Improving model representation of rapid ozone deposition over soil in the central Tibetan Plateau

Abstract

Ozone soil deposition contributes a major part to the total deposition of ozone on land covered by low vegetation and perturbs the ozone budget on both regional and global scales. Large model-observation divergences in ozone soil deposition require continuous efforts to improve the mechanical understanding and model representation. Observation of ozone deposition over bare soil directly meets the requirement. Here, we performed field observation of ozone deposition over bare soil first available in the Tibetan Plateau (TP) using the aerodynamic gradient method. A top ozone deposition velocity with a daily mean of 0.49 ± 0.11 (1 sd) cm s⁻¹ (1 May to 10 July 2019) and an hourly mean maximum across the diel pattern of 0.73 ± 0.67 cm s⁻¹ in the afternoon were recorded. Such rapid ozone deposition was mainly attributed to extremely low soil resistance (R_soil), which was further regulated by median low soil clay content, dry conditions, and strong solar radiation in the central TP. Parameterization of R_soil in the newly developed Stella scheme was demonstrated to be effective according to our verification. An updated scheme was further attained with the inclusion of our observation and better represents the R_soil variability than the Stella scheme. More verification is therefore encouraged and hopefully to improve the Stella scheme. Finally, both the Stella scheme and our updated scheme showed great advantages over the oversimplified scheme in current models and should be considered more seriously for the sake of better representation of ozone soil deposition and its variability.

---

Environmental significance

Tropospheric ozone is an important gaseous pollutant and a short-lived climate forcer that affects air quality, human health and climate. Ozone soil deposition represents a non-negligible fraction of ozone budget. However, it is highly underestimated by current models and therefore leads to ill-representation of ozone distribution. To explore the soil deposition mechanism and improve the model representation, we performed direct measurements of ozone deposition velocity on bare soil in the central Tibetan Plateau (TP), an ideal experimental field for its pristine nature in terms of weak human perturbation on both land use and ozone photochemistry. For the first time, our measurement recorded in the TP the top ozone deposition velocity on soil, which could be accounted for by measured meteorological and edaphic parameters. An updated parameterization scheme based on Stella et al. was summarized and better represented ozone deposition velocity and its environmental variability, relative to the schemes commonly used in chemical transport models. The Stella scheme is therefore worthy of more serious consideration for the sake of better representation of ozone soil deposition and its variability. Our data has also confirmed that ozone soil deposition in the TP is underestimated by current model evaluations.

1 Introduction

Tropospheric ozone is an important gaseous pollutant and a short-lived climate forcer that plays a key role in atmospheric oxidative capacity, oxidative damage to humans and vegetation, and climate change.^1–4 Obtaining accurate temporal and spatial variability of ozone is essential for assessing its role. While dense ozone measurement networks have been established in the United States, Europe, and eastern Asia, and great progress has been made in China in recent years, observational constraints on ozone are still absent in other regions, especially in background areas globally.⁵ Chemical transport models (CTMs) are alternative tools to evaluate the temporal–spatial variability in ozone in those regions where measurement is absent. Simulations of ozone by CTMs are numerical approximations of the budgets of ozone, including chemical reactions that lead to ozone formation and destruction, tropospheric influx from stratosphere–troposphere exchange, deposition loss, and convectional and advectional transport. However, the simulation representations are limited by our current incomplete understanding of and parameterization methods for these budget items.⁶

Simulation representation of tropospheric ozone, especially ozone at the surface, is sensitive to the representation of ozone dry deposition.^7–11 The resistance-analogy parameterization framework for ozone deposition developed by Wesely,¹² is widely accepted by different deposition schemes to explain and parameterize ozone deposition on various land cover types.^13–18 Briefly, the resistance analogy framework conceptualized the dry deposition of ozone or other pollutants as a three-step process including (1) ozone transport downward to a given surface by atmospheric turbulence, (2) mass transfer through the quasi-laminar sublayer around the interface of air and the surface, and (3) uptake on the surface. The ozone deposition velocity, v_d, can be expressed as the reciprocal of the total mass transfer resistance of these three-step processes, namely, aerodynamic resistance R_a, quasi-laminar sublayer resistance R_b, and surface resistance R_c.¹²

Studies comparing different deposition schemes at site-specific sites illustrate that the representation of R_c and its variability cause the largest differences of v_d.^19,20 Among different deposition schemes, R_a is typically modelled with different empirical formulae of the Monin–Obukhov similarity theory (MOST), and R_b is modelled with a same formula. Wu et al.¹⁹ compared R_a + R_b over a forest modelled by four different MOST-based models and found similar results. Combined with the relatively smaller contribution of R_a + R_b than R_c to the total resistance (see Section 3.2 below), more attention is focused on the parameterization of R_c. The parameterization of R_c used in the resistance analogy considers the resistance of ozone uptake by stomatal, cuticular, and ground surfaces, the last of which mainly refers to soil surfaces. Relatively intensive field observations of ozone deposition have been performed over developed canopies.^21–28 Canopy related stomatal and cuticular uptake are the main ozone deposition paths in these observations. However, studies on crops found that the determining role of stomatal and cuticular resistances highly depends on canopy development and abundance.^29,30 On surfaces with lower and sparser canopies, such as growing croplands and grasslands, soil resistance (R_soil) could contribute up to 55% to total ozone deposition resistance; thus, v_d is more sensitive to R_soil for these land cover types.^31,32

Ozone soil deposition contributes heavily to the global budget of ozone deposition because a considerable fraction of global land cover is desert, bare soil, and sparse vegetation. However, the modeling of ozone soil deposition remains highly uncertain. Gross model underestimation of ozone deposition on bare soil (equivalent to ozone soil deposition) was consistently found.^33,34 The large discrepancy of ozone soil deposition and its variability between measurements and simulations implies the poor representations of ozone soil deposition by current CTMs. The poor representation is mainly attributed to the oversimplified R_soil parameterization scheme that uses two-level prescribed values for dry and wet surfaces without more complex response relationships with soil moisture, soil temperature or soil clay content.¹⁹ Modification attempts have been made with the aim of enabling model power to better describe the variability in R_soil. Either a positive dependence of R_soil on soil moisture or a negative dependence on temperature was reported based on site-specific observations, but the expressions between R_soil and moisture or temperature should not be extrapolated to other sites.^35–37 A few studies introduced a universal corrected function of soil moisture or temperature based on mathematical assumptions; however, field observation validation is rare.^7,38 Stella et al.^33,34 summarized their six field observations over bare soil and proposed a dual-parameter semiempirical R_soil parameterization (known as the “Stella scheme”) involving soil clay content and surface relative humidity (RH_surf). In the Stella scheme, RH_surf is believed to better represent the varied and complex effect of meteorological parameters, such as surface temperature (T_surf), soil moisture, and solar radiation. Soil clay content, as a second influencing factor, reasonably accounts for the spatial heterogeneity of soil uptake reactivity toward ozone and therefore R_soil.

Although the Stella scheme seems to have the potential to describe R_soil and its variability, limited field observational validation hinders direct extrapolation or CTMs inclusion. The observations of ozone soil deposition were mostly conducted over agricultural fields after harvest to avoid interference of the canopy.³³ However, frequent agricultural activities, such as plowing, fertilization, and irrigation, can cause transformations in surface soil physics and chemical characteristics and potentially soil uptake reactivity toward ozone. Additionally, strong NO_x and/or VOCs emissions from agricultural soil can participate in the fast atmospheric chemical conversion of ozone and therefore significantly interfere with the measurement of ozone soil deposition.^42,43 Field observation of ozone deposition in environments with negligible NO or highly reactive VOCs sources is ideal in terms of less chemistry interference and human activity perturbation of soil but is sporadically available.³³ In addition, R_soil is highly affected by soil clay contents and hydrothermal conditions based on the previous summary, which are spatially heterogeneous in different climate regions. Previous observations were performed mostly in subtropical climate regions. Field observations in more climate regions are therefore highly demanded (Fig. 1). The TP is a typical highland climate region where both soil clay content and moisture are expected to be different from the subtropical climate regions. And TP is characterized as a highly pristine area and covered with sparse and low vegetation, implying weak perturbation of land use and ozone chemistry on ozone soil deposition.

Fig. 1 Simplified concept of climatic regions with different soil clay contents and hydrothermal conditions. Previous measurements of soil resistance (R_soil) have been conducted mostly in subtropical climate regions in the figure. Data of climate classification is from Chen et al.³⁹ Soil clay content data is from Harmonized World Soil Database v2.0.⁴⁰ Humidity data is referenced to Kummu et al.⁴¹

The @Tibet series field campaign supports ozone deposition measurement over the Tibetan Plateau (TP). In this study, we took advantage of the ideal experimental field and performed the first measurement of ozone deposition velocity on bare soil in the TP, acquired from the aerodynamic gradient method. The aims of this study are (i) to measure v_d in the TP and to identify the factors controlling v_d and R_soil and (ii) to evaluate and improve the Stella scheme by inclusion of our verifications.

2 Methods

2.1 Measurements

The field measurement of ozone deposition was performed from 1 May to 10 July 2019 at Nam Co Comprehensive Observation and Research Station, Chinese Academy of Sciences (CAS) (NMC site, 30°46.30′N, 90°59.31′E, 4730 m a.s.l.), as an essential part of the @Tibet 2019 campaign. Due to power outages, instrument malfunctions and other factors, the valid data cover 50 days. Detailed descriptions of NMC site and @Tibet 2019 campaign can be found in previous studies.^44–46 NMC site is located in an open flat field ∼1 km southeast of Nam Co Lake and more than 15 km to the north and west of the Nyainqêntanglha Mountains. In addition, a sublake of Nam Co Lake is ∼280 m away from the site. NMC site is characterized by a continental background of the atmosphere with few surrounding populations, and isolated from Lhasa, the largest city in the TP.

High-altitude and mountainous environment results in a typical arctic–highland climate of this area. The monthly mean air temperature ranges from −7.8 °C to 12.2 °C, and solar radiation reaches 1200 W m⁻² in May.⁴⁷ Annual precipitation of approximately 400 mm is concentrated during the Asian Summer Monsoon (ASM) period from June to September.⁴⁸ In this cold and arid area, the soil is poorly developed because of sporadic vegetation coverage and weak chemical weathering, which is embodied in a loose soil structure and low clay content.⁴⁹

Air temperature and relative humidity (HMP155A, Vaisala, FI) and wind speed (010C, Met One, USA) profiles were measured at 1.8, 3.8, and 5.8 m height, and wind direction (020C, Met One, USA) was measured at 5.8 m. Solar radiation, including incoming and reflected shortwave and longwave radiation (CNR4, Kipp & Zonen, NL) and j(NO₂) (ultrafast CCD-detector spectrometer, Metcon, GER), were measured. Soil temperature and soil water content (CS655, Campbell Scientific Inc., USA) profiles were also measured at 5, 15, and 25 cm below ground. NO₂ concentration was measured by a sensitive incoherent broadband cavity enhanced absorption spectroscopy NO₂ analyzer.⁵⁰ As the NO concentration was too low to be measured, photostationary state calculation of the NO concentration (NO_PSS) was carried out with measured O₃, NO_2, and j(NO₂).

Ozone flux was measured using the aerodynamic gradient (AG) method, with characterization of the micrometeorological environment by the eddy covariance (EC). The AG method is based on K-theory, an application of MOST. Similar to Fick's Law, the K-theory assumes that the turbulence flux can be expressed as the product of the turbulence exchange coefficient K and vertical concentration gradient. Therefore, the ozone flux can be calculated as follows:

where F is the ozone flux (ppbv m s⁻¹ or nmol m⁻² s⁻¹), K is the turbulence exchange coefficient (m² s⁻¹), and is the vertical ozone gradient between the two measurement heights (ppbv m⁻¹ or nmol m⁻²). A UV photometric ozone analyzer (Model 49i, Thermo-Environmental Instruments Inc., USA) measured the ozone concentration. Air was sampled at 6.8 m and 1.8 m of the tower through a 27 m Teflon sampling tube with sequential sampling in a 30 min duty cycle. A solenoid valve was used to control the measurement time of 15 min at each height. The measurement gap of 15 min at each height was filled by linear interpolation of adjacent data at the same height. Therefore, continuous ozone gradient was obtained. The sampling tube was covered by black foam for heat preservation and light shading. K is calculated as follows:⁵¹where κ is von Karman's constant (0.4), u_* is the friction velocity (m s⁻¹), z₁ and z₂ are the lower and greater heights of ozone measurement respectively (m), d is the zero-plane displacement height (m), L is the Obukhov length (m), and ψ_h(z/L) is the integrated stability correction function for heat:where y = (1 − 16z/L)^1/2. As EC is a reliable method to characterize atmospheric turbulence, u_* and L were provided by the EC method and used in the AG method. The EC method was also used to measure fluxes of momentum, sensible heat (H), latent heat (LE), water vapor, and CO₂. The EC system includes an integrated CO₂/H₂O open-path gas analyzer and 3D sonic anemometer (IRGASON, Campbell Scientific Inc., USA), electronics for synchronizing gas and wind data (EC100, including a temperature thermistor probe and a barometer, Campbell Scientific Inc., USA), and a data logger (CR3000, Campbell Scientific Inc., USA). The IRGASON was mounted at 6.8 m. Data were sampled and recorded at 10 Hz and processed by EddyPro software (version 7.0.8, LiCor, USA) for a 30 min flux averaging interval. Standard data processing methods were used, including spike removal, double axis rotation, linear detrending, humidity correction of sonic temperature, WPL correction and spectral correction.⁵² Quality checks of raw data and flux results were also performed. The footprint was estimated by the Kljun et al.⁵³ method.

Ozone deposition velocity can be determined as follows:

where v_d is the ozone deposition velocity (cm s⁻¹), and [O₃] is the average concentration of ozone between two measurement heights (ppbv).

2.2 Resistance analogy

The resistance analogy method describes the mechanism and kinetics of ozone deposition. It also provides a numerical model representation of v_d. By considering that the reciprocal of v_d is the total resistance to mass transfer, the influence of the above mentioned three processes on v_d can be expressed in terms of the following electrical analogy:

R _c consists of the resistances of these different surfaces, such as soil, leaf cuticular or ice. As the surface at the study area of NMC site is bare soil, R_c in this study is equal to the soil resistance R_soil. R_a and R_b are calculated as follows:

R_b(O₃) = 2(κu_*)⁻¹(S_c/P_r)^2/3 (7)

where z is the reference height (mean ozone measurement height, m), z₀ is the roughness length for momentum (m), S_c is the Schmidt number, and P_r is the Prandtl number for air (0.72).

2.3 Calculation of surface parameters

Temperature and moisture at the surface, T_surf and RH_surf, instead of atmospheric or edaphic parameters, are recommended as measures of temperature and moisture conditions of soil uptake of ozone by Stella et al.³⁴ and have shown less variability than other measures. Thus, the same scheme is used in this study following the calculation method in Stella et al.:³²

χ_{H2O, surf} = E(R_a(z) + R_b,H2O) + X_{H2O, a} (9)

where ρ is the air density (kg m⁻³), C_p is the air specific heat (J kg⁻¹ K⁻¹), χ_H2O,surf and X_H2O,a are the air concentration of water (g m⁻³) at the surface and reference height, respectively, X_H2O,a is calculated from the air relative humidity, E is the water vapor flux (kg m⁻² s⁻¹), P_vap,surf is the water vapor pressure at the surface (Pa), R is the universal gas constant (J mol⁻¹ K⁻¹), M_H2O is the molecular weight of water (g mol⁻¹), P_sat(T_surf) is the saturation vapor pressure at T_surf (Pa), P is the atmospheric pressure (Pa), γ is the latent heat of vaporization of water (J kg⁻¹), and T₀ is the boiling temperature of water at P (°C).

2.4 Data quality assurance

The use of the AG method requires some prerequisites, such as stationary but strong enough turbulence, no chemical interference and uniform underlying surface.

First, the AG method under extremely unstable and stagnant conditions were thought to suffer from large uncertainties due to the invalidation of MOST. The ratio of the measurement height and the Obukhov length (z/L) is used to evaluate the atmospheric stability. z/L between −2 and 1 are required. Valid data coverage was 79% with a 30 min time resolution.

Second, the NO titration of ozone, which might cause a substantial surface gradient of ozone, is often regarded as an uncertainty source in ozone flux measurements in some areas.⁴³ As ozone titration by a considerable NO near the surfaces in agricultural fields could occur in minutes, it might compete with the vertical transport of ozone in determining the surface gradient of ozone. A comparison between the turbulence transport time (τ_trans) and the chemical reaction time (τ_chem) can be used to evaluate the influence of the chemical reaction on the ozone flux. Following the method of Stella et al.,³⁴ the τ_trans can be expressed as the transfer resistance through each layer multiplied by the layer height. The layer height of the quasi-laminar boundary layer, (z₀ − z_0′), is so small that the contribution of the quasi-laminar boundary to τ_trans is negligible. τ_chem is calculated as the lifetime of ozone reacting with NO.

τ_trans = R_a × (z_m − z₀) + R_b × (z₀ − z_0′) = R_a × z_m (13)

where z_m is the measurement height, χ_NO(z_m) is the concentration of NO_PSS (ppbv) at the measurement height. The reaction rate constant (k_r in ppbv⁻¹ s⁻¹) is calculated as k_r = 0.0444 × exp(−1370/(T_a + 273.15)).⁵⁴ NMC site is almost free from interference from anthropogenic emissions due to the extremely low value of NO_PSS with an average of only 58 pptv. τ_chem is calculated to be 10-fold that of τ_trans (Fig. S1a†), suggesting that this chemical perturbation of the ozone flux measurement is negligible at NMC.

Third, the footprint area of the ozone flux measurement shows ca. 3% attribution to the water surface to the west of our measurement site 280 m away (Fig. S1b†). Ozone deposition on the water surface was roughly evaluated based on the turbulence measurement and typical water surface resistance for ozone uptake of 1000 s m⁻¹. The calculated value of v_d is 1/4–1/8 of that on soil. However, considering the contribution of the water surface to the footprint area, corrections of less than 3% for ozone flux measurement are thus abandoned.

The ozone flux is determined from the product of K and (eqn (1)), thus the uncertainty of ozone flux includes the uncertainty of these two factors, σ_K and σ_(d[O3]/dZ). σ_K contains uncertainty from measurements and those arising from the parameterization (eqn (2) and (3)), both of which have larger uncertainty under stable atmospheric conditions, i.e. during nighttime. Quantify of σ_K is tough⁵⁵ and out of our scope. Constant relative uncertainties of 20% and 50% are given to K, which is robust.^55,56 To validate the estimation of σ_K, we compared H and LE derived separately from the EC method and AG method. It was indicated that the diel profiles were similar, and the magnitudes of H and LE were comparable between the two methods (Fig. S1c and d†). The AG method overestimated H by approximately 10%, and biased LE by a constant systematic error of approximately 18 W m⁻² compared to the EC method. Further examination of the temperature and moisture gradients suggested that inconsistencies between different hygrothermometers might account for the divergences.

σ _(d[O3]/dZ) is determined by the uncertainty of the difference of ozone concentration σ_Δ[O3], because measurement of height is accurate. Ozone concentration at two heights were measured by one ozone analyzer and thus σ_Δ[O3] is equal to 2σ_[O3], which supposed to be 0.35 ppbv. Whether the ratio larger than 1 is used to judge whether Δ[O₃] is significant. 10% of ozone gradient data were insignificant but still retained to avoid a misestimation of the average flux. It is noteworthy that linear interpolation of the 15 min ozone measurement gaps could be an uncertainty source to the ozone gradient but could not be accurately evaluated. Extreme and unreasonable values of the ozone gradient occasionally appeared from the interpolation method. However, it was plausible to see a typical diel variation of the ozone gradients (Fig. S1e†). Therefore, to reduce the impact of outliers on the analysis, outlier spikes of v_d and maximal or minimal 2.5% of R_soil are excluded from the data analysis.

Using Gaussian uncertainty propagation, the uncertainty of ozone flux and ozone deposition velocity could be calculated as follows:

were dependent on during nighttime and dependent on both and during daytime. Diel varies from 25.7% during daytime to 51.1% during nighttime, with an average of 39.8%. As is less than 1%, is approximately equal to .

2.5 Simulation of ozone deposition velocity with different R_soil parameterizations

The resistance analogy with the Stella scheme (hereafter called RA_SS) is performed to evaluate the reproducibility of v_d. R_a and R_b are calculated by eqn (6) and (7) with meteorological parameters measured on site. The expression of the Stella scheme is as follows:

R_soil = R_{soil min} × e^(k×RH_surf) (17)

R_{soil min} = 702 × (clay content)^−0.98 (18)

k = 0.0118 e^{0.0266×(clay content)} (19)

where R_{soil min} (s m⁻¹) is the soil resistance when RH_surf is 0%, k (unitless) is the coefficient of the exponential function, and clay content (%) is the fraction of clay content in the topsoil. The soil clay content of NMC site, 14.5%, is obtained from the soil dataset by Ito and Wagai.⁵⁷

3 Results and discussion

3.1 Environmental conditions and ozone deposition velocity

The environmental conditions of ozone deposition are largely determined by climate regions. As a typical highland mountain climate region, NMC site features strong solar radiation and cold, arid, and windy meteorological conditions. The measurements covered two periods with distinct meteorological parameters, namely, the premonsoon period before 1 July and the ASM period after 1 July (shaded area in Fig. S2†). In the premonsoon period, solar radiation (R_g) at NMC site was even larger than the solar constant (1367 W m⁻²) at cloudless noon (Fig. S2a†), with a maximum of 1552.6 W m⁻² on 11 June. However, with such strong R_g, air temperature (T_a) is quite low due to the buffering effect of plateau permafrost and could be below 0 °C in early May (Fig. S2b†). T_a slowly increased from 4.3 ± 4.3 °C to 11.8 ± 4.0 °C in the premonsoon period. Under the additive effects of low soil water content (SWC) and strong R_g, typical air relative humidity (RH_a) was lower than 25% during the daytime in the premonsoon period (Fig. S2c†). Strong winds occurred frequently, and the wind speed (WS) exceeded 10 m s⁻¹ on some days (Fig. S2d†). In the ASM period, the ASM brought abundant moisture to the TP and resulted in more frequent precipitation. R_g showed a dramatic decline compared to the premonsoon period, but it still remained at a high level compared with regions of the same latitude. Volumetric soil water content (SWC) zoomed from 0.01 to 0.2 after 1 July. Weakened R_g and enhanced SWC in the ASM period caused a slight decrease in T_a and a significant increase in RH_a. WS showed obvious differences from the premonsoon period, with the peak dropping, but WS mean at night increased.

Diel pattern of key parameters that related to calculation of R_a, R_b, R_soil and derivation of v_d according to eqn (6)–(12) and (17) are performed in Fig. 2. Over the entire observation period, the hourly mean T_a ranged from 1.6 ± 4.6 °C at night to 10.6 ± 4.3 °C in the daytime (Fig. 2b). Significant enhancements of T_surf compared to T_a occurred with a maximal hourly enhancement of 15.6 °C during daytime. This was caused by strong R_g, of which the hourly mean maximal reached 1252.4 ± 348.5 W m⁻² (Fig. 2a and b). The enhancement of T_surf further caused lower RH_surf according to eqn (12). RH_surf at noon was as low as 23.2 ± 17.8%, which is 15.3% of decrease compared with RH_a (Fig. 2c). The increased T_surf and consequently decreased RH_surf may result in a small R_soil, and thus favors ozone deposition.

Fig. 2 Diel pattern of meteorological parameters and edaphic parameters associated with ozone deposition velocity derivation and deposition resistance derivation. Diel pattern of (a) solar radiation (R_g), (b) air temperature at 1.8 m (T_a), surface temperature (T_surf) and soil temperature at −5 cm (T_soil), (c) air relative humidity at 1.8 m (RH_a), surface relative humidity (RH_surf) and soil water content (SWC), (d) wind speed at 1.8 m (WS) and wind direction (WD), (e) friction velocity (u_*), (f) atmospheric stability criterion (z/L), (g) turbulence exchange coefficient (K), (h) gradient of ozone between 6.8 m and 1.8 m (d[O₃]/dZ), ozone concentration at 6.8 m and ozone concentration at 1.8 m, (i) ozone flux (F), and (j) ozone deposition velocity (v_d). The solid line is the diel mean, and the shading is variation of our observations.

Solar radiation and wind supply energy to atmospheric turbulence which affects R_a and R_b. A campaign average R_g of 411.9 ± 469.2 W m⁻² and wind speed of 3.83 ± 0.72 m s⁻¹ (Fig. 2a and d) favored strong atmospheric turbulence. The influence of solar radiation and wind on atmospheric turbulence strength is reflected in the diel pattern of friction velocity (u_*) and z/L (Fig. 2e and f). u_* remained fairly large throughout the day with a maximum of 0.43 ± 0.19 m s⁻¹ around noon. The mean u_* during the night dropped to 0.2 m s⁻¹. An opposite diel trend was shown in z/L, with an average of 0.1 at night and −0.5 in the daytime. A nighttime z/L of 0.1 also confirmed that the atmospheric stability at night was nearly neutral and did not seriously block the development of turbulence. Active atmospheric turbulence resulted in rapid vertical turbulence transport, which was characterized by K (eqn (2), (3), and Fig. 2g). The diel pattern of K is similar to those of WS and R_g. Nighttime K remained at approximately 0.3–0.4 m² s⁻¹, while K increased to 1.36 ± 0.24 m² s⁻¹ at noon. It can be inferred from the diel pattern of K and R_g that surface heating driven by solar radiation plays a dominant role in turbulence development at NMC site.

F and v_d are determined based on the K, as well as the measurements of d[O₃]/dZ between 6.8 m and 1.8 m (eqn (1) and (4)). The ozone measured at the two heights showed similar diel profiles but distinct abundances (Fig. 2h). The ozone at 1.8 m ranged from 45.6 ± 11.4 ppbv at night to 67.4 ± 10.4 ppbv in the daytime, with a mean of 56.9 ± 8.6 ppbv. The ozone at 6.8 m ranged from 50.7 ± 12.1 ppbv at night to 69.0 ± 10.3 ppbv in the daytime, with a mean of 60.2 ± 7.0 ppbv. A similar diel profile of F was derived compared to that of K, WS, and R_g (Fig. 2i). Negative values suggested a downward deposition of ozone. The daily mean F was −7.09 ± 2.36 nmol m² s⁻¹, with a noontime maximum of −11.78 ± 10.28 nmol m² s⁻¹. As the low air pressure counteracts high volume concentrations of ozone to F, the large F could be explained mainly by high v_d. v_d showed a bridge-shaped diel profile (Fig. 2j) with the mean of 0.49 ± 0.11 cm s⁻¹ and a noontime peak of 0.73 ± 0.67 cm s⁻¹. The v_d observed here is comparable to typical v_d in forests and one of the highest previously reported over bare soils.^33,58,59 Notably, a high nighttime v_d of approximately 0.4 cm s⁻¹ is observed, which is among the top values ever reported.³³ Although nighttime v_d has larger uncertainty than that of daytime, quite high u_* and z/L much less than 1 during nighttime favored high nighttime v_d at NMC.

3.2 Determinative role of R_soil in ozone deposition and its variability

The constitution of deposition resistance provides the information to explain the magnitude of v_d and its diel variation. The diel profile of deposition resistance and its components is shown in Fig. 3a. The deposition resistance of ozone, R_all, showed a typical U-shaped diel profile, which accounts for the bridge-shaped v_d. Both R_soil and R_a + R_b showed a typical U-shaped diel profile as well. The daytime low R_soil at NMC site (106.0 ± 22.4 s m⁻¹) was even comparable to the typical daytime R_c over forests (Fig. 3c), which was the lowest resistance due to effective stomatal uptake of ozone by leaves. The nighttime R_soil, with a mean of 159.5 ± 30.7 s m⁻¹, was approximately 0.5 times higher than the daytime R_soil, but still remained the lowest in literature.³³ Such diel structure of R_soil at NMC site was also worthy of highlighting. In comparison, the nighttime R_c on forest canopies was generally 1.5 times (sometimes up to 10 times) higher than the daytime R_c due to the closure of leaf stomata.²⁷ Mechanism exploration and parameterization of R_soil will be discussed in next section. A comparison of daytime R_a + R_b at NMC site (62.3 ± 15.8 s m⁻¹) with previous studies in various environments (56.5 ± 25.7 s m⁻¹) indicated efficient atmospheric turbulent transport at NMC site (Fig. 3b). A much smaller nighttime R_a + R_b here (110.4 ± 15.3 s m⁻¹vs. 200.5 ± 101.0 s m⁻¹ in various environments) implied the more rapid ozone deposition at night as well. R_a + R_b is reasonably well defined and can be expressed by a function of u_* and z/L (eqn (6) and (7)), which are affected by WS and R_g. Strong WS and R_g at NMC site allow neither a high R_a + R_b nor a very low R_a + R_b due to the insensitive response of R_a + R_b to WS at its high value end.

Fig. 3 (a) Median diel pattern of deposition resistance of observation at NMC site. Hourly medians are utilized instead of hourly means due to the skewed distribution of R_all and R_soil. (b) Comparison of R_a + R_b and reported R_a + R_b during day and night. (c) Comparison of observed R_soil and reported R_c over different conditions during day and night. Parts of reported R_c are obtained by 1/v_d minus prescribed R_a + R_b (50 s m⁻¹ during daytime and 200 s m⁻¹ during nighttime). Data from.^{24,26,58–67}

Similar to previous impressions, R_soil is the major fraction (62 ± 7%) of R_all during both the daytime and the nighttime, suggesting R_soil plays the dominant role in ozone deposition at NMC site. The low R_soil and R_a + R_b together explain large v_d in both daytime and nighttime. The spatial variability in R_soil and v_d over different climates across 3 orders of magnitude (Fig. 3c and S3†), whereas the span range of R_soil at NMC site is the smallest and distributed at the lower end of reported R_soil. To date, our study reported the only observation of R_soil in highland climate regions, and the significant differences from previous studies highlighted the heterogeneity of R_soil over different climates.

3.3 Variability of R_soil and its parameterization

In-depth understanding of soil uptake of ozone could help to establish a more robust empirical parameterization of R_soil. Clifton et al.²⁷ proposed a mechanistic model of R_soil, which represents ozone reaction with soil surfaces and gases in soil pore spaces. Although it is tough to measure all of the parameters required for the mechanistic model in a field observation, it provides a roadmap to identify key parameters that control R_soil. Both the mechanistic model and previous observations^34–37 have found that moisture and temperature are two key parameters that control the temporal variability of R_soil.

Moisture is suggested a major influencing factor of R_soil based on both mechanical deduction and observations.^27,34,37 The inhibition of soil ozone uptake by moisture is reflected in two ways.³⁴ First, soil moisture can block the diffusion of ozone in soil. Second, water molecules can be adsorbed by soil and occupy the reactive surface site, thus competing with ozone absorption and reactive uptake. The two mechanisms account for a positive dependence of R_soil on moisture as widely observed in field observations and laboratory experiments.²⁷ Our observation at NMC site also showed a significant positive dependence of R_soil on RH_surf (Fig. 4a). The dry climate at NMC site is conducive to keeping R_soil small.

Fig. 4 Dependence of R_soil on (a) RH_surf and (b) T_surf. The block processing ranges of means and medians of RH_surf and T_surf are 10% and 5 °C, respectively.

Temperature is a positive-going parameter for ozone uptake, as the reaction of ozone on the surface is endothermic. Thus, R_soil decreases with increasing temperature. The mechanical deduction requires an Arrhenius-like response of R_soil to temperature, which is verified in field observations but with large fitting uncertainty and variability.^34–36 The negative relationship between R_soil and temperature was also observed at NMC site (Fig. 4b). Notably, the dependence of R_soil on temperature could be a misconception due to the highly negative correlation between temperature and moisture. As a result, it is difficult to distinguish the individual effect of temperature from RH_surf on R_soil. Nevertheless, high T_surf is conducive to lower R_soil.

As shown in Fig. 5a and b, R_soil can be described by an exponential equation of RH_surf and an Arrhenius-like function of T_surf (eqn (20) and (21)):

R_soil = 71.0 × e^{0.012×RH_surf} (20)

Fig. 5 R _soil as a function of (a) RH_surf and (b) T_surf. Dots are block median data with ranges of 10% for RH_surf and 5 °C for T_surf. (c) Comparison of observed v_d and simulated v_d using R_soil as a function of RH_surf and T_surf with input meteorological data from observations. The shading area for observation represents 1 sd of observed v_d. Time alignments are performed on the data used for comparison.

These two parameterization schemes were tested by simulating v_d with observation-constrained meteorological data by employing eqn (5)–(7) (Fig. 5c). Similar results of v_d are obtained from the two parameterization schemes and both results are closely comparable to the observed v_d in terms of both the magnitude and the temporal variability. Stella et al.³⁴ have found that the parameterization scheme with RH_surf performed more robustly than the parameterization scheme with T_surf. A potential reason for this could be that RH_surf itself contains the influence of T_surf due to the high correlation between RH_surf and T_surf, as previously mentioned. Herein, we choose the parameterization scheme with RH_surf for further discussion. According to eqn (20), R_soil is predicted to be lower than 240 s m⁻¹ at NMC site even at a high RH_surf range, which explains the rapid ozone deposition at night. In addition, strong solar radiation and dry conditions during the daytime maintain high T_surf and low RH_surf, which are also conducive to low R_soil.

Block medians instead of block means of R_soil were chosen to fit the function among R_soil and RH_surf (or T_surf) in this study, since R_soil at NMC was positively skewed. The fitting function based on block medians turned out to better reproduce our observations than that based on block means (Fig. 5 and S4†). As a matter of fact, R_soil is consistently found to be positively skewed.^26,65,67,68 Thus, it needs more attention about whether the function among R_soil and RH_surf (or T_surf) are fitted with block medians or block means. Block means of R_soil were used in the establishment of the Stella scheme,^33,34 which should be further examined.

3.4 Evaluation and improvement of the Stella scheme

RH_surf explains the variability of R_soil at specific sites, while soil clay content determines the site-by-site variability by affecting R_{soil min} and k. According to the fitted results in eqn (20), R_{soil min} and k of NMC are 71.0 s m⁻¹ and 0.012, respectively. We applied the Stella scheme reversely (eqn (17)–(19)) and derived R_{soil min} = 51.0 s m⁻¹ and k = 0.017 at NMC. Compared with the observed R_{soil min} (71.0 s m⁻¹) and k (0.012), the Stella scheme underestimated R_{soil min} by 28% and overestimated k by 42% (red dots in Fig. 6a and b). By including our observations on R_{soil min} and k, we updated the Stella scheme (hereafter called the updated Stella scheme, eqn (17), (20) and (21). To be noted, observations at La Crau were not included fitting in both the Stella scheme and the updated Stella scheme due to the reason mentioned in Stella et al.,³³ where the reported soil clay content is not representative.

R_{soil min} = 661 × (clay content)^−0.86 (22)

k = 0.0093 e^{0.0325×(clay content)} (23)

Fig. 6 Relationships between soil clay content and two coefficients in the Stella scheme, (a) minimum soil resistance (R_{soil min}) and (b) k. The black dots are values collected from previous observations, and the red dots are values from our observation at NMC site. The Stella scheme are shown as black lines, and the updated Stella scheme are shown as red lines. (c) Comparison of observed v_d and simulated v_d using the RA_SS and resistance analogy with the updated Stella scheme (RA_uSS) with input meteorological data from observations. Time alignments are performed on the data used for comparison.

The updated Stella scheme of course improved the scheme representation of R_soil at NMC, relative to the Stella scheme. The simulation error of R_{soil min} and k was reduced from 28% and 42% to 7% and 25%. A better representative R_soil led to a better representation of v_d (Fig. 6c). The resistance analogy with the updated Stella scheme (RA_uSS) reproduced the observed daytime mean of v_d of 0.58 cm s⁻¹. Almost the same diel pattern of v_d was modelled by the RA_SS and overestimated daytime v_d by only 14%. Our observations validate the feasibility of the Stella scheme and illustrate the need to include more observations in the Stella scheme.

The above analysis suggests that v_d is sensitive to both R_{soil min} and k. Assuming ±25% spread in R_{soil min} and k, the spread in daytime and nighttime v_d are calculated by the RA_uSS with typical RH_surf (40% during daytime and 80% during nighttime) and R_a + R_b (50 s m⁻¹ during daytime and 200 s m⁻¹ during nighttime). v_d generally varied in a narrow range responding to R_{soil min} changes across the entire soil clay content range (Fig. 7a). The sensitivity of v_d to k increases exponentially with increasing clay content (Fig. 7b). In the high soil clay content range, the ±25% spread in k could cause more than 200% change in v_d. The sensitivity of v_d to k also increases with RH_surf. As a result, nighttime v_d is more sensitive than daytime v_d for a much higher RH_surf range over the night. Therefore, observation verification on high soil clay contents and high RH_surf are especially useful, though all observation verifications on varied climatic environments worthy to be highlighted, for the sake of continuous test and improvement of the updated Stella scheme.

Fig. 7 Sensitivity test of daytime and nighttime v_d responding to a ±25% spread (−25% and +25% change are shown in solid lines and dash lines, respectively) in (a) R_{soil min} or (b) k as calculated by the updated Stella scheme with typical RH_surf (40% during daytime and 80% during nighttime) and R_a + R_b (50 s m⁻¹ during daytime and 200 s m⁻¹ during nighttime).

3.5 Implications of the updated Stella scheme

The updated Stella scheme provides a comprehensive view of the distributions of R_soil and v_d over soil in various soil and hydrothermal conditions (Fig. 8). The rough range of soil clay content and hydrothermal conditions in different climates are also performed in Fig. 8a and b. A commonly used R_soil in CTMs (black solid line in Fig. 8a) overestimates R_soil in most climate regions and thus causes the underestimation of v_d over soil (black solid line in Fig. 8b). And the great variability of R_soil and v_d over soil among different climate regions are not reproduced.

Fig. 8 Dependence of R_soil and v_d over soil on soil clay content and hydrothermal conditions. Theoretical distribution of (a) R_soil and (b) v_d over soil with different soil clay contents and RH_surf. R_soil is calculated with the updated Stella scheme. v_d is the mean of daytime and nighttime v_d. The daytime and nighttime v_d are calculated with the assumed R_a + R_b (50 s m⁻¹ during the daytime and 200 s m⁻¹ during the nighttime) and calculated R_soil. The solid black line in (a) represents a commonly used R_soil in CTMs (500 s m⁻¹) and in (b) represents the v_d calculated with R_soil = 500 s m⁻¹. Data of climate classification and soil clay content are the same with Fig. 1. The HadISDH data^69,70 is used to estimate the approximate RH_surf ranges of different climates.

The delineation of different climate regions in Fig. 8 is helpful to establish a general impression of the distribution of R_soil and v_d over soil in real environments. In desert, coarse soil could cause high R_{soil min} and R_soil, and thus low v_d. As for tropical forest climate regions, high RH_surf causes extremely high R_soil and low v_d over soil, thus deposition on canopy would be the major path. In vast areas of other climates, moderate soil clay content and RH_surf supply low R_soil and therefore high v_d over soil, especially in highland climate regions and steppe. Our observation in fact provides the confirmation of high v_d over soil in highland climate regions. Stella et al.³³ provides five observations in the subtropical climate region. More observations in different climate regions, especially in desert, steppe, and highland are needed to evaluate and improve the Stella scheme or the updated Stella scheme.

4 Conclusions and perspectives

This study reported the first measurement of ozone deposition in a typical highland background site with bare soil in the central TP. Unexpectedly high v_d with a daily mean of 0.49 ± 0.11 cm s⁻¹ and an hourly mean maximum of 0.73 ± 0.67 cm s⁻¹ was found. Strong solar radiation and high wind speed provided efficient turbulence energy, especially in the daytime, suggesting that atmospheric transports to the surface were not limiting steps most of the time. The soil uptake process was the main controlling process of ozone deposition at NMC. Soil uptake rate, as well as 1/R_soil, was found among the top in the literature. Such high 1/R_soil was regulated by proper soil clay content, dry conditions, and strong solar radiation in the TP. These favorable conditions of soil ozone deposition are common in the TP and other highland regions. Therefore, ozone deposition in the TP and highland regions was validated as being underestimated by current models.

The Stella scheme is proven to be an effective parameterization of R_soil by our verification at NMC site. The Stella scheme was updated through the inclusion of our observations and then better represented R_soil and thus v_d at NMC. The updated Stella scheme is recommended as a replacement for the prescribed values of R_soil in current CTMs. Notably, both the Stella scheme and the updated Stella scheme showed the potential for continuous improvement. More quality and representative field observations of R_soil are encouraged to further minimize the fitting uncertainty of R_{soil min} and k in the Stella scheme. Observation verifications that meet the following requirements will be helpful. First, soil deposition should be isolated from other paths of ozone deposition, such as observations over bare soils or chambers in the ground. Second, to avoid interference of reactive gases, pristine bare soils or inactivated soils are recommended. Third, perform measurements under different climatic environments to obtain different combinations of soil clay content and hydrothermal conditions. Finally, long-term and even year-round observations should be conducted to obtain a wider range of variations in the combination of RH_surf and T_surf.

Author contributions

Chong Zhang: data curation, formal analysis, investigation, methodology, software, visualization, writing – original draft, writing – review & editing; Jianshu Wang: investigation, writing – review & editing; Yingjie Zhang: conceptualization, funding acquisition, investigation, methodology, project administration, writing – review & editing; Wanyun Xu: data curation, investigation, writing – review & editing; Gen Zhang: investigation, writing – review & editing; Guofang Miao: software, writing – review & editing; Jiacheng Zhou: investigation, writing – review & editing; Hui Yu: investigation, writing – review & editing; Weixiong Zhao: writing – review & editing; Weili Lin: writing – review & editing; Ling Kang: resources, writing – review & editing; Xuhui Cai: writing – review & editing; Hongsheng Zhang: writing – review & editing; Chunxiang Ye: conceptualization, funding acquisition, investigation, methodology, project administration, resources, supervision, writing – review & editing.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work was supported by the Second Tibetan Plateau Scientific Expedition and Research Program (2019QZKK060604) and the National Natural Science Foundation of China (grants no. 42105110).The authors acknowledge all researchers involved in the @Tibet 2019 campaign. Thanks for the support from the Nam Co Comprehensive Observation and Research Station of Chinese Academy of Sciences.

References

1. E. A. Ainsworth, C. R. Yendrek, S. Sitch, W. J. Collins and L. D. Emberson, Annu. Rev. Plant Biol., 2012, 63, 637–661.
2. T.-M. Fu and H. Tian, Curr. Pollut. Rep., 2019, 5, 159–171.
3. N. E. Grulke and R. L. Heath, Plant Biol., 2020, 22, 12–37.
4. M. Jerrett, R. T. Burnett, C. Arden Pope, K. Ito, G. Thurston, D. Krewski, Y. Shi, E. Calle and M. Thun, N. Engl. J. Med., 2009, 360, 1085–1095.
5. M. G. Schultz, S. Schröder, O. Lyapina, O. R. Cooper, I. Galbally, I. Petropavlovskikh, E. von Schneidemesser, H. Tanimoto, Y. Elshorbany, M. Naja, R. J. Seguel, U. Dauert, P. Eckhardt, S. Feigenspan, M. Fiebig, A.-G. Hjellbrekke, Y.-D. Hong, P. C. Kjeld, H. Koide, G. Lear, D. Tarasick, M. Ueno, M. Wallasch, D. Baumgardner, M.-T. Chuang, R. Gillett, M. Lee, S. Molloy, R. Moolla, T. Wang, K. Sharps, J. A. Adame, G. Ancellet, F. Apadula, P. Artaxo, M. E. Barlasina, M. Bogucka, P. Bonasoni, L. Chang, A. Colomb, E. Cuevas-Agulló, M. Cupeiro, A. Degorska, A. Ding, M. Fröhlich, M. Frolova, H. Gadhavi, F. Gheusi, S. Gilge, M. Y. Gonzalez, V. Gros, S. H. Hamad, D. Helmig, D. Henriques, O. Hermansen, R. Holla, J. Hueber, U. Im, D. A. Jaffe, N. Komala, D. Kubistin, K.-S. Lam, T. Laurila, H. Lee, I. Levy, C. Mazzoleni, L. R. Mazzoleni, A. McClure-Begley, M. Mohamad, M. Murovec, M. Navarro-Comas, F. Nicodim, D. Parrish, K. A. Read, N. Reid, L. Ries, P. Saxena, J. J. Schwab, Y. Scorgie, I. Senik, P. Simmonds, V. Sinha, A. I. Skorokhod, G. Spain, W. Spangl, R. Spoor, S. R. Springston, K. Steer, M. Steinbacher, E. Suharguniyawan, P. Torre, T. Trickl, L. Weili, R. Weller, X. Xiaobin, L. Xue and M. Zhiqiang, Elem. Sci. Anth., 2017, 5, 58.
6. A. T. Archibald, J. L. Neu, Y. F. Elshorbany, O. R. Cooper, P. J. Young, H. Akiyoshi, R. A. Cox, M. Coyle, R. G. Derwent, M. Deushi, A. Finco, G. J. Frost, I. E. Galbally, G. Gerosa, C. Granier, P. T. Griffiths, R. Hossaini, L. Hu, P. Jöckel, B. Josse, M. Y. Lin, M. Mertens, O. Morgenstern, M. Naja, V. Naik, S. Oltmans, D. A. Plummer, L. E. Revell, A. Saiz-Lopez, P. Saxena, Y. M. Shin, I. Shahid, D. Shallcross, S. Tilmes, T. Trickl, T. J. Wallington, T. Wang, H. M. Worden and G. Zeng, Elementa, 2020, 8, 1–53.
7. O. E. Clifton, F. Paulot, A. M. Fiore, L. W. Horowitz, G. Correa, C. B. Baublitz, S. Fares, I. Goded, A. H. Goldstein, C. Gruening, A. J. Hogg, B. Loubet, I. Mammarella, J. W. Munger, L. Neil, P. Stella, J. Uddling, T. Vesala and E. Weng, J. Geophys. Res.: Atmos., 2020, 125, e2020JD032398.
8. A. Y. H. Wong, J. A. Geddes, A. P. K. Tai and S. J. Silva, Atmos. Chem. Phys., 2019, 19, 14365–14385.
9. D. Simpson, Atmos. Environ., Part A, 1992, 26, 1609–1634.
10. J.-T. Lin, D. Youn, X.-Z. Liang and D. J. Wuebbles, Atmos. Environ., 2008, 42, 8470–8483.
11. S. Solberg, Ø. Hov, A. Søvde, I. S. A. Isaksen, P. Coddeville, H. De Backer, C. Forster, Y. Orsolini and K. Uhse, J. Geophys. Res.: Atmos., 2008, 113(D7) DOI:10.1029/2007JD009098.
12. M. L. Wesely, Atmos. Environ., 1989, 23, 1293–1304.
13. T. Emmerichs, A. Kerkweg, H. Ouwersloot, S. Fares, I. Mammarella and D. Taraborrelli, Geosci. Model Dev., 2021, 14, 495–519.
14. S. Falk and A. Søvde Haslerud, Geosci. Model Dev., 2019, 12, 4705–4728.
15. I. Lagzi, R. Mészáros, L. Horváth, A. Tomlin, T. Weidinger, T. Turányi, F. Ács and L. Haszpra, Atmos. Environ., 2004, 38, 6211–6222.
16. R. J. Park, S. K. Hong, H.-A. Kwon, S. Kim, A. Guenther, J.-H. Woo and C. P. Loughner, Atmos. Chem. Phys., 2014, 14, 7929–7940.
17. L. Zhang, J. R. Brook and R. Vet, Atmos. Chem. Phys., 2003, 3, 2067–2082.
18. S. Galmarini, P. Makar, O. E. Clifton, C. Hogrefe, J. O. Bash, R. Bellasio, R. Bianconi, J. Bieser, T. Butler, J. Ducker, J. Flemming, A. Hodzic, C. D. Holmes, I. Kioutsioukis, R. Kranenburg, A. Lupascu, J. L. Perez-Camanyo, J. Pleim, Y.-H. Ryu, R. San Jose, D. Schwede, S. Silva and R. Wolke, Atmos. Chem. Phys., 2021, 21, 15663–15697.
19. Z. Wu, D. B. Schwede, R. Vet, J. T. Walker, M. Shaw, R. Staebler and L. Zhang, J. Adv. Model. Earth Syst., 2018, 10, 1571–1586.
20. D. Schwede, L. Zhang, R. Vet and G. Lear, Atmos. Environ., 2011, 45, 1337–1346.
21. E. Lamaud, A. Carrara, Y. Brunet, A. Lopez and A. Druilhet, Atmos. Environ., 2002, 36, 77–88.
22. K. F. Zeller and N. T. Nikolov, Environ. Pollut., 2000, 107, 1–20.
23. S. Fares, F. Savi, J. Muller, G. Matteucci and E. Paoletti, Agric. For. Meteorol., 2014, 198–199, 181–191.
24. K. Matsuda, I. Watanabe, V. Wingpud, P. Theramongkol, P. Khummongkol, S. Wangwongwatana and T. Totsuka, Atmos. Environ., 2005, 39, 2571–2577.
25. A. J. Visser, L. N. Ganzeveld, I. Goded, M. C. Krol, I. Mammarella, G. Manca and K. F. Boersma, Atmos. Chem. Phys., 2021, 21, 18393–18411.
26. M. Coyle, E. Nemitz, R. Storeton-West, D. Fowler and J. N. Cape, Agric. For. Meteorol., 2009, 149, 655–666.
27. O. E. Clifton, A. M. Fiore, W. J. Massman, C. B. Baublitz, M. Coyle, L. Emberson, S. Fares, D. K. Farmer, P. Gentine, G. Gerosa, A. B. Guenther, D. Helmig, D. L. Lombardozzi, J. W. Munger, E. G. Patton, S. E. Pusede, D. B. Schwede, S. J. Silva, M. Sörgel, A. L. Steiner and A. P. K. Tai, Rev. Geophys., 2020, 58, e2019RG000670.
28. L. Horváth, P. Koncz, A. Móring, Z. Nagy, K. Pintér and T. Weidinger, Bound. -Layer Meteorol., 2017, 167, 303–326.
29. P. Stella, E. Personne, E. Lamaud, B. Loubet, I. Trebs and P. Cellier, J. Geophys. Res.: Biogeosci., 2013, 118, 1120–1132.
30. E. Lamaud, B. Loubet, M. Irvine, P. Stella, E. Personne and P. Cellier, Agric. For. Meteorol., 2009, 149, 1385–1396.
31. S. J. Silva and C. L. Heald, J. Geophys. Res.: Atmos., 2018, 123, 559–573.
32. P. Stella, E. Personne, B. Loubet, E. Lamaud, E. Ceschia, P. Béziat, J. M. Bonnefond, M. Irvine, P. Keravec, N. Mascher and P. Cellier, Biogeosciences, 2011, 8, 2869–2886.
33. P. Stella, B. Loubet, C. de Berranger, X. Charrier, E. Ceschia, G. Gerosa, A. Finco, E. Lamaud, D. Serça, C. George and R. Ciuraru, Atmos. Environ., 2019, 199, 202–209.
34. P. Stella, B. Loubet, E. Lamaud, P. Laville and P. Cellier, Agric. For. Meteorol., 2011, 151, 669–681.
35. J. N. Cape, R. Hamilton and M. R. Heal, Atmos. Environ., 2009, 43, 1116–1123.
36. I. Fumagalli, C. Gruening, R. Marzuoli, S. Cieslik and G. Gerosa, Agric. For. Meteorol., 2016, 228–229, 205–216.
37. W. J. Massman, Atmos. Environ., 2004, 38, 2323–2337.
38. S. Bassin, P. Calanca, T. Weidinger, G. Gerosa and J. Fuhrer, Atmos. Environ., 2004, 38, 2349–2359.
39. D. Chen and H. W. Chen, Environ. Dev., 2013, 6, 69–79.
40. FAO and IIASA, Harmonized World Soil Database version 2.0, 2023.
41. M. Kummu, M. Heino, M. Taka, O. Varis and D. Viviroli, One Earth, 2021, 4, 720–729.
42. P. Stella, B. Loubet, P. Laville, E. Lamaud, M. Cazaunau, S. Laufs, F. Bernard, B. Grosselin, N. Mascher, R. Kurtenbach, A. Mellouki, J. Kleffmann and P. Cellier, Atmos. Meas. Tech., 2012, 5, 1241–1257.
43. R. M. Vuolo, B. Loubet, N. Mascher, J. C. Gueudet, B. Durand, P. Laville, O. Zurfluh, R. Ciuraru, P. Stella and I. Trebs, Biogeosciences, 2017, 14, 2225–2244.
44. J. Wang, Y. Zhang, C. Zhang, Y. Wang, J. Zhou, L. K. Whalley, E. J. Slater, J. E. Dyson, W. Xu, P. Cheng, B. Han, L. Wang, X. Yu, Y. Wang, R. Woodward-Massey, W. Lin, W. Zhao, L. Zeng, Z. Ma, D. E. Heard and C. Ye, Environ. Sci. Technol., 2023, 57, 5474–5484.
45. X. Xu, H. Zhang, W. Lin, Y. Wang, W. Xu and S. Jia, Atmos. Chem. Phys., 2018, 18, 5199–5217.
46. X. Yin, S. Kang, B. de Foy, Z. Cong, J. Luo, L. Zhang, Y. Ma, G. Zhang, D. Rupakheti and Q. Zhang, Atmos. Chem. Phys., 2017, 17, 11293–11311.
47. W. Lin, X. Xu, X. Zheng, J. Dawa, C. Baima and J. Ma, J. Environ. Sci., 2015, 31, 133–145.
48. J. Huang, S. Kang, Q. Zhang, H. Yan, J. Guo, M. G. Jenkins, G. Zhang and K. Wang, Atmos. Environ., 2012, 62, 540–550.
49. L. Wang, C. Yi, X. Xu, B. Schütt, K. Liu and L. Zhou, J. Mt. Sci., 2009, 6, 354–361.
50. B. Fang, W. Zhao, X. Xu, J. Zhou, X. Ma, S. Wang, W. Zhang, D. S. Venables and W. Chen, Opt. Express, 2017, 25, 26910.
51. Z. Y. Wu, L. Zhang, X. M. Wang and J. W. Munger, Atmos. Chem. Phys., 2015, 15, 7487–7496.
52. M. Mauder, M. Cuntz, C. Drüe, A. Graf, C. Rebmann, H. P. Schmid, M. Schmidt and R. Steinbrecher, Agric. For. Meteorol., 2013, 169, 122–135.
53. N. Kljun, P. Calanca, M. W. Rotach and H. P. Schmid, Bound. -Layer Meteorol., 2004, 112, 503–523.
54. S. Walton, M. W. Gallagher and J. H. Duyzer, Atmos. Environ., 1997, 31, 2915–2931.
55. V. Wolff, I. Trebs, T. Foken and F. X. Meixner, Biogeosciences, 2010, 7, 1729–1744.
56. J. Rinne, J.-P. Tuovinen, T. Laurila, H. Hakola, M. Aurela and H. Hypén, Agric. For. Meteorol., 2000, 102, 25–37.
57. A. Ito and R. Wagai, Sci. Data, 2017, 4, 170103.
58. S. Cieslik, Environ. Pollut., 2009, 157, 1487–1496.
59. R. Wada, S. Yonemura, A. Tani and M. Kajino, J. Agric. Meteorol., 2023, 79, 38–48.
60. D. Fowler, C. Flechard, J. N. Cape, R. L. Storeton-West and M. Coyle, Water, Air, Soil Pollut., 2001, 130, 63–74.
61. G. Gerosa, R. Marzuoli, S. Cieslik and A. Ballarin-Denti, Atmos. Environ., 2004, 38, 2421–2432.
62. T. P. Meyers, P. Finkelstein, J. Clarke, T. G. Ellestad and P. F. Sims, J. Geophys. Res.: Atmos., 1998, 103, 22645–22661.
63. M. Michou, P. Laville, D. Serça, A. Fotiadi and V.-H. Peuch, Atmos. Res., 2005, 74, 89–116.
64. T. N. Mikkelsen, H. Ro-Poulsen, M. F. Hovmand, N. O. Jensen, K. Pilegaard and A. H. Egeløv, Atmos. Environ., 2004, 38, 2361–2371.
65. C. A. Pio, M. S. Feliciano, A. T. Vermeulen and E. C. Sousa, Atmos. Environ., 2000, 34, 195–205.
66. M. Zapletal, P. Cudlín, P. Chroust, O. Urban, R. Pokorný, M. Edwards-Jonášová, R. Czerný, D. Janouš, K. Taufarová, Z. Večeřa, P. Mikuška and E. Paoletti, Environ. Pollut., 2011, 159, 1024–1034.
67. L. Zhang, J. R. Brook and R. Vet, Atmos. Environ., 2002, 36, 4787–4799.
68. L. Zhang, M. D. Moran, P. A. Makar, J. R. Brook and S. Gong, Atmos. Environ., 2002, 36, 537–560.
69. K. M. Willett, R. J. H. Dunn, P. W. Thorne, S. Bell, M. de Podesta, D. E. Parker, P. D. Jones and C. N. Williams Jr, Clim. Past, 2014, 10, 1983–2006.
70. A. Smith, N. Lott and R. Vose, Bull. Am. Meteorol. Soc., 2011, 92, 704–708.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3ea00153a

‡ School of Ecology and Nature Conservation, Beijing Forestry University, Beijing, China.

---