#### 2.3.1. Eddy Covariance

The sensible heat flux (H), latent heat flux (LE), flux of CO

_{2} (

${F}_{C{O}_{2}}$) and

${F}_{{O}_{3}}$ were calculated with the software EddyPro 6.2.0 (Licor Bioscience) [

37] (see Equation (1),

Section 2.3.2, for the calculation of the ozone flux). Raw data processing and flux calculation included de-spiking [

38], double rotation of the 3D wind components, block averaging, lag correction with covariance maximization within defined bounds, corrections for sensor separation according to Horst and Lenschow [

39], high-frequency corrections according to Ibrom et al. [

40], and high-pass filtering effects according to Moncrieff et al. [

41]. Additionally, the ozone fluxes were corrected for the dilution of the ozone concentration by water vapor, which was not measured within the instrument [

42]. The calculated fluxes were quality controlled for stationarity and integral turbulence characteristics [

43]. The CO

_{2} storage term was estimated using the 1-point storage correction and then added to the CO

_{2} fluxes to derive the net ecosystem exchange (NEE). Nocturnal data were filtered for low friction velocity (u*) values. The u* threshold (0.1 m s

^{−1}) was estimated following Papale et al. [

44].

Data quality control was performed according to Mauder and Foken [

45] and resulted in three classes: QC_0 data are considered to be of very good quality, QC_1 data are of acceptable quality, and QC_2 data should not be considered for scientific data analysis. Further, data sets that are associated with low u* values (u* < 0.1 m s

^{−1}) are considered problematic because the respective turbulence is not sufficiently established. After removal of QC_2 data and those with low u* values (see

Section 3.1 for details), the remaining data set was gap-filled with marginal distribution sampling as described in Reichstein et al. [

46]. Statistical analyses (

Section 3.2 and

Section 3.3) were performed with non-gap-filled data.

Subsequently,

NEE was partitioned into gross primary productivity (

GPP) and ecosystem respiration (

R_{eco}) by using daytime flux partitioning with light response curves [

47] as implemented in REddyProc 1.0.0 for R [

48].

The flux footprints were calculated as described in Kljun et al. [

36] for half-hour fluxes when the u* threshold was exceeded. Nighttime fluxes were calculated whenever the global radiation (i.e., the shortwave, downwelling radiation

R_{g}) was below or equal to 5 W m

^{−2}, while daytime conditions were defined for global radiation values above 200 W m

^{−2}. With this classification, the transition periods around sunrise and sunset were excluded from the analysis.

Co-spectra were computed for the vertical wind component in conjunction with temperature and O

_{3}, respectively, for each half hour. Frequency-class binned averages were then used to calculate ensemble co-spectra from good-quality data and well-developed turbulence conditions for individual stability classes. Reference co-spectra for temperature and vertical wind were calculated according to Kaimal and Finnigan [

49].

The stability of the boundary layer was determined from the ultrasonic anemometer data as (z-d)/L with L being the Monin-Obukhov length and d the displacement height (d = 0.67 · z). Neutral stability conditions were defined as −0.03 ≤ z/L ≤ 0.03, stable conditions as z/L > 0.03, and unstable conditions as z/L < −0.03, respectively.

#### 2.3.2. Resistance Terms and Deposition Velocity

${F}_{{O}_{3}}$ was measured with the eddy covariance method as described above on the basis of high-frequency measurement of

${c}_{{O}_{3}}$ and the vertical wind component

w (Equation (1) in [

24]),

where the primes describe the deviations of individual data points from the half-hour means and the overbar indicates the half-hour averages. Further,

${F}_{{O}_{3}}$ can be expressed in terms of concentration gradients and resistance terms (Equation (4) in [

24]):

Here,

${c}_{{O}_{3}}$ is the ambient O

_{3} concentration at measurement height and

${c}_{0}$ is the concentration within the sub-stomatal cavities, which can be assumed to be zero [

50]. The total resistance against the O

_{3} deposition

${R}_{tot}$ is equal to the sum

${R}_{a}+{R}_{b}+{R}_{c}$ (Equation (2)), where the aerodynamic resistance (

${R}_{a}$) and the bulk leaf boundary layer resistance for ozone (

${R}_{b}$) are computed from ultrasonic anemometer data as (Equations (8) and (9) in [

50])

and (Equation (6) in [

50])

Here, σ

_{v} is the standard deviation of the lateral wind component,

u* is the friction velocity,

u is the horizontal wind velocity,

Sc = 1.07 is the Schmidt number for ozone and

Pr = 0.72 is the Prandtl number [

51]. The surface resistance

${R}_{c}$ can now be calculated from Equation (2) with

${c}_{0}=0$ and by using the measured ozone flux

${F}_{{O}_{3}}$ (Equation (1)) and the ambient ozone concentration

${c}_{{O}_{3}}$:

Following the reasoning of Gerosa et al. [

25] and Fares et al. [

17],

${F}_{{O}_{3}}$ can be divided into

${F}_{sto}$ and

${F}_{nsto}$ (Equation (9) in [

24]),

${F}_{sto}$ happens together with the exchange flux of CO

_{2} and H

_{2}O vapor through the plant stomata.

${F}_{nsto}$ includes all other O

_{3} sinks in the vegetation including plant, water, soil surfaces, and chemical reactions. It is further presumed that the flux-resistance concept can be applied for both terms on the right-hand side of Equation (6) separately while using an ozone concentration near the surfaces

${c}_{{O}_{3}surf}$ (Equation (9) in [

24]):

Note that

${c}_{{O}_{3}surf}$ is not measured but will be computed as a residual once

${R}_{sto}$ and

${R}_{nsto}$ are known (see below, Equations (8)–(12). The surface resistance

${R}_{c}$, which is known from Equation (5), can be partitioned into two resistances acting in parallel, the stomatal (

${R}_{sto}$) and the non-stomatal (

${R}_{nsto}$) resistances (Equation (9) in [

21]):

To calculate the stomatal resistance of ozone,

${R}_{sto}$, we use the estimate of the stomatal conductance for water vapor from the Penman-Monteith approach (

g_{sto,PM}) and the molecular diffusivities for water vapor

${D}_{{H}_{2}O}$ and ozone

${D}_{{O}_{3}}$ (Equation (4) in [

11]),

where β is the Bowen ratio H / ET,

s is the slope of the water vapor saturation curve (Pa K

^{−1}), and γ is the psychrometric constant (kPa K

^{−1}).

${R}_{b,{H}_{2}O}$ is the bulk leaf boundary layer resistance for water vapor, which is computed from Equation (4) with the Schmidt number for water vapor (0.68; [

51]). We calculated the water vapor pressure deficit (VPD) in Equation (9) from the difference between

$e{s}_{{T}_{surf}}$, which is saturation vapor pressure (kPa) at the temperature at canopy surface level

T_{surf}, and

e, which is the vapor pressure (kPa) at measurement level

z. This procedure was proposed by Gerosa et al. (Equation (12) in [

24]), who suggest calculating

$e{s}_{{T}_{surf}}$ as

and this calculation is valid as long as no direct evaporation takes place. To account for “computational contamination” of the stomatal water vapor conductance in Equation (9) from direct evaporation, we applied a correction after Lamaud et al. [

11]. For this purpose, the slope of the regression line α between

${g}_{sto,PM}$ and

GPP is computed for conditions during daytime and

rH < 60% only. Under these conditions, the direct evaporation from wet surfaces is at its minimum (zero), and the stomatal conductance for water vapor and CO

_{2} are related by the constant α. This ratio is used to compute the stomatal conductance of ozone for all conditions (Equation (5) in [

11]):

Now,

${R}_{sto}$ can be directly calculated as

and

${R}_{nsto}$ can be calculated from Equation (8),

${c}_{{O}_{3}surf}\text{}$from Equation (7), and the flux partitioning can be derived from Equation (6). This routine was applied for each 30-min interval of the measurement period.

The deposition velocity

${v}_{d}$ is typically considered as the inverse of

${R}_{tot}$, while, by convention, a minus sign is introduced into Equation (11) (Equation (7) in [

50]):

Additionally, the deposition velocity can be calculated from measured data directly as (Equation (9) in [

11])

In analogy to eqs. 8 and 11, ozone deposition velocities towards the stomata are defined as ${v}_{d,sto}=\frac{1}{{R}_{sto}}$, and towards all non-stomatal surfaces as ${v}_{d,nsto}=\frac{1}{{R}_{nsto}}$. Note that ${v}_{d,sto}$ equals ${g}_{sto}$.