Quantifying the Impact of Fertilizer-Induced Reactive Nitrogen Emissions on Surface Ozone Formation in China: Insights from FEST-C* and CMAQ Simulations

Abstract

The emissions of reactive nitrogen (Nr) from cropland links the pedosphere and atmosphere, playing a crucial role in the Earth’s nitrogen cycle while significantly impacting regional climate change, air quality, and human health. Among various Nr species, nitrogen oxide (NO) and nitrous acid (HONO) have garnered increasing attention as critical precursors to surface ozone (O₃) formation due to their participation in photochemical reactions. While most studies focus on Nr emissions from soils, the specific contributions of cropland Nr emissions considering planting activities to regional O₃ pollution remain insufficiently investigated. This study applied the enhanced process-based agroecological model (FEST-C*) coupled with the air quality (CMAQ) model to quantify cropland Nr emissions and assess their contributions to regional O₃ formation across China in June 2020. The simulated results indicated that the fertilizer-induced total Nr emission was estimated at 1.26 Tg in China, with NO emissions accounting for 0.66 Tg and HONO emissions for 0.60 Tg. North China was identified as a hotspot for cropland Nr emissions, contributing 43% of the national total. The peak emissions of cropland NO and HONO occurred in June, with emissions of 169 and 192 Gg, respectively. Cropland Nr emissions contributed approximately 8% to the national monthly mean MDA8 O₃ concentration, with localized enhancements exceeding 9% in agricultural hotspots in summer. North China experienced the largest MDA8 O₃ increase, reaching 11.71 μg m⁻³, primarily due to intensive fertilizer application and favorable climatic conditions. Conversely, reductions in nighttime hourly O₃ concentrations were observed in southern North China and northern Southeast China due to the rapid titration of O₃ via NO. In this study, the contributions of cropland Nr emissions to MDA8 O₃ concentrations across different regions of China have been further constrained. Incorporating cropland Nr emissions into the CMAQ model improved simulation accuracy and reduced mean biases in MDA8 O₃ predictions. This study offers a detailed quantification of the contribution of cropland Nr emissions in regional ozone formation across China and highlights the critical need to address cropland NO and HONO emissions in air quality management strategies.

1. Introduction

Surface O₃ is a major atmospheric pollutant that poses a considerable threat to the environmental, human health, and the climate [1,2,3,4]. It is mainly produced through sunlight-driven chemical reactions involving volatile organic compounds (VOCs) and nitrogen oxides (NO_x) [5]. These precursors are emitted from anthropogenic activities such as petroleum and natural gas extraction, the combustion of fossil fuels, chemical and industrial processes, solvent applications, residential boilers, and agricultural fertilization and pesticides [6,7,8,9], as well as biogenic sources like forests, vegetation, biomass burning, and soil bacteria [10,11,12,13].

Reactive nitrogen species (Nr, including NH₃, N₂O, NO_x, HONO, and N₂O₅) are key precursors to surface O₃ formation and play a significant role in atmospheric chemistry [14]. Soil is a major source of Nr emissions, predominantly emitting NH₃, N₂O, NO, and HONO [15]. NH₃ is the sole alkaline gas, and N₂O is an important greenhouse gas in the atmosphere, having received extensive attention in previous studies [16,17], while NO and HONO remain insufficiently investigated, especially from cropland [18,19,20,21]. Cropland NO and HONO emissions primarily stem from the biological processes in the soil [22] and are influenced by human activities like fertilization and irrigation and natural factors such as precipitation and temperature [23,24]. It is well established that soil emissions contribute to 12% to 20% of global NO_x emissions, with over 40% of this amount originating from agricultural sources [14]. In California, the United States, soil Nr emissions are a significant contributor to NO_x and O₃ pollution, contributing to a 176.5% increase in surface NO₂ concentrations and resulting in an additional 23.0% increase in surface O₃ levels [25,26]. Long-term field observations in China have also validated that soil Nr emissions accelerate in summer ozone increase and notably decrease the sensitivity of O₃ to anthropogenic emissions [27]. Unregulated soil Nr emissions impeded a NO₂ decrease in rural areas and offset the national efforts in China to mitigate NO₂ and O₃ levels by 20.9% and 15.4%, respectively, from 2011 to 2020 [28].

China is the world’s largest consumer of fertilizers [29], with fertilizer-induced Nr emissions exerting a notable influence on regional O₃ pollution formation and receiving widespread attention [20,30,31]. Due to the absence of complete temporal profiles for fertilizer-driven Nr emissions, the quantitative assessment of the impact of Nr emissions on regional O₃ pollution formation was impeded until 2021 [27,32]. Recent studies have increasingly utilized mechanistic representations of the biogeochemical processes, such as Yienger–Levy parameterization and Berkeley–Dalhousie Soil NO_x Parameterization (BDSNP) in models like GEOS-Chem, WRF-Chem, and Model of Emissions of Gases and Aerosols from Nature (MEGAN) to calculate soil NO_x emissions and assess their impacts on air quality in China [15,31,33,34,35]. The latest research has estimated soil NO_x and HONO emissions and evaluated the impact of soil Nr emissions on O₃ pollution, and temperature rises by integrating an updated soil Nr emissions scheme within the WRF-Chem model [36]. These studies have indicated that the increase in soil Nr emissions could significantly counteract the reductions in NO₂ and O₃ achieved by clean air policies. However, current methodologies treat NO and HONO as a part of the total soil available N without distinguishing between the fractions of soil N present as organic N, NH₄, or NO₃ [37]. Furthermore, these parametric schemes treat soil NO and HONO emissions as constant factors across diverse ecosystems based on global literature and field estimates [38,39]. These deficiencies resulted in significant uncertainties in NO and HONO emissions estimates and their contribution to regional O₃ formation. While current studies have explored the impact of soil NO and/or HONO on air quality, comprehensive research on cropland Nr emissions that fully account for anthropogenic management activities remains scarce.

To address this gap, we simulated cropland NO and HONO emissions across China in 2020 using FEST-C*. The FEST-C* model has been previously applied by Luo et al. (2022) [40] to simulate Nr emissions from fertilized soils over the continental United States. This work is the second to apply FEST-C* over China and the first to use FEST-C*-derived Nr emissions in air quality simulations. We applied the CMAQ model to quantify their impact on regional O₃ formation in June 2020. The objectives of this study were (1) to estimate the spatial and temporal distribution of cropland NO and HONO emissions in China, and (2) to assess their contributions to regional O₃ formation. The remainder of this paper is organized as follows: Section 2 describes the data and methodologies; Section 3 analyzes and discusses the cropland Nr emissions, model performance, the impact on regional O₃ formation, and limitations and uncertainties; and the conclusions are presented in Section 4.

2. Data and Methodology

Figure 1 presents the schematic diagram of this study. To simulate Nr emissions from cropland, we first adapted a process-based agroecological model (FEST-C*) for China. We processed land use, crops, soils, climate and meteorological data, and N input in China to drive the FEST-C* model. The model was then used to simulate NO and HONO emissions of China, which were subsequently incorporated as input for the CMAQ simulations.

2.1. Model Description and Configurations

FEST-C* is a process-based agroecological model developed by Luo et al. [40]. Unlike traditional schemes with empirical formulas such as the YL95 (Equation (1)) [41], BDSNP [42], and BDISNP [24] (Equations (2)–(5)), FEST-C* incorporates comprehensive biogeochemical processes of microbial nitrification and denitrification (Equations (6)–(14)) to constrain uncertainties in daily NO and HONO emissions from soil carbon, nitrogen, and phosphorus cycles on agricultural lands by accounting for atmospheric conditions and management practices [43].

$$F_N\left({NO}_{YL}\right)=f{\displaystyle \frac{w}{d}}\left(A_{biome({\displaystyle \frac{w}{d}})}, Tsoil\right)\times P(precipitation)\times CRF(LAI, SAI)$$

(1)

where $f{\displaystyle \frac{w}{d}}\left(A_{biome({\displaystyle \frac{w}{d}})}, Tsoil\right)$ represents the emissions factor, which depend on whether the soil is wet) or dry. The wet factor is appliedwhen rainfall exceeds 1 cm within the preceding two weeks. The factor is further influenced by temperature ($Tsoil$).

$$F_N\left({NO}_{BDSNP}\right)=A_{biome}^{\prime }\left(N_{avail}\right)\times f\left(T\right)\times g\left(\theta \right)\times P\left(l_{dry}\right)\times \left(1-CRF\right)$$

(2)

where $N_{avail}$ represents the fertilizer N, the standing pool N, and deposited N; $A_{biome}^{\prime }$ coefficients are the functions of $N_{avail}$; $f\left(T\right)$ and $g\left(\theta \right)$ are the temperature and soil moisture dependencies, respectively, where $\theta$ is water-filled pore space; $P\left(l_{dry}\right)$ is the pulsing factor depending on dry spell length; and $CRF$ is the canopy reduction factor.

$$F_N\left(HONO\right)=F_{N,opt}\left(HONO\right)\times f\left(T\right)\times f\left(SWC\right)$$

(3)

$$f\left(T\right)=e^{{\displaystyle \frac{E_a}{R}}\left({\displaystyle \frac{T}{T_{opt}}}-{\displaystyle \frac{1}{T}}\right)}$$

(4)

$$f\left(SWC\right)=1.04\times {exp}^{\left({-e}^{{\displaystyle \frac{SWC-11.32586}{5.27335}}}-{\displaystyle \frac{SWC-11.32586}{5.27335}}+1\right)}$$

(5)

where $F_{N,opt}\left(HONO\right)$ represents the optimal flux of soil HONO; the factors of $f\left(T\right)$ and $f\left(SWC\right)$ account for the influence of soil temperature and water content, respectively; $E_a$ denotes the activation energy of HONO; $R$ is the universal gas constant; and $T_{opt}$ is the temperature at which the optimal flux is emitted.

$$F_{NO}=F_{{nt}_{NO}}+F_{{dt}_{NO}}$$

(6)

$$\begin{array}{c}{F}_{n{t}_{NO}}=F_{n{t}_{N_{2}O}}\ast r_{{\displaystyle \frac{NO}{N_{2}O}}}\left(soil porosity,soil moisture,soil texture\right)\ast P\left(length of dry period\right) \ast \\ CRF\left(LAI,meterology\left({\displaystyle \genfrac{}{}{0pt}{}{wind speed, surface pressure,surface air temperature,}{ surface radiation,snow cover,cloud fraction}}\right)\right) \ast \\ \left(1-f_{r,\mathit{max}\left(HONO\right)}\ast f_{SWC}\right)\end{array}$$

(7)

$$F_{d{t}_{NO}}=F_{d{t}_{N_{2}O}}\ast r_{{\displaystyle \frac{NO}{N_{2}O}}}\left(soil porosity, soil moisture, soil texture\right)$$

(8)

where $F_{n{t}_{NO}}$, $F_{d{t}_{NO}}$, and $F_{d{t}_{N_{2}O}}$ are NO emission flux from nitrification and denitrification and N₂O emission flux from the denitrification process, respectively. $P\left(length of dry period\right)$ is the pulsing effect of NO emission triggered by re-wetting after the drying period. $CRF$ is the canopy reduction, which is mainly affected by leaf area index and meteorological factors.

$$F_{N_{2}O}=F_{{nt}_{N_{2}O}}+F_{{dt}_{N_{2}O}}$$

(9)

$$F_{{nt}_{N_{2}O}}=F_{nit}\ast 2\%$$

(10)

$$F_{{dt}_{N_{2}O}}=R_{den,max}\ast WD\ast f_{N{O_3}^{-}}\ast f_{respiration}\ast f_{moisture}\ast { f}_{r{N}_{2}O} \left(f_{N{O_3}^{-}, }{f}_{respiration},f_{moisture}\right)$$

(11)

$$F_{nit}=\left[ {WNH}_3\ast {\displaystyle \frac{\left(1-e^{-\left(k_{vol}+k_{nit}\right)dt}\right)}{dt}}\right]\ast \left(1-f_{vol}\right)$$

(12)

where $F_{N_{2}O}$ represents the total N₂O emissions from microbial nitrification and denitrification ($\mathrm{k}\mathrm{g} \mathrm{N} {\mathrm{h}\mathrm{a}}^{-1} {\mathrm{d}}^{-1}$). It is expressed as a function of the empirical maximum denitrification rate ($R_{den,max }$), soil nitrate ($f_{N{O_3}^{-}}$), soil respiration ($f_{respiration}$), and soil moisture ($f_{moisture}$).

$$F_{{nt}_{HONO}}=F_{n{t}_{NO}}\ast f_{r,\mathit{max}\left(HONO\right)}\ast f_{SWC}$$

(13)

$$f_{SWC}=\left\{\begin{array}{c}{\displaystyle \frac{WFPS}{0.1}}, WFPS\le 0.10\\ {\displaystyle \frac{0.4-WFPS}{0.4-0.1}}, 0.1<WFPS\le 0.40\\ 0, WFPS>0.4\end{array}\right.$$

(14)

where $F_{{nt}_{HONO}}$ and $F_{n{t}_{NO}}$ are the emission fluxes of HONO and NO from nitrification, respectively; $f_{r,\mathit{max}\left(HONO\right)}$ is the maximum value of HONO in the soil nitrification process, which is 0.429 [11]; and $f_{SWC}$ is the soil water content factor, which reflects the effect of soil pore water content on HONO emissions.

FEST-C* integrates the EPIC, WRF, and CMAQ model with land use and agricultural management, inducing an approach for evaluating the effects of fertilizer-induced Nr emissions on air quality. FEST-C* outputs daily Nr emissions from rain-fed and irrigated crops. Detailed model framework and formulations of Nr calculations are provided by Luo et al. [40].

WRFv4.3 (https://www.mmm.ucar.edu/wrf-model-general, last access: 10 November 2024) and CMAQv5.3.3 (https://www.epa.gov/cmaq, last access: 10 November 2024) were used to simulate the meteorology and atmospheric ozone concentrations across China (Figure 2). The simulation domain encompassed the entirety of China (62°~143° E, 14°~58° N) with a grid resolution of 27 km, utilizing a Lambert conformal conic projection centered at 37° N and 103° E. These simulations were conducted with a horizontal resolution of 27 km with 30 vertical levels.

Table 1. Model configurations and inputs for the WRF-CMAQ model.

Categories	Description	Schemes and Data
Domain	Grid numbers	190 (x) × 170 (y)
	Horizontal resolution	27 km
	Vertical layers	30 levels
Physics parameterization	Shortwave radiation	RRTMG [44]
	Longwave radiation	RRTMG [44]
	Cloud microphysics	Morrison [45]
	Planetary boundary layer	ACM2 [46]
	Cumulus	Kain–Fritsch [47]
	Surface	Pleim–Xiu [48]
	Land surface	Pleim–Xiu LSM [48]
	ICloud	Xu–Randall [49]
Chemistry scheme	Aerosol mechanism	AERO6
	Gas-phase chemistry	CB6
Emissions	Anthropogenic emission	MEIC (2020) [52], EDGAR (2018) [53]
	Dust emission	Foroutan
	Sea-salt emission	Gong
Input data	Meteorological ICs and BCs	FNL
	Chemical ICs and BCs	WACCM

lists the parameterization schemes and data for WRF and CMAQ simulations. The Rapid Radiative Transfer Model for General Circulation Models (RRTMG) for both longwave and shortwave radiation schemes [44], the Morrison scheme for the cloud microphysics scheme [45], the Asymmetric Convective Model (ACM2) for the planetary boundary layer scheme [46], the Kain–Fritsch scheme for weather prediction models to simulate deep convection and precipitation processes [47], the Pleim–Xiu Land Surface Model for the surface scheme and the land surface process scheme [48], and the Xu–Randall method for convective processes [49] were selected for the WRF model, respectively. The aerosol mechanism and aqueous-phase chemistry used in CMAQ were the AERO6 and CB06, respectively. Two scenarios with and without considering cropland Nr emissions were conducted in June 2020 to explore the impact of cropland Nr emissions on surface O₃ formation in China. We chose June as that is when intensive fertilizer-induced Nr emissions and severe ozone pollution have been observed [31,50]. In this study, we evaluated the impact of cropland Nr emissions on O₃ formation at both the national and regional levels across China. The six regions (Figure 2) represent diverse socioeconomic, geo-climatic conditions, and various crop types and rotation practices [51]. These regions include Northeast China (NE), Northwest China (NW), North China (NC), Southeast China (SE), Southwest China (SW), and the Tibetan Plateau (TP).

2.2. Data Sources

The input data of the FEST-C* model include land use, crops, soils, climate and meteorological information, and N input. Land use data from the China Land Cover Dataset [54] (http://doi.org/10.5281/zenodo.4417810, last access: 9 November 2024) were utilized in this study. The data provide 10 landcover categories in which the “Cropland” category was the input for the BELD4 Data Generation tool in FEST-C* [55]. Crop-related data, including crop types, cropping intensity, and irrigated/rain-fed crops, were from the latest publicly available data. The Harmonized World Soil Database version 2.0 [56] (https://gaez.fao.org/pages/hwsd, last access: 11 November 2024) was used to provide soil data at 1 km resolution. The climate dataset used in the model spin-up stage was from the National Climatic Data Center, National Oceanic and Atmospheric Administration (ftp://ftp.ncdc.noaa.gov/pub/data/noaa/isd-lite/, last access: 12 November 2024). This dataset provides historical climate data, while daily meteorological data used in the year-specific simulation were from METCRO2D files generated by MCIP using WRF outputs. N input data were from China Statistical Yearbooks, China Rural Statistical Yearbooks, and fertilizer application recommendations.

Anthropogenic emissions were sourced from the Multi-resolution Emission Inventory model for Climate and air pollution research (MEIC, http://meicmodel.org.cn/?page_id=560, last access: 10 October 2024) for China in 2020 and the Emissions Database for Global Atmospheric Research (EDGAR, https://edgar.jrc.ec.europa.eu/, last access: 10 November 2024) for regions outside China in 2018. These databases provide pollutant emissions from various sectors, including industrial, energy system, transportation, residential, and agricultural sources with spatial resolutions of 0.25° and 0.1°, respectively. Cropland Nr emissions were simulated in China during 2020 using FEST-C* as mentioned in Section 2.1. VOCs emissions from biogenic sources and biogenic NO_x emissions excluding cropland were calculated using MEGAN v3.0, (https://bai.ess.uci.edu/megan/, last access: 11 November 2024). Open biomass burning emissions are adopted from the Fire INventory from NCAR version (FINN, version 1.5, https://www2.acom.ucar.edu/modeling/finn-fire-inventory-ncar, last access: 11 November 2024). The allocation of emission data was performed using linear interpolation. These anthropogenic emission data were matched to the simulation domain, intercepted to calculate area ratios, and weighted to allocate emissions for each model grid [57].

The meteorological ICs and lateral BCs for this study were obtained from the National Center for Environmental Prediction Final Analysis datasets (http://rda.ucar.edu/datasets/ds083.2, last access: 11 November 2024). These data have a horizontal resolution of 1° × 1° and an interval of 6 h. Chemical ICs/lateral BCs for CMAQ simulations were downscaled from the Whole Atmosphere Community Climate Model (https://data.rda.ucar.edu/, last access: 11 November 2024) by using the mozart2camx (https://www.camx.com/download/support-software/, last access: 21 November 2024) and mozbc (https://www.acom.ucar.edu/wrf-chem/download.shtml, last access: 21 November 2024) tools, respectively.

2.3. Ground-Based Observation

The China National Environmental Monitoring Center (CNEMC, https://quotsoft.net/air, last access: 20 November 2024) provided hourly O₃ concentration to evaluate the model performance in simulating surface O₃. The maximum daily 8 h averaged ozone (MDA8 O₃) concentrations were calculated using Pythonv3.12.2.

2.4. Data Analysis

In this study, statistical metrics including Pearson correlation coefficient (R), mean bias (MB), root mean squared error (RMSE), normalized mean bias (NMB), and normalized mean error (NME) were selected to evaluate hourly and MDA8 O₃ concentrations across different regions of China.

The BASE and InNr scenarios were established to quantify the contribution of cropland Nr emissions on O₃ concentrations. The BASE scenario only included anthropogenic emissions except for croplands, while InNr scenario included NO and HONO emissions from croplands as estimated by FEST-C*. We calculated the difference of O₃ concentrations between the InNr and BASE scenarios (InNr–BASE). These differences indicate the impact of cropland Nr emissions on surface ozone. Also, we further quantified the promotion or inhibition of surface O₃ formation due to fertilizer-induced Nr emissions to the surface O₃ concentration by calculating the ratio of the difference to the O₃ concentration simulated under the BASE scenario.

3. Results and Discussion

3.1. Cropland Nr Emissions in China

Total fertilizer-induced Nr emissions simulated by FEST-C* in China during 2020 were 1.26 Tg, with 43% (0.54 Tg) of them emitted in the NC region, followed by SE (23%), SW (13%), NE (10%), NW (7%), and TP (4%) (

Table 2. Cropland Nr emissions (Tg) across different regions of China during 2020.

Region	NO	HONO	Nr
Northeast China (NE)	0.06	0.07	0.13
Northwest China (NW)	0.06	0.03	0.09
North China (NC)	0.24	0.30	0.54
Southeast China (SE)	0.19	0.10	0.29
Southwest China (SW)	0.09	0.07	0.16
Tibetan Plateau (TP)	0.02	0.03	0.05
Total China	0.66	0.60	1.26

). NO and HONO emissions were comparable across most regions, including NE, NW, SW, and TP, ranging from 0.02 to 0.09 Tg. However, NO emissions in SE (0.19 Tg) exceeded those in NC (0.10 Tg), whereas HONO emissions in SE (0.24 Tg) were lower than in NC (0.30 Tg).

The spatial distribution of cropland Nr emissions varied significantly across different regions of China (Figure 3). The eastern NC region (>7 kg/ha) and northern SE region (>5 kg/ha) emerged as hotspots for Nr emissions, followed by certain areas in NE, NW, and SW. In eastern NC, estimated NO (Figure 3a) and HONO (Figure 3b) emissions reached 0.21 and 0.26 Tg yr⁻¹, accounting for 35% and 48% of the national total emissions, respectively. In contrast, Nr emissions were minimal in regions such as most TP, the western NW, and the northwestern SW. These patterns of Nr emissions were predominantly attributable to N fertilizer application in cropland. In 2020, China applied a total of 25.89 Tg N in N fertilizers, with over 65% concentrated in major grain-producing regions, including the North China Plain (NCP) (30.2%), northern SE (20.4%), and western NE (15.6%) (Figure 3c). The high fertilizer consumption in these regions contributed significantly to their elevated Nr emissions. Interestingly, specific areas such as the central and eastern NCP exhibited NO and HONO emissions comparable to the central part of the region, with values exceeding 5 kg/ha. However, the fertilization intensity in these areas was notably lower than in the central NCP. This suggested that Nr emissions in these regions were influenced by additional factors, which will be discussed further in this section.

In 2020, cropland Nr emissions in China exhibited a bimodal temporal pattern, with notable peaks in June (361 Gg) and October (280 Gg) (Figure 4). Cropland NO emissions nearly tripled from 71 Gg to 192 Gg during January and June before sharply declining to 80 Gg in September. November and December recorded the lowest NO emissions, contributing just 3% and 2% of the annual total NO emissions. The emission trends in cropland HONO were consistent with those of NO. These variations were largely attributed to the crop rotation systems in China and corresponding FC in croplands. For example, high FC levels in May in NE China, June and October in NC, and parts of SE China led to high NO and HONO emissions of 21, 48, 37 Gg, and 7, 65, and 49 Gg in the corresponding periods, respectively. An increase of 2.1 Tg in FC from February to March resulted in NO and HONO emissions increasing by 77% and 145%, respectively. Similar fertilizer-induced NO and HONO emissions trends were observed from April to May. These emission trends have also been confirmed in previous studies [20,32,58].

However, except for FC, cropland NO and HONO emissions were also significantly affected by other factors such as soil moisture and temperature. For instance, despite a 3% increase in FC from January to February, NO and HONO emissions decreased by 16% and 62%, respectively. This reduction was attributed to the suppression nitrification by soil ammonia-oxidizing archaea and soil ammonia-oxidizing bacteria, which was driven by increased soil moisture under a low (11 °C) soil temperature [59,60,61]. Moreover, although the peak FC in May (5.4 Tg N) exceeded that in June (4.4 Tg N), NO and HONO emissions in May were 36% and 38% lower. Similarly, while FC decreased by over 88% from June to August, NO and HONO emissions declined by only 48% and 76%. These findings indicated that NO and HONO emissions during non-fertilization periods were more dominated by soil temperature and moisture than by FC. Although May had the highest fertilizer application during seeding (Figure 4), cropland Nr emissions were not higher than those in June because of the effects of other factors except for fertilization, such as soil temperature. Huang et al. [31] indicated that cropland Nr emissions in June were in North China due to fertilization before the pre-planting period, affecting regional O₃ formation. Therefore, we chose to quantify the contribution of cropland Nr emissions to O₃ concentration this month in the following section.

3.2. Evaluation of CMAQ Model Performance

Figure 5 presents the spatial distribution of monthly average O₃ and MDA8 O₃ concentrations from CNEMC observations and simulations under BASE and InNr scenarios for June 2020 in China to evaluate the model performance. Both simulations captured the hotspots of hourly O₃ and MDA8 O₃ in the NCP region and Sichuan Basin, consistent with observed data and indicative of extensive agricultural activities.

Under the BASE scenario, simulations significantly underestimated hourly O₃ and MDA8 O₃ concentrations by 12.10 μg m⁻³ (NMB: −9.84%) and 10.81 μg m⁻³ (NMB: −9.57%), respectively, with the central NCP showing notable underestimations of 40 μg m⁻³ for hourly O₃ and 20 μg m⁻³ for MDA8 O₃. Incorporating cropland Nr emissions (InNr scenario) improved the model performance in simulating hourly O₃ and MDA8 O₃, reducing MB to −8.00 and −2.34 μg m⁻³. The model performed optimally in the NCP region under the InNr scenario, with a negative mean bias of 20 μg m⁻³ for MDA8 O₃ in June 2020. Conversely, the performance was poorer in regions such as southeastern SW, southern SE, and most of the TP.

This study found that cropland Nr emissions significantly impacted regional O₃ formation in China during June 2020, primarily concentrated in the NC region and some parts of SW and NE, as described in Section 3.1. To further assess the influence of cropland Nr emissions on O₃ formation, we selected agriculture-dominated cities in these regions, such as Suihua, Songyuan, Jilin, and Siping in the NE region, Baoding, Cangzhou, Tangshan, and Handan in the NC region, and Mianyang, Ziyang, Zigong, and Yibin in the SW region, and analyzed the hourly variation in simulated and observed O₃ and MDA8 O₃ under the BASE and InNr scenarios (Figure 6).

Our results indicated that the variations in simulated hourly O₃ and MDA8 O₃ during this period were consistent with the corresponding observed hourly O₃ and MDA8 O₃ variations in the four cities, with correlation coefficients exceeding 0.68. In the Baoding, Cangzhou, and Tangshan of the NC region, similar hourly O₃ and MDA8 O₃ variations were found, and peak concentrations were observed on 7 June, 13 June, and 16 June, respectively. During these periods, the hourly O₃ concentrations exceeded 250 μg m⁻³, which was 25% higher than the national Level II standard (200 μg m⁻³, https://www.mee.gov.cn/ywgz/fgbz/bz/bzfb/, last access: 26 November 2024). Compared with the BASE scenario, the monthly O₃ and MDA8 O₃ concentrations under the InNr scenario in these cities increased by approximately 4 and 7 μg m⁻³, respectively, underscoring the influence of fertilizer-induced emissions on O₃ formation. However, the simulated hourly O₃ and MDA8 O₃ concentrations exhibited a normalized mean bias ranging from −29% to −40% across the three cities. These underestimations may arise from simplifying complex chemical mechanisms, physical processes, and input data uncertainties, such as emission inventories. In Mianyang, Sichuan Basin, the average hourly O₃ and MDA8 O₃ concentrations in June were 83.7 and 84.7 μg m⁻³, respectively. The simulated average hourly O₃ and MDA8 O₃ concentrations under the InNr scenario were 85.1 μg m⁻³ and 88.7 μg m⁻³, 5% and 9% higher than the those under the BASE scenario.

3.3. Impact of Cropland Nr Emissions on Regional Ozone Formation

To quantify the impact of cropland NO and HONO emissions on O₃ formation, the spatial distribution of the differences between simulated hourly O₃ and MDA8 O₃ was depicted in Figure 7. Elevated ∆MDA8 O₃ (>10 μg m⁻³) (Figure 7b) concentrations were predominantly observed in most of the NC region, in parts of the NW and SW regions, and in northern SE, corresponding closely to regions with high cropland Nr emissions. These findings were also confirmed by Sha et al. [36], who showed that soil Nr emissions significantly influence surface O₃ in the FWP and Beijing–Tianjin–Hebei (BTH) region. In the northwestern part of the NCP, cropland Nr emissions increased MDA8 O₃ concentrations by 6–14 μg m⁻³, respectively, which were 3–5 times higher than those in non-agricultural areas. This notable increase can be attributed to the intensive fertilization preceding the crop planting stage, which releases substantial Nr emissions. Wang et al. [34] reported that cropland NO_x emissions contributed 27.7% of surface NO₂ and elevated MDA8 O₃ by an average of 8 μg m⁻³ in the NCP region. Two critical factors led to North China becoming the region where cropland NO and HONO emissions have the most significant impact on O₃ formation. Previous studies have shown that cropland NO and HONO emissions were concentrated in North China, contributing more than 40% and 60% of the total NO and HONO emissions [62]. North China is a hotspot for NO_x and VOCs emissions from both industrial activities and motor vehicles. After fertilization, the rate of ·OH generation from HONO photolysis significantly exceeds the ·OH generation rate from O₃ photolysis, with the maximum ratio reaching 4.7 [63]. This accelerates the conversion of NO_x and VOCs into O₃, leading to a more significant impact of cropland NO and HONO emissions on O₃ formation in North China compared to other regions. Notably, while cropland Nr emissions increased hourly O₃ concentrations by 2 μg m⁻³ to 6 μg m⁻³ across most areas of China, a significant inhibitory effect was observed in the southern NC and the northern SE (Figure 7a). To investigate this phenomenon, we analyzed the spatial distribution of average hourly O₃ concentrations during daytime and nighttime and found that the inhibitory effect occurred primarily at night (Figure 7c,d). This inhibitory effect was also confirmed by Shen et al. [33] and Chen et al. [64], who showed that the soil NO emissions increased the atmospheric NO_x concentrations and caused the O₃ concentration to decrease by 2 μg m⁻³ to 10 μg m⁻³ in eastern China. Song et al. [65] and Niu et al. [66] suggested that the nocturnal titration of surface O₃ by NO, leading to the formation of nitrate radicals (NO₃), played a key role in reducing O₃ concentrations. Additionally, HONO can undergo conversion into NO through non-photolytic pathways, further increasing atmospheric NO levels and accelerating the reaction of O₃ with NO (Figure 8).

To further investigate ozone responses to cropland Nr emissions at the region level, we analyzed monthly simulated hourly O₃ and MDA8 O₃ concentrations and the contribution of cropland Nr emissions to the surface O₃ concentration across different regions of China in June 2020 (

Table 3. Monthly average concentrations (µg m⁻³) of hourly O₃ and MDA8 O₃ with and without cropland Nr emissions and the contribution (%) of cropland Nr emissions to hourly O₃ and MDA8 O₃ across different regions of China in June 2020.

Region	Hourly O3				MDA8 O3
	BASE	InNr	InNr-BASE	(InNr-BASE)/BASE	BASE	InNr	InNr-BASE	(InNr-BASE)/BASE
NE	65.66	68.05	2.39	3.65	91.00	97.71	6.71	7.37
NW	78.39	81.74	3.34	4.27	94.31	102.36	8.05	8.54
NC	89.03	90.78	1.76	1.97	127.41	139.12	11.71	9.19
SE	65.64	66.83	1.19	1.81	92.31	99.45	7.14	7.73
SW	70.62	73.21	2.59	3.66	86.90	93.71	6.81	7.84
TP	86.95	89.98	3.03	3.49	95.40	102.20	6.80	7.13
Total China	77.00	79.63	2.63	3.42	97.89	105.76	7.87	7.97

). Under the BASE scenario, hourly O₃ concentrations ranged from 65.66 μg m⁻³ to 89.03 μg m⁻³. The NC and TP regions exhibited the highest O₃ concentrations, with monthly concentrations of 89.03 μg m⁻³ and 86.95 μg m⁻³, respectively, followed by NW (78.39 μg m⁻³), SW (70.62 μg m⁻³), NE (65.66 μg m⁻³), and SE (65.64 μg m⁻³). Similar regional patterns were observed for MDA8 O₃ concentrations. The NC region recorded the maximum MDA8 O₃ concentration of 127.41 μg m⁻³, representing a 38.4% increase over hourly O₃ levels.

Under the InNr scenario, national average hourly O₃ and MDA8 O₃ concentrations in June increased by 2.63 µg m⁻³ and 7.87 µg m⁻³, respectively, compared to the BASE scenario, highlighting the positive impact of cropland Nr emissions on surface ozone formation. Regionally, NC and NW showed the highest enhancements in MDA8 O₃ concentrations, with increases of 11.71 µg m⁻³ and 8.05 µg m⁻³, exceeding the national averages of 48% and 3%, respectively. In contrast, the SE, SW, NE, and TP regions displayed smaller increases, with MDA8 O₃ remaining below 7.20 µg m⁻³. Regarding regional variation in hourly O₃ concentrations, the largest increases were in the NW and TP regions, with enhancements of 3.34 µg m⁻³ and 3.03 µg m⁻³, respectively, followed by SW (2.59 µg m⁻³), NE (2.39 µg m⁻³), NC (1.76 µg m⁻³), and SE (1.19 µg m⁻³). These regional differences can be attributed primarily to variations in fertilization intensity. Sha et al. [36] reported that Nr emissions from fertilization during the planting season contributed to a 16% increase in MDA8 O₃ concentrations in the BTH region. Notably, the regional increments in hourly O₃ concentrations due to cropland Nr emissions differed significantly from those of MDA8 O₃. For instance, while the NC region exhibited the largest increase in MDA8 O₃, its hourly O₃ increment was lower compared to other regions, such as NW, TP, SW, and NE. This discrepancy can be explained by the inhibitory effect of cropland Nr emissions on hourly O₃ production in the NC region, where NO rapidly titrates O₃ at night. In contrast, MDA8 O₃, which reflects the maximum daily 8 h average ozone concentration, is more relevant to human health impacts and less affected by nocturnal titration processes.

3.4. Comparison with Previous Studies

The magnitudes of cropland NO and HONO emissions in this study were compared with estimates from previous studies employing bottom-up, top-down, or data-driven approaches [67,68,69,70,71,72,73,74,75,76]. The NO emission estimates estimated in our study were 0.66 Tg, falling within the range of 0.36–1.09 Tg reported by earlier research. However, simulated HONO emissions were approximately three times lower than those reported by Wu et al. [67]. This discrepancy likely arises from earlier statistical models lacking the detailed parameterization of soil temperature effects on HONO emissions, introducing greater uncertainty into HONO estimates.

A limited number of studies have been conducted to quantify the contribution of fertilizer-induced Nr emissions to regional O₃ pollution, particularly in cropland-dominated areas where such emissions are closely linked to fertilization practices (

Table 4. Comparison of the contributions of soil and cropland Nr emissions to MDA8 O₃ across different regions of China in this study and previous studies.

Species	Sources	Methods	Regions	Simulation Periods	Contributions to MDA8 O3 Enhancement	References
NO	Soil	BDSNP + CAMx	China	June 2018	5.7%	[31]
		YL95 + WRF-Chem	NW, NC, NE, SW	July 2017	32.5%, 38.3%, 41.9%, 31.2%	[33]
	Cropland	BDSNP + GEOS-Chem	NCP	July 2017	3.7%	[27]
		BDSNP + WRF-Chem	NCP, BTH	24 May−7 June 2020	5.4%, 4.2%	[34]
Nr (NO + HONO)	Soil	BDISNP + GEOS-Chem	China, NCP	June−July 2019	9.7%, 10.1%	[15]
		BDISNP + WRF-Chem	BTH, FWP	July 2018	16.9%, 17.2%	[36]
	Cropland	FEST-C* + CMAQ	China, NC	June 2020	8.0%, 9.2%	This study

). Existing studies have assessed the impact of NO and/or HONO emissions from agricultural soils on O₃ formation. For instance, Lu et al. (2021) [27] utilized a mechanistic parameterization of soil NO emissions with multiple air quality models and found that cropland NO emissions contributed 3.7% to MDA8 O₃ concentrations in the NCP. Similarly, Huang et al. [31] used the BFM and OSAT model to demonstrate that soil NO emissions contributed 5.7% to MDA8 O₃ concentrations in June. However, Shen et al. [33] demonstrated that omitting soil NO emissions could lead to a reduction of more than 30% in O₃ concentrations across various regions of China, which was 8 and 5 times higher than the estimates of Lu et al. [27] and Huang et al. [31], respectively. This discrepancy primarily arises from Shen et al. [33] underestimating the simulated O₃ concentrations (50 µg m⁻³) in July 2017 under the BASE scenario compared to the O₃ concentrations obtained by other studies (130 µg m⁻³).

Except for the single effect of NO to O₃ formation, Sha et al. [36] assessed the combined influence of Nr (including NO and HONO) on O₃ pollution. Their findings indicated that soil Nr emissions increased MDA8 O₃ by 17.2% and 16.9% in the FWP and BTH regions, respectively, surpassing the effects of NO or HONO alone. In this study, cropland Nr emissions were found to contribute 8.0% and 9.2% to MDA8 O₃ in China and NC, respectively, which were comparable to the findings of Tan et al. [15], who reported that soil Nr emissions increased MDA8 O₃ by 10.1% and 9.7% in China and the NCP, respectively. However, these estimates were significantly lower than the 16.9% reported by Sha et al. [36]. These inconsistencies can be attributed to the lower cropland Nr emissions applied in our study due to a 12% reduction in nitrogen fertilizer inputs from 2018 to 2020 (https://www.stats.gov.cn/sj/ndsj/, last access: 28 November 2024). Variations in model parameterization schemes may also lead to discrepancies among studies. Therefore, to improve the accuracy of simulated cropland Nr emissions and their impact on O₃ formation, there is an urgent need for the continuous and dynamic observations of Nr fluxes in agricultural areas to better validate and refine model outputs.

3.5. Limitations and Uncertainties

Although the major findings of this study highlight the contribution of cropland Nr emissions to surface ozone concentrations across different regions of China, substantial uncertainties remain in the current model estimates of cropland HONO emissions. For instance, the formation of cropland HONO is a complex process involving both abiotic and biotic pathways and is influenced by a multitude of factors such as soil type, soil acidity, soil temperature, and microbial activity. However, the FEST-C* model adopts simplified representations of these intricate processes. Specifically, HONO emissions are estimated based on NO emissions and adjusted using a water-filled pore space (WFPS) function. This function assumes a linear increase in HONO emissions when WFPS ≤ 0.10 and a linear decrease when 0.10 < WFPS ≤ 0.40 [40]. This approach fails to account for the nuanced interactions between soil moisture and temperature. For example, Wu et al. [59] demonstrated that HONO emissions are significantly affected by the combined effects of soil moisture and temperature, which exhibit a Gaussian distribution. Consequently, relying solely on soil moisture as an indicator may not accurately capture the dynamic variations in HONO emissions.

Additionally, the effects of microbial species and their denitrification intensities on HONO emissions are not effectively parameterized in the current FEST-C* model, which may increase the uncertainty in HONO estimates. Future research should prioritize optimizing the parameterization mechanisms for NO and HONO production by various denitrifying microbiomes within the FEST-C* model to enhance its predictive accuracy. Furthermore, manure application was also not incorporated in the cropland Nr emission simulations. Integrating detailed data on manure use from livestock farming into the FEST-C* model is essential to reduce uncertainties in NO and HONO estimations. Another critical limitation is the scarcity of comprehensive observational data for validating simulated NO and HONO emissions. Continuous field measurements of NO and HONO fluxes should be conducted to establish robust datasets for model validation. These efforts will help improve the reliability of simulated cropland Nr emissions and their contributions to regional ozone formation, thereby supporting the development of more effective air quality management and mitigation strategies.

4. Conclusions

Because of the continuous growth of the global population, the extensive application of fertilizers is necessary to ensure sufficient food production; however, this results in substantial Nr emissions during cropland planting, significantly impacting the regional air quality. This study presents a comprehensive analysis of cropland Nr emissions and their impact on O₃ formation across different regions of China in June 2020, utilizing the enhanced process-based agroecological model (FEST-C*) integrated with the Eulerian air quality model (CMAQ). Fertilizer-driven total Nr emission were estimated at 1.26 Tg, with North China accounting for 43% of the total. Cropland Nr emissions contributed about 8% to the national monthly MDA8 O₃ concentration, with localized increases exceeding 11.71 μg m⁻³ in agricultural hotspots. The region most impacted by cropland Nr emissions in terms of MDA8 O₃ concentration was North China, followed by Northwest China, Southeast China, Southwest China, Northeast China, and the Tibetan Plateau. The contributions of cropland Nr emissions to regional MDA8 O₃ in these regions ranges from 7% to 10%, reflecting the regional variability influenced by fertilization intensity, soil conditions, and climatic factors. Notably, cropland Nr emissions led to reductions in hourly O₃ concentrations of 2–6 μg m⁻³ in southern North China and northern Southeast China, primarily during nighttime. This reduction was attributed to the extensive nocturnal titration of surface O₃ by NO, resulting in the formation of nitrate radicals. Combined with previous studies, the impact of cropland Nr emissions on regional MDA8 O₃ has been constrained to approximately 8%, within the range of 3% to 17%. These findings will offer valuable insights into developing strategies to balance agricultural production, fertilizer usage, and environmental pollution mitigation in China.