The experiments were carried out at the Program for Research on Oxidants: PHotochemistry, Emissions, and Transport (PROPHET) research site at UMBS, which occupies about 10 000 acres on the northern tip of the Lower Michigan Peninsula (45.56∘ N, 84.71∘ W; Fig. 1). The area was heavily logged until the end of the 19th century. It also experienced several severe wildfires from 1880 to 1920. Natural reforestation started when the location was acquired for the research station in 1909. Today, bigtooth aspen, trembling aspen (Populus tremuloides), red maple, red oak, and white pine dominate within about a 1 km radius of the PROPHET site, whereas within a 60 m radius of the site, there are more white pine trees and almost no trembling aspen.

The northern part of the peninsula is fairly remote. The air is free from anthropogenic pollutants unless meteorological conditions result in the advection of air masses from surrounding major cities: to the southwest, Chicago, IL, and Milwaukee, WI; to the southeast, Detroit, MI; and to the east, Toronto, ON. During the field experiments, about 35 % of the time, air masses were coming from these directions. However, since the enclosures were purged with scrubbed ambient air (see Methods section), the direct influence of pollutants on the enclosed plant material was minimal.

The enclosure measurements were carried out from 20 July to 3 August 2016. Sky conditions were sunny to mixed sun and clouds most of the time. The average ambient temperature measured in the canopy (∼2 m from the ground) was 24 ∘C during daylight and 18 ∘C at night, with maximum and minimum temperatures of 31 and 12 ∘C, respectively. There were two main rainfall events, with the most recent being 2 d prior to the start of the enclosure experiments. The average soil temperature was near 19 ∘C throughout the experiment period, and the soil moisture decreased gradually after the rainfall. These conditions are within the normal ranges for this site in July.

Branch enclosure experiments were conducted sequentially on branches of white pine, red maple, bigtooth aspen, and red oak. The estimated ages of the white pine, red maple, and red oak trees were about 15–20 years, and the bigtooth aspen, 5–10 years. All tree branches were selected based on their sun exposure, accessibility, and size. The height of the enclosed branches ranged from 3 to 10 m above the ground.

The enclosure system was composed of three parts: the enclosures, the airflow system, and the trace gas measurement instruments (Fig. 2). The enclosure, essentially a flow chamber, was constructed using a 61×91 cm bag made of Tedlar^® (polyvinyl fluoride) (Jensen Inert Products, Florida, USA) with three factory-installed 0.95 cm diameter ports to attach tubing and sensor wires. The branch was carefully enclosed by the bag so that it was situated as close to the middle of the bag as possible. The open end of the bag was then closed around and tied onto the main stem of the branch, tight enough to secure the enclosure when it was inflated by the purge air, but also with enough leakage to allow air to escape during purging. Each branch enclosure was paired with an identical enclosure assembly without any plant material as the background reference to account for wall effects and other factors that may affect trace gas concentrations. The reference enclosure was placed adjacent to the branch enclosure but without obstructing the sunlight to the enclosed tree leaves.

Air delivery and air sample lines (polytetrafluoroethylene or PTFE), each about 30 m long, were connected to the enclosures and to the instruments housed in an air-conditioned trailer at the site. Between the trailer and the enclosures, the air and sample lines were bundled together and sheathed inside black flexible insulation hoses linked together end to end. The hoses were wrapped in aluminum foil to keep the sample lines from absorbing heat from sunlight.

Ambient air from outside the trailer, scrubbed free of dust, O3, and NOx, was used as the purge gas. An oil-free air compressor (Medo USA, now Nitto Kohki USA) was used to pull the ambient air through an organic vapor/acid gas respirator cartridge (Magid, Illinois, USA), which functioned as a dust filter. Downstream of the compressor, the air was further filtered by an ozone scrubber (Thermo Fisher Scientific), activated charcoal, and a NOx scrubber (Purafil, Inc., Georgia, USA). After the filters, polytetrafluoroethylene (PTFE) tubing (12.7 and 9.52 mm outside diameter – OD) was used to carry the air to the enclosure chambers. The tubing was connected to the port on the Tedlar bag at the end near the tip of the enclosed branch, opposite the bag opening. Inside the bag connected to the same port was an air distributor made from a loop of tubing (9.52 mm OD) with pinholes (∼1.5 mm diameter) about 1 cm apart drilled along its entire length. This allowed even distribution and mixing of the purge air inside the enclosure. The flow rate of the purge air was maintained at 37 L m-1. The volume of the inflated enclosure was ∼57 L, giving the air a residence time of ∼1.5 min. Additional residence time of sample air due to the sample line (30 m, 3.175 mm inside diameter – ID) was ∼6 s.

For the trace gas exchange experiments, known amounts of NO2, NO, or O3 were added into the scrubbed air stream. NO2 and NO were from compressed standard gas cylinders (Scott-Marrin, Inc., California, USA), and O3 was made in situ using a mercury Pen-Ray lamp O3 generator (UVP, California, USA) and compressed zero air. A KOFLO^®-type mixer was placed just downstream of the trace gas inlet to ensure even mixing of the added trace component with the scrubbed air.

NO, NO2, O3, CO2, and H2O mixing ratios before and after the enclosures were measured, with the air sample selected using a set of solenoid valves. The concentrations of these gases were calculated using the ideal gas law and the measured air temperature. The time for each sample was 5 min, alternating between the enclosure inlet and outlet. The reference enclosure inlet and outlet were sampled once an hour. The environmental conditions were also recorded, including ambient and enclosure temperatures (S-THB, Onset Computer Corp., Massachusetts, USA), leaf temperatures (thermocouple wire sensors, Omega Engineering, Connecticut, USA), relative humidity (S-THB, Onset), leaf wetness (for qualitative assessment of leaf conditions only) (S-LWA, Onset), and photosynthetically active radiation (PAR) (S-LIA, Onset). Standard commercially available instruments were used for O3 (Model 49i, Thermo Fisher Scientific, USA) and CO2/H2O (LiCor 840, Li-Cor Corp., Nebraska, USA). NO and NO2 were measured using a home-built chemiluminescence detector that utilizes the light-emitting reaction of NO with O3 (Ryerson et al., 2000).

The NOx instrument was programmed to run in 5 min cycles, each with a 1 min measurement of zero air (UHP, Airgas, USA), followed by a 2 min measurement of NO and a 2 min measurement of NO2. In NO2 mode, NO2 was first converted to NO and then measured the same as in NO mode. The conversion was done using an LED UV light source (L11921-500, Hamamatsu Photonics). The peak light emission of this LED was at 385±5 nm, matching the absorption peak of NO2 and minimizing the interference from the unwanted photolysis of HONO. The NO2-to-NO conversion efficiency was ∼0.68. Because the ambient air was scrubbed to remove O3 (and other trace gases) before entering the enclosures, the effect of ambient O3 on NOx measurements was negligible. A high-concentration (1.5 ppm) NO standard dynamically diluted with ultra-high purity zero air (Airgas, USA) was used to calibrate the NO measurement. For the NO2 calibration, NO in the same diluted standard was partially converted to NO2 by adding a controlled amount of ozone (generated in situ using a Pen-Ray ozone generator and 99.98 % oxygen). The instrument calibration runs were initiated automatically about every 7 h during regular operation. The overall 1σ precision for a 5 min measurement cycle was ∼2 ppt for NO and ∼4 ppt for NO2. The accuracy of the NO and NO2 measurements was ∼30 ppt.

Water vapor at the enclosure inlet and outlet was measured using a LiCor 840. The instrument was calibrated using a LiCor dew point generator. The ambient relative humidity results from the Onset sensors were compared with the data from a nearby AmeriFlux tower (within 100 m) (Vogel, 2016), and the agreement was within 3 %. It was noticed that on particularly hot and humid afternoons, there was condensation of water in the sample line leading to the instruments. The condensed water was removed promptly with gentle warming of the affected section of the sample line. The data recorded during these times were excluded.

After installation, each set of branch and reference enclosures was first purged with scrubbed ambient air at least overnight and through the early morning hours (up to ∼10:00 local time) to allow the branches to acclimate and also to reduce the amount of any possible surface-deposited photochemically labile compounds that might interfere with the measurements (Raivonen et al., 2006). The gas exchange experiments were then started and carried out for the following 2 to 4 d. A known amount of the trace gas was introduced into the purge air flow. This included zero concentration, i.e., purging with scrubbed air between the trace gas additions. The maximum mixing ratios of the trace gases in the purge air were kept within the range of typically observed ambient measurements, i.e., NO2 <1.2 ppb, NO <300 ppt, and O3 <60 ppb.

The enclosed leaves were harvested after the completion of the measurements and immediately placed in an oven to be air-dried at 65 ∘C. The leaf area was then measured by forming a monolayer of the dried leaves on graph paper. The enclosed single-sided leaf area for white pine, red maple, bigtooth aspen, and red oak was 0.35, 0.26, 0.11, and 0.44 m2, respectively.

Leaf-level uptake or emission of the trace gas leads to a trace gas concentration difference between the enclosure inlet and outlet. In the enclosure, the flux of the trace gas with respect to the leaf surface is 1Fx=QAco-ci, where Fx (pmol m-2 s-1) is the flux of the trace gas x; ci and co, in pmol m-3, are the trace gas concentration measured at the enclosure inlet and outlet, respectively. Q (m3 s-1) is the purge air flow rate. A (m2) is the one-sided area of the enclosed leaves, as the stomata, the part of leaf anatomy most relevant to gas exchange, generally are located on the underside of tree leaves (Kirkham, 2014). A resulting flux with a negative sign reflects the loss of the trace gas at the leaf surface, and a positive flux, emission from the foliage. All the trace gas concentration changes through the branch enclosure (co-ci) were corrected against the background obtained from the reference enclosure before the fluxes were determined according to Eq. (1). The detection limit of flux, i.e., the minimum absolute value above which the flux is significantly non-zero (p<0.05, or at the 95 % confidence level), was determined using the flux data obtained during the scrubbed air purge at nighttime when no emission from the leaves was expected because generally the leaf stomata are closed at night. These detection limits (Table 1) reflect the measurement precision of the instruments, variations of the actual enclosure conditions over time, and fluctuations of the purge air flow rate. Nighttime transpiration in trees and shrubs has been measured in prior work, with reports of nighttime transpiration rates ranging from 0 % to as much as 25 % of the daytime value (Dawson et al., 2007), suggesting that leaf stomata may remain open at night for some plants. However, this possibility did not affect the above results as there was no evidence of a consistent concentration difference above zero between the enclosure outlet and inlet measurements.

The representative tree species were determined based on the basal and leaf area coverage within the 60 m radius of the research site. Tree branches for the enclosure experiments were selected for their accessibility from the ground. Preferences were given to those with adequate sun exposure and to mature trees whenever possible. Enclosed branches of white pine (Pinus strobus) and red maple (Acer rubrum) were ∼7 m above ground and from trees that were over 10 m in height. The enclosed branches of red oak (Quercus rubra, ∼6 m) and bigtooth aspen (Populus grandidentata, ∼4.5 m) were ∼3 m above ground.

Nitrogen dioxide (NO2) was introduced into the purge air at different concentrations between zero and 40 nmol m-3 (∼1 ppb). Generally, when NO2 was added to the enclosure, there was a negative flux, indicating uptake of the trace gas by the plant material (Fig. 4a). The magnitude of the flux was proportional to the input NO2 concentration. In addition, when the input concentration was held constant for several hours or overnight, the flux had a diurnal pattern, e.g., bigtooth aspen on 30 to 31 July. It was lowest at night, increased through the morning hours, and peaked around midday before diminishing again toward nighttime. This behavior strongly suggests that the NO2 uptake by these trees is in large part controlled by leaf stomatal aperture and, at the same time, driven by the NO2 concentration gradient from the air around the leaf surface to the leaf internal space.

There are a couple of factors that complicated the NO2 gas exchange experiment. First, the NO2 standard used for delivering NO2 to the enclosure contains about 5 % NO that was unavoidably added to the enclosure. Secondly, when there was intense direct sunlight, some NO2 in the enclosure was photolyzed. While corrections for these interferences were done using the measurements from the reference enclosure, it is difficult to completely remove the artifact caused by NO2 photolysis. This is because the sunlight exposure of the two enclosures, although situated side by side, was often uneven, and the measurements of the enclosures were done not simultaneously, but sequentially. This problem is particularly pronounced for clear-sky conditions with strong contrasts in sunlit and shaded conditions inside the canopy. The branch enclosure was always positioned to get more sun exposure than the reference enclosure if choices needed to be made. Therefore, the branch enclosure likely received more sunlight overall, even though it might be more shaded during some measurement cycles. Generally, for the periods of strong sunlight, there is residual NO after the correction against the reference enclosure is made. If we assume all this is due to an underestimation of NO2 photolysis and make a further correction by combining the changes in NO and NO2, the data quality is not improved, while more noise is introduced to the data. Because of this and because we are not absolutely certain about all possible sources of NOx from the branch enclosures, we prefer to adhere to the correction using only the reference enclosure measurements and view the resulting NO2 flux as an upper bound, with possibly as much as 20 % overestimation under direct sunlight conditions, which accounts for ∼16 % of all data during the NO2 exchange experiments. The size of the stomatal aperture, regulated by the plant's need to optimize photosynthesis and simultaneously minimize water loss, can be gauged by stomatal conductance of water using Eq. (2) (Weber and Rennenberg, 1996): 2gH2O=FH2O/cH2O_leaf-cH2O_enclosure, in which the flux of water (FH2O, in mmol m-2 s-1) due to plant transpiration is calculated by applying the measured water concentration difference at the inlet and outlet of the enclosure to Eq. (1). cH2O_leaf (mmol m-3), the water concentration inside the leaf air space, is calculated using the measured temperature of the enclosed leaves, assuming the air in the leaf internal space is saturated with water vapor. cH2O_enclosure (mmol m-3), the water concentration of the branch enclosure, is evaluated using the measured enclosure relative humidity and temperature data. The resulting stomatal conductance of water, gH2O (mm s-1), for the four enclosed branches is shown in Fig. 5a. The stomatal conductance has a clear diurnal pattern, mainly following the daily cycles of sunlight and photosynthesis. The magnitude varies from tree to tree. The conductances of the white pine and the red maple branches were similar, ranging from near zero at night to about 3 mm s-1, while the conductance of red oak was 0 to ∼1 mm s-1, and the bigtooth aspen, 0 to ∼6 mm s-1. When the conditions are such that the difference between the leaf and air temperatures is small and the enclosure humidity is high, the difference between cH2O_leaf and cH2O_enclosure is also reduced, increasing the uncertainty in gH2O. In our measurements, this happened mostly from dawn to sunrise, accounting for ∼10 % of the total data points, where the (cH2O_leaf-cH2O_enclosure) was within 1 standard deviation from 0.

Knowing the stomatal conductance of water, the expected rate of NO2 deposition through the plant stomata can be calculated. Across the stomata, the deposition is a diffusion-controlled process (Weber et al., 1998; Weber and Rennenberg, 1996), where the expected rate is the product of the stomatal conductance of water multiplied by the square root of the ratio of the molecular weight of water to the molecular weight of NO2: 3gx=gH2O×MWH2OMWx, where gx (mm s-1) represents the expected stomatal uptake rate for the trace gas species x (here x=NO2); gH2O (mm s-1) is the stomatal conductance of water; and MW (g mol-1) represents molecular weight.

From the measurements, the leaf-level NO2 deposition velocity, vdNO2 (mm s-1), is the NO2 flux toward leaf surface (FNO2, in pmol m-2 s-1) normalized to the corresponding NO2 concentration in the enclosure (co,NO2, in pmol m-3): 4vdNO2=FNO2/co,NO2. If NO2 deposition is exclusively controlled by stomatal uptake, agreement between the measured deposition velocity and the calculated stomatal uptake rate is expected, i.e., vdNO2=gNO2. If not, additional factors, such as internal mesophyll resistance (Gut, 2002; Thoene et al., 1996) or leaf cuticular adsorption (Geßler et al., 2002; Coe, 1995; Rondón et al., 1993), may also play a role, as the former decreases, and the latter increases the overall foliar deposition velocity.

In Fig. 5b, the measured foliar deposition velocity of NO2 is plotted together with the calculated stomatal uptake rate for comparison. The agreement is generally good for all experiments, suggesting that the foliar deposition of NO2 for these tree species is indeed closely related to stomatal aperture. This also suggests that the effects of internal mesophyll resistance and cuticular uptake of NO2 are relatively minor. The strength of correlation between NO2 deposition and stomatal conductance is evaluated using the Pearson correlation coefficient, ρ, which has a possible value between -1 and 1, with a value of 0 indicating no correlation and a value that is away from 0 indicating increasing positive or negative correlation. In Fig. 6, the foliar NO2 deposition velocity is plotted against the stomatal conductance for each tree. The correlation coefficient for bigtooth aspen is 0.96, and for red oak it is 0.85, both showing a strong positive correlation between the deposition velocity and stomatal conductance. The correlation is also evident but relatively weaker for the white pine (ρ=0.73) and red maple (ρ=0.71).

The relationship between the deposition velocity and stomatal conductance is also examined using linear regression analysis. If NO2 deposition is entirely controlled by stomata, the deposition rate at zero conductance (gH2O=0), when the stomata are closed, should be zero, and the slope of the deposition rate vs. stomatal conductance should be equal to MWH2OMWNO2 or 0.62 (recall Eq. 3). This relationship is shown in Fig. 6 with the solid blue line. The best fit and the 95 % confidence bounds are represented by the red solid and dashed lines. Also listed in the figure are the slope (m), the intercept (b), and the r2 value of each fit. The linear relationship for bigtooth aspen appears to be the tightest, where over 90 % of the data variation can be explained by the fit. The intercept is nearly zero, and the slope of 0.56 is close to 0.62, making it reasonable to conclude that for the bigtooth aspen, stomatal uptake dominates NO2 loss at the leaf surface. A similar conclusion can be made for red oak, where r2 is 0.72 and the slope and the intercept are 0.54 and 0.03, respectively.

The red maple is different. The slope of NO2 deposition rate to stomatal conductance is 0.25, far less than 0.62. The data also appear to have more scatter. In the time series plot (Fig. 5a), the stomatal conductance on the morning of 25 July (from 08:30 to 12:30) shows high variability that is not reflected by the NO2 deposition rate at the same time. Possibly an unknown measurement issue for water concentration during this time or sources of water exchange at the leaf surface other than stomata (see Discussion Sect. 4.1) led to the high variability. However, excluding this portion of the observations and using only the data obtained prior to this time window, from 13:00 on 24 July to 08:00 on 25 July, resulted in a modestly improved linear fit with a slope still below 0.3. For white pine, the slope of vdNO2 vs. gH2O is 0.40, also lower than the expected value of 0.62 based on stomatal-controlled diffusion. These lower-than-expected slopes imply there may exist mesophyll resistance to NO2 uptake for these tree species. Such resistance to stomatal uptake of NO2 has previously been observed on some trees such as European Picea abies (Norway spruce) seedlings (Thoene et al., 1996) and Amazonian Laetia corymbulosa and Pouteria glomerata (Gut, 2002). However, in a separate study of Norway spruce seedlings (Rondón and Granat, 1994), no evidence of internal resistance to NO2 stomatal uptake was found. Research on CO2 diffusion and H2O transport into leaf internal spaces has revealed that mesophyll resistance is subject to environmental perturbations, and the responses among and within species can vary (Xiao and Zhu, 2017). It is reasonable to assume that the mesophyll resistance to NO2 uptake may also be subject to environmental conditions, and systematic observations under different conditions are needed to obtain more general conclusions.

The y intercept of the fitted line accounts for any possible additional foliar deposition when the stomata are closed and consequently the stomatal conductance is zero. Of all four trees studied, only white pine has an intercept significantly larger than zero at 0.43±0.09 mm s-1, indicating a possible role of wet leaf surfaces and/or cuticular uptake. (See Discussion Sect. 4.1 below.) The nighttime stomatal conductance of white pine is relatively high, with a median value of 0.57 mm s-1, compared to 0.05–0.19 mm s-1 for the other trees, probably due to incomplete stomatal closure at night (Dawson et al., 2007). There is corresponding nighttime deposition of NO2, with a higher rate for white pine relative to the other trees (Fig. 5).

Nitric oxide (NO) was added to the purge air at concentrations up to ∼10 nmol m-3 (∼250 ppt) to the white pine, red maple, and bigtooth aspen enclosures (Fig. 4b). Red oak was not included in this experiment. No significant NO flux toward the leaf surface was observed. This agrees with observations made on Scots pine (Rondón et al., 1993), corn leaves (Hereid and Monson, 2001), and Quercus agrifolia (Delaria et al., 2018). In contrast, for the white pine, a positive flux up to 2.7 pmol m-2 s-1 from the enclosure was measured when NO was added (Fig. 4b, white pine), indicating emission from the enclosed plant material. This flux also appears to increase with the enclosure NO mixing ratio. Although the photolysis of surface-deposited nitrogen oxides may cause such positive NO flux, during the scrubbed air purge prior to the addition of NO, there was no significant NO emission from the pine enclosure. That said, this experiment was done only once in a span of 5 h from late morning to early afternoon. We cannot absolutely rule out possible interference from nitrogen-containing chemical components in the system.

Up to 2.2 µmol m-3 (∼55 ppb) of ozone (O3) was introduced to the enclosures. As in the case of NO2, there was an O3 loss within the enclosure, and it increased with input O3 concentration (Fig. 4c). Shown in Fig. 5c is the comparison of the measured foliar O3 deposition velocity to the expected stomatal uptake rate calculated using leaf stomatal conductance and the square root of the ratio of the O3 and H2O molecular weight, MWH2OMWO3. For red maple, bigtooth aspen, and red oak, these two values agree reasonably well, implying that foliar ozone loss is mainly through leaf stomata and closely related to stomatal conductance. Correlation analyses were not performed here due to the limited number of data points.

For white pine, the measured leaf-level O3 deposition velocity is significantly greater than the expected stomatal uptake rate by a factor of 2 or more. It is known that on average up to 60 % of ozone deposition in vegetated areas is through non-stomatal pathways (Clifton et al., 2019). Within a branch enclosure, non-stomatal pathways can include deposition to wet leaf surfaces (Zhou et al., 2017; Altimir et al., 2004), cuticular uptake, chemical reactions at the leaf surface (Jud et al., 2016; Fares et al., 2010), and in the gas phase with biogenic organic compounds (BVOCs). Estimation of the possible contribution from gas-phase reactions with BVOCs was made as follows. The upper bounds of typical emission rates at 30 ∘C and PAR level at 1000 µmol m-2 s-1 for monoterpenes and other BVOCs (excluding isoprene) are 3 and 5 µg C g-1 h-1, respectively (Guenther et al., 1994). The speciation of major BVOCs emitted by white pine at UMBS is based on Kim et al. (2011), including α- and β-pinene, limonene, linalool, α-humulene, and β-caryophyllene. Using the rate constants of the BVOCs with ozone reactions (Burkholder et al., 2015) and the residence time of 1.5 min in the enclosure plus ∼6 s in the sample line before reaching the detector, the estimated ozone loss due to gas-phase chemical reactions was less than 1 %. Even with optimal light and temperature conditions for BVOC emission, the estimated gas-phase chemical removal would only be on the order of a few percent.

For 2 weeks during the summer PROPHET2016 campaign, we examined the leaf-level NO2, NO, and O3 gas exchange of four different tree species. This work provided a first insight into the general characteristics of the gas exchange of these North American trees in their natural habitat. The trees used in the enclosure measurement represent 80 % of the total leaf area within a 1000 m radius of the research site and 90 % within the 60 m radius. It is evident from the results that bidirectional foliar gas exchange depends on the trace gas in question and tree type and is influenced by diverse and complex environmental conditions, similar to the findings from previous studies mainly on European tree species and on annual plants (grasses and crops). The foliar uptake rates of NO2 and O3 vary from tree to tree and even within the same tree. Leaf stomatal conductance of H2O emerges as a strong indicator of the uptake efficiency. The foliar NO2 deposition of bigtooth aspen and red oak is almost entirely controlled by stomatal aperture. For red maple and white pine, the correlation coefficient is over 0.7, even though the measured NO2 foliar deposition velocity is 40 %–50 % of the predicted stomatal uptake rate. Except for white pine, the O3 foliar deposition velocities of all the studied trees also covary with stomatal conductance (Fig. 5c). Generally, the leaf-level NO2 and O3 deposition velocity can largely be inferred from the stomatal conductance of water only, as also concluded from earlier studies of European tree species (Breuninger et al., 2013; Rondón and Granat, 1994).

Thus, the factors controlling leaf stomatal conductance would in turn greatly influence NO2 and O3 deposition in a forested environment. These factors include PAR level, ambient temperature, moisture, soil conditions, as well as ambient CO2 (Jarvis, 1976). Further, the capability of the foliar uptake of trace gases would also depend on the intrinsic characteristics of leaf stomata, such as their size and density on the leaf surface, determined by plant species and stage of maturity, and factors such as growth history, leaf age, tree height, and the vertical location of the leaf on the tree (Kirkham, 2014; Sparks et al., 2001; Schäfer et al., 2000). In this work, the stomatal conductance of the bigtooth aspen was 3–5 times higher than that of the other trees. Biological features, such as plant and leaf age and stomatal density, may have contributed to this difference. Compared with the other three trees in this work, the aspen was younger and smaller. The enclosed branch was in the upper part of the crown containing developing new leaves. Past measurements, albeit on different species, have shown that for the same species under similar environmental conditions, leaves of young trees generally have higher stomatal conductance than old ones (Niinemets, 2002; Hubbard et al., 1999; Fredericksen et al., 1995; Yoder et al., 1994). Another possible reason for the observed high gH2O, while direct evidence has yet to be found, is the number of stomata. Many trees have stomata on only the lower (abaxial) leaf surface; however, trees that belong to the genus Populus, which includes aspen, are an exception. They have stomata on both sides (amphistomatous), a feature that allows increased photosynthetic rate and fast growth (Kirkham, 2014). If the bigtooth aspen leaves are indeed amphistomatous, a relatively high gH2O can be expected. We compared the environmental conditions of the enclosures. The integrated PAR exposure levels were similar. The daily variation of the relative humidity in the bigtooth aspen enclosure was not significantly different from the others. In contrast, the average daily temperature was 19.2 ∘C, cooler than the temperatures (23.9, 22.6, and 21.6 ∘C) in the other enclosures, similar to the average ambient air temperature outside the enclosure during the same time, 19.1 and 23.6 ∘C, 22.4 and 21.3 ∘C. The combined conditions of moisture and temperature led to a relatively low vapor pressure deficit (VPD) in the aspen enclosure, 0.8 kPa, compared to 1.2 kPa (white pine), 1.0 kPa (red maple), and 1.4 kPa (red oak) in the others. Generally, VPD and gH2O are inversely correlated, and a low VPD corresponds to a relatively high gH2O (Urban et al., 2017a, b; Hubbard et al., 1999). However, because here we are comparing different tree species, we consider the observed results to stem from the combination of biological and environmental factors. Further examination of these factors is beyond the scope of this paper; nevertheless, it would be beneficial to take this temporal and spatial variability and inhomogeneity into account in model parameterizations of trace gas dynamics since plant stomata are the main conduit of NOx and O3 deposition over vegetation.

When extrapolated to zero stomatal conductance, the deposition velocity of NO2 to white pine was 0.43 mm s-1 (Fig. 6a), implying deposition unrelated to leaf stomata, possibly to wet leaf surfaces and/or to leaf cuticula. This observation does not exclude the possible existence of these pathways when the stomata are open. A deposition velocity higher than expected based on the stomatal conductance would result if there is significant non-stomatal deposition. On the other hand, mesophyll resistance renders a lower deposition velocity than the expected value. There is no mechanistic reason why the deposition velocity associated with either a non-stomatal pathway or mesophyll resistance should remain constant or vary linearly with stomatal conductance. The relationship of deposition velocity, vdNO2, and stomatal conductance, gH2O, would remain essentially linear as long as stomatal deposition dominates or the non-stomatal deposition term is constant while mesophyll resistance is small. However, if mesophyll resistance is significant, it would limit the increase in vdNO2 with stomatal conductance.

To assess the role of wet leaf surfaces in non-stomatal deposition, we calculated the white pine enclosure dew point using the temperature and relative humidity data and compared it to the measured leaf temperature. The leaf temperature was always higher than the dew point during the experiments, excluding the possibility of a wet leaf surface from the condensation of pure water. However, a microscopic water film may nevertheless form at a relative humidity as low as 50 % if there are hygroscopic deposits on the leaf surface (Sun et al., 2016; Burkhardt and Hunsche, 2013; Burkhardt and Eiden, 1994). The microscopic water film could potentially modify gas exchange rates of water-soluble trace gases in the air. Data from this work do not contain information that can be used to delineate the possibilities of trace gas dissolution into microscopic water films or cuticular uptake. Further investigations with appropriately designed experiments, better measurement precisions, longer observation time, and under different environmental conditions are necessary to delineate the various possible deposition pathways and their dependencies.

To put our results into perspective, we compare our measured daytime maximum foliar deposition velocity of NO2 with the results from previous studies on American trees (Table 3). Although the development stages of the trees, PAR, humidity, and temperature conditions are different, the results are comparable, ranging from 0.76 to 1.6 mm s-1 from this work and from 0.4 to 1.8 mm s-1 from earlier work. We also compare our results to those of several native European trees – Scots pine, evergreen oak, common oak, European beech, and silver birch – measured under the conditions of PAR = 900 µmol m-2 s-1, maximum temperature 27.7 ∘C, and relative humidity 31.2 %–99.9 % (Breuninger et al., 2013). The maximum NO2 deposition rates were ∼0.5–1 mm s-1 for all but the birch tree, which was ∼1.5 mm s-1. These numbers are also fairly similar to those of pine, maple, and oak reported here. What stands out but without a direct or closely related comparison is the high rate of trace gas uptake by the aspen leaves. Although the comparisons show reasonable agreement, it is evident that the NO2 (and O3) foliar uptake is highly variable depending on a myriad of conditions, both environmental and intrinsic to tree species and developmental stage. Measurement results and comparisons from different studies are probably also sensitive to experimental protocols and environmental conditions. These factors should be taken into consideration if more comparisons are to be made in future work.

There was no significant leaf-level deposition of NO for all the tree species studied here. Instead, relatively small NO emissions were detected from white pine when up to ∼250 ppt NO was added to the enclosure. Delaria et al. (2018) reached the same conclusion from their study on Quercus agrifolia. We searched for possible errors that might have led to the results but could not find an obvious explanation. Certainly, additional measurements are necessary to verify this observation. Using the leaf area index of white pine at UMBS, 0.11 m2 m-2 (Chris Vogel, personal communication, 2016), and the maximum measured flux, 2.7 pmol m-2 s-1, we estimated the potential canopy-wide NO flux from this emission to be 0.3 pmol m-2 s-1, less than 10 % of the reported minimum soil NO emission flux of 4–10 pmol m-2 s-1 at UMBS (Nave et al., 2011).

Although this observation seems counterintuitive, in previous publications, emission of NO has been reported from leaves of individual corn plants exposed to 0.1–0.3 ppb NO (Hereid and Monson, 2001), from leaves of California live oak exposed to air containing NO (Delaria et al., 2018), from several nitrate-nourished plant species (Wildt et al., 1997), as well as from pesticide-treated soybean leaves (Klepper, 1979). Additionally, recent plant physiological studies have started to reveal the mechanism of plant NO production and its importance for regulating growth and development, immunity, and signaling (Astier et al., 2017; del Río, 2015; Yu et al., 2014) as well as for responding to pollutants and stress (Bison et al., 2018; Farnese et al., 2017; Velikova et al., 2008). In light of these advances, more targeted observations of foliar NO exchange probably should be conducted while taking these biological factors into consideration.