (Chang et al 2016), regional (Hao and Song 2018), national (Li et al 2020), continental (Chen et al 2022), and even global (Meng et al 2017) levels, have highlighted the rising potential impact of vehicularemitted ammonia.Accurate estimation of ammonia bottom-up inventory, like computer programme to calculate emissions from road transport (COPERT) and Motor Vehicle Emission Simulator (MOVES), still relies on the measurement of emission factors (EFs, NH 3 emitted per unit mass of fuel).Previous studies have used various methods to determine EFs.For instance, chassis dynamometers for individual vehicles (Huang et al 2018), tunnel measurement (Liu et al 2014) and roadside remote sensing (Zhang et al 2021).However, the huge discrepancy (ten fold) in EFs reported in the same tunnel under similar traffic condition suggested that the tunnel measurement results still contain significant uncertainties (Liu et al 2014, Li et al 2021b).