Ozone deposition measurements over wheat fields in the North China Plain: variability and related factors of deposition flux and velocity

Ozone (O₃) deposition is the main sink of surface O₃, exerting great influences on air quality and ecosystems. Due to instrument limitations and method shortages, O₃ deposition was less observed and investigated in China, where O₃ has been reported to be continuously and significantly rising. Here, we conducted comprehensive measurements of O₃ deposition over a wheat canopy at a typical polluted agricultural site in the North China Plain using a newly developed relaxed eddy accumulation system. For the main wheat growing season in 2023, O₃ deposition flux and velocity (V_d) averaged −0.25 ± 0.39 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{s}}^{-\mathrm{1}}$ and 0.29 ± 0.33 cm s⁻¹, respectively. Daytime V_d (0.40 ± 0.38 cm s⁻¹) was obviously higher than in the nighttime (0.17 ± 0.26 cm s⁻¹). The temporal changes in V_d were mainly determined by crop growth, and V_d significantly increased with decreasing relative humidity and increasing friction velocity and soil water content, enhanced by a higher leaf area index. With rapid increases in soil moisture, simultaneous and following overall increments in V_d were detected, attributed to remarkably strengthening O₃ stomatal uptake under increased stomatal conductance and extended opening into the night, and more non-stomatal O₃ removal at night resulted from strengthened soil NO emission in moist conditions. This study confirms the leading effects of crop growth on O₃ deposition modulated by environmental conditions and the non-negligible influences of nocturnal plant activities, and it emphasizes the need for O₃ deposition observation over different surfaces and accurate evaluation of O₃ agricultural impacts based on deposition fluxes.

1 Introduction

Surface ozone (O₃) is a key secondary air pollutant, generated in photochemical reactions involving volatile organic compounds (VOCs) and nitrogen oxides (${\mathrm{NO}}_x=\mathrm{NO}+{\mathrm{NO}}_{\mathrm{2}}$) (Seinfeld et al., 2006). Over the past 2 decades, China's rapid economic development and increasing anthropogenic emissions of NO_x and VOCs have led to significantly upward trends in O₃ concentrations (Monks et al., 2015; Li et al., 2019; Lu et al., 2020; Xu et al., 2020), especially in the North China Plain (NCP) (Tai et al., 2014; Ma et al., 2016; Wang et al., 2017, 2022; Lu et al., 2020; Xu, 2021; Lyu et al., 2023). Dry deposition plays one of the key roles in removing surface O₃ (e.g., Tang et al., 2017) and contributes about 20 % to the annual global tropospheric O₃ loss (Lelieveld and Dentener, 2000; Wild, 2007; Hardacre et al., 2015). Over vegetated areas, stomatal and non-stomatal uptake of O₃ may represent a major part of the total dry deposition (Fowler et al., 2001). Plant uptake of large amounts of O₃ may cause a series of deleterious oxidative reactions, damaging vegetation and threatening crop quality and production (Ainsworth, 2017; Harmens et al., 2018; Mills et al., 2018; Feng et al., 2019a). In addition, O₃ deposition to the ground surfaces (including soil, snow and water) is closely related to tropospheric chemistry, air quality and ecosystems (Clifton et al., 2020b; Stella et al., 2019; Helmig et al., 2012; Stocker et al., 1995). Under the rapid expansion of population and growing demands for food, China has become the world's largest crop producer, as well as importer (Dong et al., 2021). O₃ deposition is thus of great importance, and its accurate quantification is urgently needed to evaluate the impact of increasing O₃ levels on agricultural production, ecosystems, air quality, human health and global climate.

O₃ deposition has been measured over various ecosystems, including forest, grassland, cropland and bare-soil environments (Table S1 in the Supplement), in order to understand deposition mechanisms and evaluate its potential effects (Stella et al., 2019; Xu et al., 2018; Zhu et al., 2015; Helmig et al., 2012; Mészáros et al., 2009; Lamaud et al., 2009; Coyle et al., 2009). However, the deposition processes are controlled by various biotic (stomatal uptake) and abiotic (non-stomatal removal) activities that are simultaneously modulated by environmental factors. The relative contributions of stomatal and non-stomatal O₃ deposition varied with land cover, plant species and growth stages, as well as environmental factors. Stomatal uptake of O₃ depends on the opening and closure of stomata on leaf surfaces. For example, the fraction of diurnal maximum stomatal O₃ deposition over boreal forests ranged from 56 % to 74 % (Rannik et al., 2012) while only accounting for 31.2 % in a wheat field (Xu et al., 2018), with both of them peaking at midday during the most rigorous growth stage of vegetation (Xu et al., 2018; Rannik et al., 2012). Non-stomatal resistance of O₃ decreased with the increasing temperature and friction velocity and was ∼50 % lower under wet conditions than under dry conditions over the same potato canopy (Coyle et al., 2009). Thus, O₃ deposition is dominated by distinct deposition processes over different surfaces in different environments.

Currently, the eddy covariance (EC) method and flux-gradient (FG) approach are the most commonly used micrometeorological techniques for measuring O₃ vertical fluxes (Businger and Oncley, 1990; Altimir et al., 2006; Wu et al., 2015; Clifton et al., 2020a). However, EC requires robust fast-response measurement instruments (≥10 Hz) (Hicks and Wesely, 1978; Muller et al., 2009), while the assumption of the FG approach is dependent on surface roughness and the photochemical reactions of O₃ and its precursors (Raupach and Thom, 1981; Vilà-Guerau De Arellano and Duynkerke, 1992). These relatively high requirements have more or less limited the application of traditional micrometeorological methods to measurements of O₃ flux. The relaxed eddy accumulation (REA) method is another important micrometeorological method for observing the air–surface exchange of substances of interest over ecosystems (Desjardins, 1977; Businger and Oncley, 1990). REA overcomes the need for fast-response gas sensors and is based on the same physical principle as EC without introducing other uncertainties (Pattey et al., 1993). REA relies on the conditional sampling of air at a constant flow rate according to the instantaneous vertical velocity, which requires high-response sampling valves (∼10 Hz). The air samples associated with updrafts and downdrafts are accumulated into two separate reservoirs and accurately measured with slow-response gas analyzers (Businger and Oncley, 1990). In addition, REA sampling systems are low-cost; easily portable; and simple to operate at remote locations such as forests, croplands and grassland surfaces (Sarkar et al., 2020). Thus, REA methods have been widely applied in flux measurements of various species, such as biogenic VOCs (Mochizuki et al., 2014), reduced sulfur gases (Xu et al., 2002), HONO (Ren et al., 2011) and aerosols (Matsuda et al., 2015; Xu et al., 2021) above forest canopies; peroxyacetyl nitrate at a grassland site (Moravek et al., 2014); NH₃ above fertilized corn (Nelson et al., 2017); and Hg at an urban site and over a boreal peatland (Osterwalder et al., 2016). To the best of our knowledge, the REA method has not been applied to O₃ deposition flux measurements so far.

Although many regions in China have been experiencing severe O₃ pollution during growing seasons, measurements of O₃ flux over crop fields in the country have only been sporadically reported and were made using either chamber techniques (e.g., Tong et al., 2015) or micrometeorological approaches (Zhu et al., 2014, 2015; Xu et al., 2018). In this study, we developed a new REA flux system and applied it to obtain O₃ deposition fluxes over wheat fields in the NCP during the springtime growing season. Based on these in situ observations, we evaluated the feasibility of O₃ flux measurements using the REA method, analyzed the variation characteristics of O₃ deposition during the wheat growing season, and identified the key drivers in the variability in daytime and nighttime O₃ deposition during distinct crop growth stages and under different environmental conditions.

2 Observation and method

2.1 Site description

The flux observations were conducted at the Gucheng site (39°08^′ N, 115°40^′ E; GC), an integrated ecological–meteorological observation and research station of the Chinese Academy of Meteorological Sciences, located 35 km to the northeast of the city of Baoding, Hebei Province, and 100 km southwest of urban Beijing. The site is mainly surrounded by irrigated high-yield agricultural lands with small villages and a highway connecting Beijing and Shijiazhuang 7 km to the west of the site (Fig. S1 in the Supplement). The fields within and surrounding the yard of GC are on a winter wheat–summer maize rotation, which is typical in northern China. Observations at the site have revealed good regional representativeness of the agricultural areas in the NCP that are heavily impacted by the severe regional air pollution (Lin et al., 2009; Xu et al., 2019; Kuang et al., 2020; Zhang et al., 2022a, b).

2.2 Relaxed eddy accumulation (REA) technique

2.2.1 Theory

A self-assembled relaxed eddy accumulation system for O₃ dry deposition measurements (REA-O₃ flux system) was deployed at GC. In the REA-O₃ flux system, conditional sampling is conducted according to the direction of vertical wind (w), which separates sampled air into updraft and downdraft reservoirs at a constant flow rate (Desjardins, 1977; Businger and Oncley, 1990). The vertical fluxes of O₃ ($F_{{\mathrm{O}}_{\mathrm{3}}}$, in $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{s}}^{-\mathrm{1}}$) are calculated by the concentration differences between two reservoirs following Eq. (1):

$$\begin{array}{}\text{(1)}& F_{{\mathrm{O}}_{\mathrm{3}}}=\stackrel{\mathrm{‾}}{w^{\prime }{c}^{\prime }}=b{\mathit{\sigma }}_w(\stackrel{\mathrm{‾}}{c^{+}}-\stackrel{\mathrm{‾}}{c^{-}}),\end{array}$$

where σ_w is the standard deviation of vertical wind (in m s⁻¹); $\stackrel{\mathrm{‾}}{c^{+}}$ and $\stackrel{\mathrm{‾}}{c^{-}}$ are the averaged O₃ concentrations in the updraft and downdraft reservoirs, respectively (in µg m⁻³); and b is the eddy accumulation coefficient. The latter is obtained from CO₂ flux measured using the EC method and is calculated using Eq. (2):

$$\begin{array}{}\text{(2)}& b={\displaystyle \frac{\stackrel{\mathrm{‾}}{w^{\prime }{{\mathrm{CO}}_{\mathrm{2}}}^{\prime }}}{{\mathit{\sigma }}_w(\stackrel{\mathrm{‾}}{{{\mathrm{CO}}_{\mathrm{2}}}^{+}}-\stackrel{\mathrm{‾}}{{{\mathrm{CO}}_{\mathrm{2}}}^{-}})}},\end{array}$$

where $\stackrel{\mathrm{‾}}{{{\mathrm{CO}}_{\mathrm{2}}}^{+}}$ and $\stackrel{\mathrm{‾}}{{{\mathrm{CO}}_{\mathrm{2}}}^{-}}$ are averaged CO₂ concentration observed under upward and downward vertical winds, respectively (in mg m⁻³), and $\stackrel{\mathrm{‾}}{w^{\prime }{{\mathrm{CO}}_{\mathrm{2}}}^{\prime }}$ represents the EC CO₂ flux (in $\mathrm{mg}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{s}}^{-\mathrm{1}}$). Based on the measurements from February to June 2023, b revealed an average ± standard deviation of 0.55 ± 0.09, ranging from 0.16 to 0.80.

O₃ deposition velocity (V_d, cm s⁻¹) is estimated based on O₃ flux and concentrations using Eq. (3):

$$\begin{array}{}\text{(3)}& V_{\mathrm{d}}=-{\displaystyle \frac{F_{{\mathrm{O}}_{\mathrm{3}}}}{C_{{\mathrm{O}}_{\mathrm{3}}}}}\times \mathrm{100},\end{array}$$

where $C_{{\mathrm{O}}_{\mathrm{3}}}$ is the 30 min averaged O₃ concentration (in µg m⁻³).

2.2.2 System setup and verification

The setup of the REA-O₃ flux system is depicted in Fig. 1. A 3-D sonic anemometer (CSAT3, Campbell Scientific, Inc., USA) was used for measuring the three wind components (u, v, w) at 10 Hz, which was mounted at the height of 4.5 m on an eddy covariance tower, located in the middle of a cropland. The height of the flux tower was designed according to the result of the fetch and footprint analysis. The range of the flux source region was about 400 m, which is covered by the crop field within the GC station. The inlet (0.125 in. o.d. Teflon tubing) of the REA system was installed in the center of the anemometer. The 10 Hz wind signals, together with the signals from the ${\mathrm{CO}}_{\mathrm{2}}/{\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ analyzer (LI-7500, LI-COR, Inc., USA), were collected by a data logger (CR1000, Campbell Scientific, Inc., USA) and sent to a PC. The wind signals were processed by a program written in Python, which also sent a switch command to two fast-response three-way solenoid valves (LVM105R-5C, SMC Corporation, Japan) according to the vertical wind direction. Based on the direction of instantaneous vertical winds, sample air was drawn alternatively through updraft or downdraft sample tubes (0.25 in. o.d. Teflon) wrapped in aluminum foil and was analyzed by the two UV photometric O₃ analyzers (TE 49i, Thermo Fisher Scientific Inc., USA) installed at the ends of updraft and downdraft channels, respectively. Coarse particulate matter was filtered out of air samples using two particle filters (47 mm single-stage filter assembly, Savillex, LLC., USA) before entering the O₃ analyzers (Fig. 1). To ensure the stability of airflow in the O₃ analyzers and sampling system, zero air was supplied to the channel that was not sampling ambient air. The zero air was generated by an external air compressor (M104, Gast Manufacturing Inc., USA) and a zero-air generator (model 111, Thermo Fisher Scientific Inc., USA), which removes O₃, NO, NO₂, CO and hydrocarbons from ambient air but not water vapor. In addition, both sampling tubes were bypassed through a piston pump (Thomas 617CD22, Gardner Denver Inc., USA) in order to increase the inlet sample flow and avoid axial mixing in front of the solenoid valves. The linear velocity of the air sample in the inlet tubing was set to 22 m s⁻¹, and airflow was at the turbulent state with a Reynolds number over 2300. The estimated residence time from system inlet to the valves was 18 ms, while the response time of the fast-response sampling valves was less than 10 ms, leading to total time delays from the inlet to individual sampling tubes of below 10 ms; thus the REA system could work at a sample frequency of 10 Hz (100 ms). Total residence time of air samples from the tip of the inlet to the point of O₃ detection was about 10 s, which was much shorter that the lifetime of O₃ reacting with NO (“Supplementary Method” in the Supplement), suggesting the chemical reaction in the two channels could be neglected. The O₃ analyzers recorded 1 min averaged O₃ concentrations, which were downloaded by the PC. O₃ data were synchronized with wind data as well as sample time according to the PC time. The actual averaged O₃ concentrations under updraft and downdraft conditions were calculated according to Eq. (4) using the sample time, sample flow and 1 min averaged O₃ concentrations:

$$\begin{array}{}\text{(4)}& \stackrel{\mathrm{‾}}{c}={\displaystyle \frac{{\sum }_{i=\mathrm{1}}^{i=\mathrm{30}}{c}_i\times {\text{flow}}_i}{{\sum }_{i=\mathrm{1}}^{i=\mathrm{30}}{\text{flow}}_i\times t_{\text{sample gas},i}}},\end{array}$$

where c_i is the 1 min averaged O₃ concentration (in µg m⁻³), flow_i is the 1 min averaged sample flow (in slpm) measured by the mass flow meter (MFM) and t_{sample gas,i} is the real time for analyzing air sample within the ith minute (in fractions of a minute). The O₃ concentration was calculated by averaging O₃ concentrations under updraft and downdraft conditions using Eq. (5).

$$\begin{array}{}\text{(5)}& C_{{\mathrm{O}}_{\mathrm{3}}}={\displaystyle \frac{\stackrel{\mathrm{‾}}{c^{+}}+\stackrel{\mathrm{‾}}{c^{-}}}{\mathrm{2}}}\end{array}$$

Figure 1 Schematic of the REA-O₃ flux system (w>w₀).

To increase the concentration difference between updraft and downdraft, a wind speed threshold called the wind dead band (w₀) was used in the REA system to discard air samples associated with w close to 0. The application of w₀ promotes the sampling of larger eddies that contribute more to gas fluxes. If a proper w₀ is used, the eddy frequency spectrum shifts to the low frequencies in the sample but does not cut off all the high-frequency signals, only filtering out samples with small vertical displacements that have relatively small impacts on the overall flux (Bowling et al., 1999; Held et al., 2008; Tsai et al., 2012; Moravek et al., 2014; Grelle and Keck, 2021). Conditional sampling using w₀ also prolongs the lifetime of the fast-response solenoid valves (Pattey et al., 1993) and effectively avoids sampling error around w=0 due to the limitation of the sonic anemometer (Grelle and Keck, 2021). In the REA-O₃ system, we adopted fixed wind dead bands during the daytime (from 08:00 to 18:00 local time (LT, GMT + 8); w₀=0.05 m s⁻¹) and nighttime (from 19:00 to 07:00 LT; w₀=0.01 m s⁻¹), considering that wind speeds during the daytime were generally larger than those at night. The concentration difference increased with w₀ and led to an increase in measured fluxes. Using raw EC data, the REA CO₂, H₂O and heat fluxes were calculated using w₀ in this study (0.05 m s⁻¹ for daytime and 0.01 m s⁻¹ for nighttime) and w₀=0, respectively, with a constant b of 0.60 (Businger and Oncley, 1990). As shown in Fig. S2 in the Supplement, the two flux datasets revealed excellent correlations during the whole observation period, with a correlation coefficient close to 1, confirming the reliability and stability of the REA flux measurement system. Compared with the fluxes without w₀, CO₂, H₂O and heat fluxes with w₀ exhibited similar small overestimations, reaching around 10 %–13 % during the daytime and 4 %–10 % at night, which were comparable with the influence of a dynamic dead band ($w_{\mathrm{0}}=\frac{{\mathit{\sigma }}_w}{\mathrm{0.35}}$) in Grelle and Keck's (2021) REA system for H₂O, CO₂, CH₄ and N₂O flux measurements. This indicates that adopting the wind dead band in our REA system only had a marginal impact on observed fluxes.

To identify potential flux errors induced by any differences between updraft and downdraft channels in the REA-O₃ system (including valves, sample tubes and O₃ analyzers), the parallelism of the two sampling channels was checked through simultaneous direct sampling of ambient O₃ from the REA system inlet during 10–11 May. As shown in Fig. S3 in the Supplement, O₃ data points obtained from the two channels all aligned close to the 1:1 line (slope=1.02, p<0.01), suggesting that the difference in measured O₃ was minimal between updraft and downdraft reservoirs and that its impact on flux measurement could be ignored. Moreover, synchronous multipoint calibrations of the two channels were conducted monthly. Different O₃ concentrations generated by an O₃ calibrator (TE 49C PS, Thermo Fisher Scientific Inc., USA) were introduced into the system from the zero-air inlet and simultaneously measured by the two O₃ analyzers. Figure S4 in the Supplement presents the results of multipoint calibration, with determination coefficients (R²) reaching 1.000, indicating high stability of the two channels and O₃ instruments. O₃ concentrations in the two channels were adjusted based on the results of parallel experiments and standard calibrations.

To further verify the reliability of the REA system, the flux data derived from the REA technique were compared with those based on the EC theory. For CO₂, H₂O and heat, the averaged ratios of REA to EC fluxes were all slightly larger than 1, indicating small overestimates in the REA flux measurement system, which were expected due to the use of w₀. Most of the flux data maintained high consistency, with correlation coefficients close to 1 (Fig. S5 in the Supplement), confirming that the REA system performed reliably under most conditions.

2.3 Field measurements and ancillary data

Measurements of O₃ flux were conducted during the main wheat growing season in 2023, from the late dormancy stage (13 February 2023) to wheat harvest (18 June 2023). According to the winter wheat phenology at GC (Table S1), its entire growth season could be divided into three stages: the over-wintering (O-W, 13 February–5 March), green-flowering (G-F, 6 March–28 May) and ripening-harvest (R-H, 29 May–18 June) stages. The wheat height increased from 6.0 cm during the O-W stage to 61.2 cm at the R-H stage. Ancillary data were obtained for further analysis, including meteorology data, soil parameters, plant growth indicators and O₃-related trace gas measurement data. Meteorological variables including air temperature (T_Air), relative humidity (RH), precipitation, soil temperature (T_Soil) and volumetric water content (soil VWC) at 20 cm, global solar radiation (G), photosynthetic active radiation (PAR), and the sun elevation angle were measured by an automatic weather station at GC. The 30 min CO₂, H₂O, heat and momentum fluxes were measured by the EC system, which includes a 3-D sonic anemometer, an open-path ${\mathrm{CO}}_{\mathrm{2}}/{\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ analyzer and a data logger. The friction velocity (u_∗) was calculated using the three wind components (u, v, w) following Eq. (6), and the vapor pressure deficit (VPD) was estimated as in Eq. (7).

$$\begin{array}{}\text{(6)}& {\displaystyle }{u}_{\ast }=({\stackrel{\mathrm{‾}}{u^{\prime }{w}^{\prime }}}^{\mathrm{2}}+{\stackrel{\mathrm{‾}}{v^{\prime }{w}^{\prime }}}^{\mathrm{2}}{)}^{\mathrm{1}/\mathrm{4}}\text{(7)}& {\displaystyle }\begin{array}{rl}{\displaystyle }\text{VPD}=& {\displaystyle }\left(\mathrm{1}-{\displaystyle \frac{\text{RH}}{\mathrm{100}}}\right)\times \mathrm{611.2}\\ {\displaystyle }& {\displaystyle }\times \exp \left({\displaystyle \frac{\mathrm{17.62}\times T_{\text{Air}}}{\mathrm{243.12}+T_{\text{Air}}}}\right)÷\mathrm{100}\end{array}\end{array}$$

The leaf area index (LAI) and the fraction of photosynthetically active radiation (FPAR) were obtained from the Moderate Resolution Imaging Spectroradiometer (MODIS) Level 4 product (MCD15A3H) with a spatial resolution of 500 m and temporal resolution of 4 d (Myneni et al., 2015).

NO_x ($\mathrm{NO}/{\mathrm{NO}}_{\mathrm{2}}/{\mathrm{NO}}_x$) concentrations were monitored from 18 March to 2 June by a $\mathrm{NO}/{\mathrm{NO}}_{\mathrm{2}}/{\mathrm{NO}}_x$ trace level analyzer (model 42C-TL, Thermo Fisher Scientific Inc., USA). The analyzer is installed in an air-conditioned container on the northern edge of the cropland, which is located 200 m north of the eddy covariance tower. The air inlet is 1.8 m above the roof of the container and ∼4.5 m above ground level, with an estimated residence time of less than 3 s. Multipoint calibrations of NO_x were made using a NO standard gas obtained from National Institute of Metrology, Beijing, China. A total set of 3631 data points of 30 min averaged NO_x were obtained, as shown in Fig. S6 in the Supplement. During the measurement period, NO concentration ranged from 0.1 to 23.3 ppb with an average of 0.8 ± 1.8 ppb, while NO₂ ranged from 0.5 to 6.6 ppb with an overall average of 2.3 ± 0.9 ppb.

2.4 Stepwise multiple linear regression (MLR) model

Stepwise multiple linear regression (MLR) models were applied to identify the key environmental factors influencing O₃ deposition in the daytime (sun elevation angle >0°) and at nighttime (sun elevation angle <0°), respectively. MLR is a commonly used approach to describe the relationship between air pollution and its influencing factors (Zhang et al., 2022b; Han et al., 2020; Fu and Tai, 2015; Rannik et al., 2012). The stepwise MLR model takes the following form:

$$\begin{array}{}\text{(8)}& y={\mathit{\beta }}_{\mathrm{0}}+{\sum }_{k=\mathrm{1}}^n{\mathit{\beta }}_k{x}_k,\end{array}$$

where y is the observed V_d, x_k is the selected normalized environmental parameter, β₀ is the regression constant, β_k is the regression coefficient and n is the number of selected terms. β_k is determined by a forward stepwise method to add and delete terms to obtain the best model fit based on Akaike information criterion (AIC) statistics (Venables and Ripley, 2003). The Z-score normalization method was adopted according to the following equation:

$$\begin{array}{}\text{(9)}& x=(x_{\text{observed}}-x_{\text{mean}})÷x_{\text{standard deviation}},\end{array}$$

where x_observed, x_mean and x_{standard deviation} are the observed parameters, its overall average and its standard deviation, respectively. Environmental parameters including seven meteorological and soil factors (T_Air, RH, VPD, u_∗, T_Soil, soil VWC, PAR) and two crop-related factors (LAI, FPAR) were considered during the daytime, while PAR was unaccounted for during the nighttime. The selected variables in the stepwise MLR were considered to be the environmental factors critical for O₃ deposition at GC during the wheat growing season.

3 Results and discussion

3.1 Meteorological conditions

Figure 2 shows the temporal variations in daily meteorological and soil conditions during the whole period. Air and soil temperatures gradually increased from the lowest values (T_Air: −0.8 °C; T_Soil: 2.9 °C) in February to the highest ones (T_Air: 30.2 °C; T_Soil: 30.9 °C) in June. RH varied around a higher level before the latter part of April and at a lower level after that, with an average of 64 % ± 17 %. Calculated u_∗ fluctuated in the range of 0.05–0.30 m s⁻¹, with an average of 0.17 ± 0.04 m s⁻¹. Daily VPD was relatively stable during February–early April, with an average of 5.9 ± 4.0 hPa, and rose obviously afterwards, reaching an averaged value of 19.9 ± 7.4 hPa during May to June. During the wheat growth stage, the cropland experienced two irrigation events, occurring on 19–21 April and 18–22 May, respectively (Fig. 2c). The irrigation system mainly consists of an irrigation pump with an outlet flow of 50 m³ h⁻¹ and a total of 20 sprinklers. The total irrigation water volume was approximately 3600 m³ for 3333 m² cropland, and the duration was 3 d. Soil VWC stayed flat before the middle of April and showed dramatic boosts caused by strong precipitation or irrigation events during 9–10, 20–21 and 28–29 April and 19–20 May (Fig. 2c), followed by slow declines due to evapotranspiration under higher temperatures. Both PAR and G exhibited great fluctuations with slight increases from February to June.

Figure 2 Daily meteorological and soil conditions from 12 February to 18 June 2023: (a) T_Air and RH; (b) u_∗ and VPD; (c) irrigation (black arrows), precipitation, T_Soil and soil VWC; and (d) G and PAR.

3.2 O₃ flux, deposition and concentration

In total, 2728 pairs of O₃ deposition flux and velocity were obtained for the wheat growing season, which are presented in Fig. 3, along with 30 min and daily average O₃ concentrations. O₃ deposition flux averaged −0.25 ± 0.39 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{s}}^{-\mathrm{1}}$, with larger deposition during the daytime (−0.39 ± 0.45 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{s}}^{-\mathrm{1}}$) and smaller deposition during the nighttime (−0.08 ± 0.21 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{s}}^{-\mathrm{1}}$). The largest negative (−3.20 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{s}}^{-\mathrm{1}}$) and positive (0.14 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{s}}^{-\mathrm{1}}$) fluxes were measured around noontime on 29 April and around midnight on 15 March, respectively. Daytime V_d averaged 0.40 ± 0.38 cm s⁻¹ and was distinctly higher than nighttime V_d values (0.17 ± 0.26 cm s⁻¹). The averages of daytime and nighttime V_d obtained in this study were comparable to those from previous EC-based observations (0.42 and 0.14 cm s⁻¹) during the wheat growing season and higher than those (0.29 and 0.09 cm s⁻¹) during the maize growing season (Table S2 in the Supplement) in Shandong Province, China (Zhu et al., 2015, 2014). V_d averaged 0.29 ± 0.33 cm s⁻¹ over the whole observation period, ranging from −0.39 to 2.65 cm s⁻¹. The average O₃ deposition velocities observed over the wheat canopy did not show substantial differences from those previously reported for grasslands (Mészáros et al., 2009; Coyle, 2006), forests (Wu et al., 2015; Rannik et al., 2012) and bare soil (Stella et al., 2019) (Table S2). O₃ concentrations over the wheat canopy were significantly enhanced after April, with an overall average of 61.8 ± 34.6 µg m⁻³. In general, O₃ deposition velocities were more pronounced from mid-April to late May (Fig. 3b), when wheat was growing vigorously, while deposition fluxes were higher after late May due to overall higher O₃ concentration (Fig. 3a and c). Thus, O₃ concentration was more determinative of O₃ deposition flux than V_d on longer timescales.

Figure 3 Time series of O₃ deposition flux (a), V_d (b) and concentration (c) during the wheat growing season.

The averaged diurnal patterns of O₃ deposition flux, velocity and O₃ concentration are depicted in Fig. 4. With solar radiation and atmospheric turbulence increasing after sunrise, plant stomatal conductance increased along with H₂O and CO₂ fluxes over the cropland, reaching peaks at noon (Fig. 5). O₃ deposition rapidly rose during the morning (06:00–10:00 LT). Deposition flux and velocity both reached their peaks (−0.62 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{s}}^{-\mathrm{1}}$ and 0.54 cm s⁻¹) by 13:00 LT, when stomatal conductance and gas–leaf exchange also reached the diurnal maximums (Rannik et al., 2012; Otu-Larbi et al., 2021). O₃ deposition quickly decreased from 14:00 to 18:00 LT despite high levels of O₃ (Fig. 4). At night, atmospheric turbulence weakened and leaf stomata closed, resulting in reduced H₂O and CO₂ fluxes remaining steady throughout the night (Fig. 5). Nighttime O₃ deposition remained at relatively low levels and exhibited weak changes, with an averaged flux and V_d of −0.09 ± 0.04 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{s}}^{-\mathrm{1}}$ and 0.17 ± 0.02 cm s⁻¹, respectively. Therefore, diel variations in O₃ deposition over the wheat fields were mainly driven by stomatal opening and closing, with O₃ deposition velocity being decisive in deposition flux diurnal variations.

Figure 4 Diurnal variations in the O₃ deposition flux, V_d and concentration during the wheat growing season, with error bars representing average ± 0.2 × standard deviation values.

Figure 5 Diurnal variations in (a) H₂O flux ($F_{{\mathrm{H}}_{\mathrm{2}}\mathrm{O}}$) and CO₂ flux ($F_{{\mathrm{CO}}_{\mathrm{2}}}$) and (b) u_∗ and PAR during the wheat growth season, with error bars representing average ± standard deviation $/\sqrt{n}$.

3.3 O₃ deposition in different stages of wheat growth

To investigate the influences of wheat growth on O₃ deposition, the characteristics of O₃ deposition were further examined in connection to the different growth stages. During the O-W stage, wheat was in dormancy and leaves had not begun to turn green (LAI<0.5, Fig. 6b), with CO₂ flux in the agricultural ecosystem close to zero (Fig. 6c). V_d in the O-W stage barely changed, exhibiting a low average value of 0.20 ± 0.28 cm s⁻¹ and a median of 0.12 cm s⁻¹ (Table 1). Wheat grew vigorously in the G-F stage, with LAI and CO₂ assimilation flux exhibiting rapid increases until the early and late flowering stages, respectively, after which both of them gradually decreased (Fig. 6b and c). O₃ deposition varied nearly in synchronization with LAI and wheat growth, with V_d reaching a peak when cropland CO₂ assimilation was the highest during the G-F stage (Fig. 6a), reaching highest daytime and nighttime averages of 0.46 ± 0.41 cm s⁻¹ and 0.24 ± 0.28 cm s⁻¹, respectively (Table 1). Afterwards, with the maturing of wheat and the aging of leaves in the R-H stage, V_d quickly dropped back to a low average level of 0.20 ± 0.25 cm s⁻¹, which is similar to that observed in the O-W stage. It can be seen that the temporal variation in O₃ deposition velocity over wheat fields was predominantly determined by crop growth at GC. As for the deposition flux, both the daily and the daytime average fluxes during the G-F stage were comparable with those in the R-H stage (Table 1), which can be attributed to the high O₃ concentrations in the summer months (Zhang et al., 2022a; Lin et al., 2009). Although nighttime O₃ concentration during the G-F stage was also 58 % lower than that in the R-H stage, nighttime O₃ deposition flux during the G-F stage was still the highest among the three stages, which was related to the remarkably high nighttime deposition velocities during the G-F stage (Table 1).

Figure 6 (a) O₃ V_d, (b) LAI and FPAR, and (c) CO₂ flux ($F_{{\mathrm{CO}}_{\mathrm{2}}}$) in different wheat growing stages. The circles and error bars in (a) denote the weekly medians and quantiles of V_d, respectively. O-W, G-F and R-H represent the over-wintering, green-flowering and ripening-harvest stages, respectively.

Table 1 Summary of daily, daytime and nighttime O₃ V_d and flux during the O-W, G-F and R-H stages.

3.4 O₃ deposition relation to environmental factors

To gain deeper insights into the responses of O₃ deposition (including stomatal and non-stomatal) to environmental factors in agricultural areas, stepwise MLR models were conducted to see which factors potentially played more important roles in O₃ deposition at GC during the daytime and nighttime, respectively. As shown in Table 2, RH, u_∗, soil VWC and LAI were identified as significant environmental factors in explaining daytime O₃ deposition changes during the entire observation period. Additionally, the coefficient of determination (R²) of the linear model between all environmental factors and V_d was 0.46, implying that these meteorological and plant-growth-related factors could explain approximately 46 % of the variance of daytime O₃ deposition, while R² was only slightly lower (0.43) with the four selected factors. Distinct key environmental factors for O₃ deposition were identified for different wheat growth stages, while LAI was only among the most important factors during the G-F stage (Table 2), confirming the significant effect of crops on O₃ deposition during its most vigorous growing stage. During the O-W stage, wheat played a minor role in O₃ deposition as LAI<0.5. Solar radiation (PAR) and wind (u_∗) supplied energy for atmospheric turbulence, which transported O₃ to the soil surface, while soil moisture (soil VWC) largely affected the diffusion and absorption of O₃ in soil (Stella et al., 2011a). Therefore, O₃ deposition was more sensitive to u_∗, soil VWC and PAR during the O-W stage. In the R-H stage, the land surface was covered by ripe wheat, which reduced LAI and stomatal conductance. O₃ deposition was therefore mainly dependent on turbulent transport, which was more affected by T_Air and u_∗ under sufficient solar radiation (Fig. 2d). During the nighttime, O₃ deposition mainly commenced through non-stomatal pathways such as cuticular and soil deposition (Xu et al., 2018), which are affected by turbulence strength and surface conditions. The most significant influencing factors for O₃ deposition were T_Air, u_∗, T_Soil and soil VWC for the whole observation period (Table 2). However, temperatures (T_Air and T_Soil) were not selected as the key factors when the stepwise MLR was separately performed for the three growth stages. Compared with the R² of the daytime MLR model, meteorological and soil conditions could explain more variance of the nighttime O₃ deposition (54 %), implying that nighttime O₃ deposition processes were less complicated than daytime ones.

Table 2 Results of the MLR models for the O-W, G-F and R-H stages. Daytime MLR models represent multiple linear regression between daily average environmental variables and daytime O₃ V_d, while nighttime models represent nighttime environmental variables and V_d. The selected MLR models refer to the stepwise MLR model based on AIC statistics. Bold numbers denote those with p value<0.05.

Further, we explored how the selected key factors controlled the temporal variability in V_d during the three wheat growth stages (Figs. 7–9). In general, the responses of both daytime and nighttime V_d to the meteorological factors were consistent throughout the entire wheat growth season (Figs. 7 and 9).

Figure 7 Dependencies of daytime O₃ V_d on (a) RH, (b) u_∗, (c) PAR and (d) soil VWC during the O-W, G-F and R-H stages. Medians of 30 min V_d with quartiles are presented.

Figure 8 The variation in daytime V_d with (a) u_∗ and (b) soil VWC under changing LAI during the G-F stage.

Figure 9 Dependencies of O₃ V_d on (a) T_Air, (b) RH, (c) VPD, (d) u_∗, (e) T_Soil and (f) soil VWC in the nighttime during the O-W, G-F and R-H stages. Medians of 30 min V_d with quartiles are presented.

Ambient RH influences the relative contributions of stomatal and non-stomatal processes to O₃ deposition. A clear negative variation in V_d with increasing RH was observed at GC for both daytime and nighttime (Figs. 7a and 9b). Due to the negative correlation of VPD to RH (Eq. 7), high VPD was conducive to high nighttime V_d (Fig. 9c). During the daytime, this negative relationship was detected for all growth stages, while during the nighttime the decrease with RH was most evident during the G-F stage. At RH above 60 %–70 %, leaf surfaces are frequently covered by a thin liquid film or by dew water, which would inhibit stomatal dry deposition but enhance aqueous reactions of O₃, leading to an enhanced relative contribution of non-stomatal deposition, albeit with high variability (Coyle et al., 2009; Lamaud et al., 2009). Under RH below 60 %, stomatal conductance contributes more to O₃ deposition, which might also be negatively dependent on RH. For instance, similar negative correlations of V_d and RH were observed over wheat and maize canopies in the NCP (Zhu et al., 2014, 2015), while O₃ deposition to a boreal forest revealed strong positive correlation with RH (Rannik et al., 2012), which was attributed to differences in plant varieties and the growth environment. The response of stomata to the changes in humidity is largely dependent on plant cultivars and plant water stress (Camacho et al., 1974; Rawson et al., 1977; Fanourakis et al., 2020). For example, the stomata of a drought-tolerant wheat cultivar closed rapidly with reduced humidity, whereas high-yield-cultivar stomatal conductance increased against decreasing humidity (Kudoyarova et al., 2007). The growth environment affects the aerodynamic roughness of the earth surfaces. Increased roughness could induce stronger turbulent transfer under low-humidity conditions, transporting gases more effectively to the leaf surfaces, thereby promoting O₃ deposition (Liao et al., 2022).

Strong turbulence (represented by elevated u_∗) can transport O₃ more efficiently to the surface (Cape et al., 2009); thus V_d almost linearly increased with u_∗ (Figs. 7b and 9d). The sensitivity of V_d to u_∗ was also affected by LAI, with daytime V_d under similar levels of u_∗ being significantly higher under higher LAI (Fig. 8a and Table S3 in the Supplement). High LAI indicates dense vegetation coverage and potential large stomatal conductance, which can provide more active (stomatal and cuticular) areas for the uptake of O₃, further promoting O₃ deposition. PAR had a positive effect on O₃ deposition during the observation period (Fig. 7c). On the one hand, increasing PAR induces automatic leaf stomatal opening, thereby determining stomatal conductance and net photosynthesis (Yu et al., 2004) and affecting the stomatal O₃ uptake (Tong et al., 2015). On the other hand, PAR (also reflecting radiation intensity) affects O₃ photochemistry directly by accelerating atmospheric photolysis reactions both above and within the canopy and indirectly by influencing the emission of biogenic VOCs (Yang et al., 2021; van Meeningen et al., 2017; Yuan et al., 2016), thus disturbing the distributions of O₃ and its precursors and contributing to non-stomatal O₃ fluxes through surface processes (Fares et al., 2008; Cape et al., 2009). Positive dependencies of V_d on u_∗ and PAR were observed over other crop fields (such as wheat, maize and potato) (Coyle et al., 2009; Zhu et al., 2014, 2015). In addition, both T_Air and T_Soil exhibited weak relationships with nighttime V_d (Fig. 9a and e), which were different from the reported positive correlations between temperature and V_d (Coyle et al., 2009; Rannik et al., 2012). These imply the variability and complexity of O₃ deposition affected by the combined influences of various environmental factors.

During the O-W and R-H stages, soil VWC was at relatively low levels. During the G-F stage, soil VWC reached beyond 0.30 % and V_d rose significantly with rising soil VWC (Figs. 7d and 9f). Although soil moisture blocks the diffusion of O₃ in soil and reduces reactive spaces for O₃ absorption, suppressing O₃ soil deposition (Stella et al., 2011b), it can also promote total O₃ deposition through several indirect pathways. From the plant physiological aspect, stomatal conductance and plant net photosynthesis are both promoted by higher soil VWC (Otu-Larbi et al., 2021; Anav et al., 2018; Medlyn et al., 2011; Jarvis et al., 1997; Ball et al., 1987). Stomatal conductance revealed overshoots after the watering of dry soil; accordingly, significantly increased transpiration and photosynthesis of crops or vegetation were detected (Wu et al., 2021; Reich et al., 2018; Ramírez et al., 2018; Rawson and Clarke, 1988; Popescu, 1967). This effect was reflected at GC by the more obvious response of V_d to soil VWC changes under higher LAI (Fig. 8b and Table S3), confirming again that O₃ deposition during the G-F stage was mainly driven by stomatal deposition rather than soil deposition. Consequently, soil VWC revealed positive coefficients in the MLR models for the G-F period at GC. Our result is qualitatively consistent with the observation-based estimation of stomatal and soil O₃ deposition relative contributions over a wheat canopy in the city of Nanjing, China, which accounted for 41.2 % and 15.4 %, respectively (Xu et al., 2018).

Overall, soil VWC at GC stayed low except for four abruptly increasing events that occurred during the G-F stage, with the highest soil VWC reaching 0.55 % (Fig. 2c). The three most prominent episodes induced by farm field irrigation were more thoroughly investigated to uncover how soil moisture affected O₃ deposition at GC. Interestingly, simultaneous increments in V_d were detected upon the increase in soil VWC, with V_d being distinctly elevated within days afterwards (Fig. 10). Soil VWC increased from 0.32 % to 0.51 % at noontime on 9 April, and V_d rapidly rose from 0.16 to 0.64 cm s⁻¹ at the same time. Dramatic increments in both averaged nighttime and averaged daytime V_d were detected on the following days. During the nighttime, V_d reached an average of 0.30 cm s⁻¹, significantly higher than the average daytime level (0.18 cm s⁻¹) on 9 April, and a maximum of 0.76 cm s⁻¹, which also exceeded the maximum of 0.64 cm s⁻¹ observed during the daytime on 9 April (Fig. 10a). The dramatic rise in V_d on 10 April resulted in a 317 % increase in daytime O₃ deposition flux (from 0.12 to 0.50 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{s}}^{-\mathrm{1}}$), with only a small change in daytime O₃ concentration from 9 April (41.4 µg m⁻³) to 10 April (48.2 µg m⁻³, Fig. S7a in the Supplement). In addition, drastic elevations in V_d during the night and morning periods were also observed following other episodes of sudden increases in soil VWC (Fig. 10b and c). Similar enhancements and disrupted daily cycles of O₃ deposition were also observed over the canopy of a pine forest during rainfall events (Altimir et al., 2006).

Figure 10 Variations in O₃ V_d (blue lines with circles), soil VWC (black lines) and NO concentration (red lines) during (a) 8–13 April, (b) 27 April–1 May and (c) 18–22 May. The shading represents the daytime.

Considering the direct effect of soil moisture on plant physiology, the temporal variations in CO₂ and H₂O fluxes were examined to characterize changes in the transpiration and photosynthesis of wheat affected by the abrupt increases in soil water contents. As shown in Fig. 11, both CO₂ and H₂O fluxes exhibited obvious increases on the days following the soil VWC increments. The daily peaks of H₂O fluxes increased from 0.08 to 0.13 $\mathrm{g}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{s}}^{-\mathrm{1}}$ between 9 and 10 April and from 0.12 to 0.19 $\mathrm{g}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{s}}^{-\mathrm{1}}$ between 28 and 29 April, while daily averaged CO₂ fluxes before and after the abruptly increasing events rose from 0.28 to 0.55 $\mathrm{mg}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{s}}^{-\mathrm{1}}$ and from 0.51 to 0.55 $\mathrm{mg}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{s}}^{-\mathrm{1}}$, respectively (Fig. 11a and b). Subsequently, CO₂ and H₂O fluxes, as well as V_d of O₃, exhibited declines with the slow loss in soil moisture (Figs. 10 and 11). This indicates that the transpiration and photosynthesis of wheat were sharply enhanced after soil water contents increased, leading to larger leaf stomatal conductance and strengthening O₃ stomatal uptake. These results were consistent with those obtained in previous conditional control experiments and field observations (Wu et al., 2021; Reich et al., 2018; Ramírez et al., 2018; Rawson and Clarke, 1988; Popescu, 1967). In addition, moist soil can extend the time window of wheat leaves' stomatal opening, both in the hours after sunset and in the hours before dawn (Schoppach et al., 2020; Ramírez et al., 2018). Stomata can even stay open during the nighttime after precipitation or irrigation events (Kobayashi et al., 2007; Rawson and Clarke, 1988). During the irrigation-induced high-soil-VWC episodes, positive H₂O fluxes were also observed at GC during the night, such as on 10 April and 28 April (Fig. 11a and b), implying that wheat transpiration might not stop over the course of the night and that leaf stomata might have not completely closed, continuing to take up O₃ at night and significantly enhancing nocturnal O₃ deposition.

Figure 11 Variations in H₂O flux ($F_{{\mathrm{H}}_{\mathrm{2}}\mathrm{O}}$, green lines) and CO₂ flux ($F_{{\mathrm{CO}}_{\mathrm{2}}}$, purple lines) during (a) 8–13 April, (b) 27 April–1 May and (c) 18–22 May, with shading representing daytime hours.

Additionally, the phenomenon of high nighttime O₃ deposition (V_d) was always accompanied by positive water vapor fluxes and high NO concentrations and occurred mainly after the rapid increase in soil VWC (Fig. S8 in the Supplement). As shown in Fig. 10, high NO became more frequent at night during the high-soil-VWC events, and nighttime V_d dramatically increased when NO_x (NO and NO₂) fluctuated at obviously higher levels and nighttime O₃ concentration was still at a low level (Fig. S7), indicating more intensive titration consumption of O₃ at night. This might be attributable to the fact that soil NO emissions were promoted by the watering process, as soil water content is a decisive factor in the transformation and emission of reactive nitrogen within soils (Schindlbacher et al., 2004; Ghude et al., 2010; Kim et al., 2012; Weber et al., 2015; Zörner et al., 2016). Enhanced nighttime soil NO emissions may inevitably cause stronger NO titration with O₃ within wheat canopies, facilitating the non-stomatal O₃ deposition at night.

In summary, both daytime and nighttime O₃ deposition fluxes and V_d were significantly affected by environmental conditions, through stomatal and non-stomatal pathways, with crop growth playing a critical role. Abrupt increments in soil moisture induced dramatic changes in V_d, which not only altered the diurnal cycle of O₃ deposition, but also caused large fluctuations in averaged O₃ deposition flux on longer timescales.

4 Conclusions and implications

In this study, we developed a relaxed eddy accumulation (REA) O₃ flux measurement system, verified its reliability and conducted measurements of O₃ deposition using this newly developed REA system over the wheat canopy at a polluted agricultural site (GC) in the NCP during the main wheat growth season. Ancillary data related to O₃ deposition were used in an integrated analysis of the influencing environmental factors. The observed O₃ deposition flux and velocity over the wheat fields at GC reached averages of −0.25 ± 0.39 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{s}}^{-\mathrm{1}}$ and 0.29 ± 0.33 cm s⁻¹, respectively. The diurnal cycle of V_d was controlled by the crop stomatal opening and turbulent transport during the day. V_d was obviously higher during the daytime (0.40 ± 0.38 cm s⁻¹) than nighttime (0.17 ± 0.26 cm s⁻¹). V_d played a decisive role in the diel pattern of O₃ deposition flux, while O₃ concentrations determined the flux variability on longer timescales. The temporal changes in V_d were synchronous with the evolutions of LAI, wheat growth and cropland CO₂ flux, suggesting the determining and enhancement effect of crop growth on O₃ deposition and the predominant contribution of stomatal uptake over wheat fields during its growth season. However, the relative contributions of stomatal and non-stomatal O₃ deposition pathways, which are of crucial importance for studies on O₃ removal and vegetation health impacts, need to be further quantified in our future investigations. While the influence of crops on O₃ deposition through stomatal uptake or surface removal has been extensively investigated in previous studies (Ainsworth, 2017; Aunan et al., 2000; Bender and Weigel, 2011; Biswas et al., 2008; Epa, 2013; Felzer et al., 2005; Harmens et al., 2018; Piikki et al., 2008), the influence of O₃ deposition on crop growth and yield under currently rising O₃ levels in the NCP remains an unsolved issue. Many researchers have assessed the crop yield loss induced by O₃ pollution based on exposure–response functions (Feng et al., 2019b, 2020; Hu et al., 2020). However, the actual exposure is more related to direct deposition flux measurements rather than concentration-based indicators (Zhu et al., 2015). Therefore, agricultural impacts of O₃ should be more accurately quantified in our following studies using stomatal O₃ deposition fluxes that might be obtained from current total O₃ deposition flux measurements using partitioning methods such as those in Fares et al. (2013).

During the wheat growth season, RH, u_∗, soil VWC and LAI were identified as the most significant factors in explaining the changes in O₃ deposition during the daytime through stomatal and non-stomatal pathways, while u_∗ and soil VWC were more important for nocturnal O₃ deposition, which mainly commenced through non-stomatal deposition. V_d significantly increased with the decrease in RH and the increases in u_∗, PAR and soil VWC, especially under higher LAI. Rapid increases in soil VWC after strong precipitation or irrigation events extended stomatal opening to nighttime hours, leading to increased stomatal conductance and enhanced transpiration and photosynthesis of wheat, which remarkably strengthened O₃ stomatal uptake during the daytime and nighttime. Stomatal opening and transpiration are typically assumed to occur specifically during the daytime. However, an increasing number of studies have shown the non-negligible effects of unclosed nocturnal stomata and transpiration for a wide range of plant species (Kukal and Irmak, 2022; Schoppach et al., 2020; Tamang et al., 2019; Ramírez et al., 2018; Hoshika et al., 2018). Therefore, how nocturnal plant activities might interact with the significantly increasing nighttime O₃ levels in China during recent years (Agathokleous et al., 2023; He et al., 2022) is also worth deeper investigation. Aside from influencing stomata opening, drastic changes in soil humidity also strengthened NO soil emissions, facilitating NO titration of O₃ within the canopy and enhancing non-stomatal O₃ removal at night. Therefore, drastically increasing soil moisture simultaneously led to strong increments in V_d. Under current climate-warming trends, extreme weather events (such as extreme precipitation and drought) have increased in frequency in agricultural areas (Yuan et al., 2016), and their effect on agriculture, NO_x emissions, O₃ formation and O₃ deposition requires future attention.

During the entire wheat growth season, O₃ deposition velocity exhibited large fluctuations under changing environmental conditions, with distinct factors determining V_d variability during different wheat growth stages. These key influencing factors and their effects on O₃ deposition would also vary with different canopy types and ground surface conditions. Aside from environmental conditions, agricultural activities also significantly affect O₃ deposition (Mészáros et al., 2009). Therefore, the actual O₃ deposition process bears large uncertainties and varies greatly in space and time. More O₃ deposition observations over different types of land surfaces and vegetation are urgently needed to facilitate the exploration of O₃ dry deposition mechanisms and to optimize current model parameterizations, whose results largely deviate from observed O₃ dry deposition fluxes in crop growth seasons (Clifton et al., 2020a; Hardacre et al., 2015).