The dynamic chamber method : trace gas exchange fluxes ( NO , NO 2 , O 3 ) between plants and the atmosphere in the laboratory and in the field

We describe a dynamic chamber system to determine reactive trace gas exchange fluxes between plants and the atmosphere under laboratory and, with small modifications, also under field conditions. The system allows measurements of the flux density of the reactive NO-NO 23 triad and additionally of the non-reactive trace gases CO 2 and H2O. The chambers are made of transparent and chemically inert wall material and do not disturb plant physiology. For NO2 detection we used a highly NO 2 specific blue light converter coupled to chemiluminescence detection of the photolysis product, NO. Exchange flux densities derived from dynamic chamber measurements are based on very small concentration differences of NO 2 (NO, O3) between inlet and outlet of the chamber. High accuracy and precision measurements are therefore required, and high instrument sensitivity (limit of detection) and the statistical significance of concentration differences are important for the determination of corresponding exchange flux densities, compensation point concentrations, and deposition velocities. The determination of NO2 concentrations at sub-ppb levels ( <1 ppb) requires a highly sensitive NO/NO2 analyzer with a lower detection limit (3σ -definition) of 0.3 ppb or better. Deposition velocities and compensation point concentrations were determined by bi-variate weighted linear least-squares fitting regression analysis of the trace gas concentrations, measured at the inlet and outlet of the chamber. Performances of the dynamic chamber system and data analysis are demonstrated by studies ofPicea abiesL. (Norway Spruce) under field and laboratory conditions. Our laboratory data show that the quality selection criterion based on the use of only significant NO 2 concentration differences has a considerable impact on the resulting compensation point concentrations yielding values closer to zero. The results of field experiments demonstrate the need to consider photo-chemical reactions of NO, NO 2, and O3 inside the chamber for the correct determination of the exchange flux densities, deposition velocities, as well as compensation point concentrations. Under our field conditions NO2 deposition velocities would have been overestimated up to 80 %, if NO2 photolysis has not been considered. We also quantified the photolysis component for some previous NO2 flux measurements. Neglecting photo-chemical reactions may have changed reported NO 2 compensation point concentration by 10 %. However, the effect on NO 2 deposition velocity was much more intense, ranged between 50 and several hundreds percent. Our findings may have consequences for the results from previous studies and ongoing discussion of NO2 compensation point concentrations.


Introduction
Nitric oxide (NO), nitrogen dioxide (NO 2 ), often denoted as nitrogen oxides (NO x ), and ozone (O 3 ) are important compounds in atmospheric chemistry.NO x has an important role in radical chemistry and in the chemical formation and destruction of tropospheric and stratospheric O 3 (Crutzen, 1979).Moreover, NO x and O 3 are coupled by chemical reactions.NO is oxidized by O 3 to NO 2 and NO is regenerated by photolysis of NO 2 under daylight conditions.Typical NO x mixing ratios in the atmosphere are a few tenth of ppb Published by Copernicus Publications on behalf of the European Geosciences Union.
(remote sites) up to 1000 ppb (urban environments).Known sources of NO x are fossil fuel combustion (energy and traffic), biomass burning, microbial activity in soils and lightning (Seinfeld and Pandis, 2006).Typical ambient non-urban NO 2 concentrations are 0.05 to 1 ppb (Lerdau et al., 2000).Mean annual mixing ratios of NO 2 are up to 20 ppb in urban or industrialized regions, or 5 ppb in regions of little industrial activity.During smog events the NO 2 concentration may exceed 1 ppm (Stulen et al., 1998).
NO x is subject to a number of local photochemical removal processes, and long range transport through the atmosphere.In addition to gas-phase oxidation of NO 2 , principally by the OH radical (forming HNO 3 ), NO 2 is removed from the atmosphere via uptake to plants.Lerdau et al. (2000) reported that depending on the leaf area indices of the relevant sites only 25 to max.80 % of the emitted/produced NO x may be exported to the atmosphere, when comparing observed canopy level NO x concentrations and measured NO soil emission rates (see Jacob and Wofsy, 1990;Yienger and Levy, 1995;Wang et al., 1998).However, these results do not agree with leaf-level measurements regarding NO 2 emission from plants (besides plant uptake of NO 2 ) and indicating the existence of a so-called "plant compensation point" for NO 2 .Corresponding compensation point concentrations of NO 2 between 0.3 and 3 ppb have been reported (Rondón et al., 1993;Thoene et al., 1996;Weber and Rennenberg, 1996a;Sparks et al., 2001;Geßler et al., 2000Geßler et al., , 2002;;Hereid and Monson, 2001) suggesting plants act as a NO 2 sink when ambient concentrations are exceeding, or as a source of NO 2 , when ambient concentrations are below the NO 2 compensation point concentration.According to Lerdau et al. (2000), these results contradict the findings of Jacob and Wofsy (1990), who demonstrated that even at ambient NO 2 concentrations of 0.2 to 0.4 ppb a strong uptake by plants (primary rainforest) is required to align measured NO 2 concentrations in the canopy with the measured NO soil emission rates.Lerdau et al. (2000) emphasized the importance of finding an explanation for this discrepancy, particularly in remote regions far away from anthropogenic NO x sources (e.g.primary rain and boreal forests under low NO x regimes).Thus investigations of the contribution of NO 2 uptake by plants are required, particularly at NO 2 compensation point concentrations of (sub-)ppb levels.A recent study of five European tree species under laboratory conditions gives reason to assume a compensation point only at very low NO 2 values, if there is a compensation point at all (Chaparro-Suarez et al., 2011).
The commonly used technique for leaf-level exchange measurements of NO 2 is the dynamic chamber technique, a technique also used for many non-reactive (e.g.CO 2 , H 2 O, COS) and reactive trace gases (e.g.NO, O 3 , VOCs, DMS, CS 2 , HONO, HNO 3 , CH 2 O, HCOOH, CH 3 COOH).An entire plant (or parts of a plant) is enclosed in a (transparent) chamber which is purged by (preferably ambient) air.Two measurements of NO 2 concentration are performed, namely (1) at the entrance of the chamber (= ambient NO 2 concentration) and (2) within the chamber.If the chamber is well mixed, the latter measurement can be replaced by that of the outlet NO 2 concentration.Alternatively, a set of two chambers, one enclosing the plant the other being empty, can be used.To relate these two concentration measurements to the exchange (i.e. the uni-or bi-directional flux) of NO 2 between the (chamber) atmosphere and the enclosed plant (or parts of plant), the full mass balance of the dynamic chamber must be considered, i.e.NO 2 fluxes entering and leaving the chamber, as well as all other fluxes due to NO 2 sinks and sources within the chamber's volume.Under typical field conditions (i.e.ambient air enters the dynamic chamber), not only NO 2 , but also ambient NO and O 3 are purged through the chamber.The fast reaction between NO and O 3 is a "chemical" source of NO 2 , while (under daylight conditions) photolysis of NO 2 (λ ≤ 420 nm) is a "chemical" sink.Depending on ambient NO 2 , NO, and O 3 concentrations and UV irradiation intensity, corresponding "gas phase fluxes" may reach the magnitude of the NO 2 flux from/to the enclosed plant(s) (Meixner et al., 1997;Pape et al., 2009).Consequently, simultaneous measurements of NO 2 , NO, and O 3 concentrations at the outlet of the chamber are required.However, since there is substantial uptake of O 3 (and to a lesser extent NO) by the plants, NO 2 , NO, and O 3 concentrations at the inlet of the chamber must also be measured.As a positive "by-product" of these additional concentration measurements, deposition velocities of O 3 (and NO) may be inferred by considering the dynamic chamber's mass balances of O 3 and NO.
In this paper we present results from a dynamic chamber system used previously for measurements of volatile organic compounds, formaldehyde, formic and acetic acid and sulfur compounds (e.g.Kesselmeier et al., 1993Kesselmeier et al., , 1996Kesselmeier et al., , 1998;;Kuhn et al., 2000).The system allows exchange measurements of NO 2 (NO and O 3 ) under field conditions (uncontrolled) as well as studies under controlled conditions including (laboratory) fumigation experiments.
Because NO 2 compensation point concentrations were reported at (sub-)ppb levels, our laboratory NO 2 fumigation experiments were performed with 3-to 4-yr old Norway Spruce trees at 0.3-3.4ppb.Such low ambient NO 2 concentrations can be expected under field conditions.Moreover, exchange fluxes derived from dynamic chamber measurements are based on generally (very) small differences of NO 2 (NO,O 3 ) concentrations between inlet and outlet of the chamber.Consequently, considerable attention has been paid to the detection limits of corresponding analyzers, statistical significance of the concentration differences, as well as the statistical goodness of measurements have a substantial impact on the identification and quantification of statistically significant deposition velocities and compensation point concentrations.Furthermore, as the exchange of NO 2 is a complex interaction of transport, chemistry and plant physiology, we determined fluxes of NO, NO 2 , O 3 , CO 2 , and H 2 O in the field experiments.
While the internal plant physiological processes of atmospheric NO 2 consumption are known (NO is oxidized to NO 2 in aqueous solutions to deliver NO 2 ; Ghaffari et al., 2005), NO 2 is converted to HNO 3 and HNO 2 in water (Takahashi et al., 2007) and NO − 3 /NO − 2 are metabolized by corresponding reductases to NH + 4 which is incorporated into amino acids (Lea and Miflin, 1974;Yoneyama et al., 2003), possible NO 2 production processes inside plants and are to our knowledge entirely unknown.

Basic considerations
We consider a small branch of a tree (leaf area A leaf ), which is enclosed in a transparent plant chamber of volume V .The air within the plant chamber is well mixed by action of one (or more) fan(s).Ambient air (containing NO 2 , NO, and O 3 ) enters the plant chamber at the inlet, flushes the chamber with the purging rate Q (m 3 s −1 ) and leaves the chamber at the outlet.Within the plant chamber trace gases of the NO-NO 2 -O 3 triad may be (a) emitted and/or taken up from/by leaves, (b) deposited to the inner walls of the plant chamber, and (c) destroyed and/or generated by (fast) photo-chemical reactions.The mass balances of the NO-NO 2 -O 3 triad of the dynamic plant chamber are derived in Appendix A.

Molar flux densities, deposition velocities, and compensation point concentrations
Equations (A7.1)-(A7.3)are formulated in terms of molar fluxes (in nmol s −1 ).However, considering the exchange of reactive trace gases between the plant chamber's atmosphere and the enclosed leaves, the exchange flux density (F ex,i ) of the molar flux (in nmol m −2 s −1 ) is commonly used rather than the molar flux itself.In the case of plant chamber studies, the appropriate reference surface (reference area) is the surface area (A leaf , in m 2 ) of the leaves.Therefore, the exchange flux density F ex,i is defined as F ex,i : = i /A leaf , and the corresponding balance equations will read as follows: In the case of defined laboratory experiments without UV radiation, where plants are fumigated with only one of the three trace gases and excluding NO emission by the plants (i.e.gas-phase production and/or destruction of the trace gas can be ruled out), Eqs.(1.1)-(1.3)will reduce to the wellknown form of where the quantity with superscript " * " describes the nonreactive case with chemical reactions excluded.
In the case of bi-directional exchange (see Eq. A2), the exchange between the plant chamber air and the leaves can be directed to or away from the leaves.This exchange process can be subject to the so-called "compensation point concentration" (m comp,i , in nmol m −3 ).According to Conrad (1994), m comp,i is "that concentration at which the consumption rate reaches the same value as the production rate, so that the result of both processes is zero flux".Similar to the resistance concept e.g. by Hicks et al. (1987) the exchange flux density F ex,i (of the plant leaves) is parameterized here by a deposition velocity v dep,i (in m s −1 or mm s −1 ) of trace gas i and its compensation point concentration m comp,i : Note, that (by convention) F ex,i is directed "downward"/ inward to the leaves, if m s,i > m comp,i , F ex,i is zero, if m s,i = m comp,i , and F ex,i is directed "upward"/outward from the leaves, if m s,i < m comp,i .
In this context it has to be mentioned that the original definition of deposition velocity, as it is commonly used, is the ratio of the flux of a given trace species measured at some height above ground surface or canopy and the concentration difference between the height where the flux has been measured and a reference height where the concentration equals zero or the compensation point concentration, e.g.within the top soil layer or inside the stomata cavity.Thus, the deposition velocity is related to ground area and includes the well defined turbulent/aerodynamic (R a ) and molecularturbulent/boundary-layer (R b ) resistances.Dynamic chamber measurements usually minimize the turbulent resistance inside the chamber (R a ≈ 0) and modify the molecularturbulent/boundary-layer resistance R b (see Pape et al., 2009).Therefore, the inverse of the "chamber deposition velocity" is the sum of the so-called "canopy resistance" (R c ) and the chamber specific R b .For that reason the "chamber deposition velocity" is not equal to the leaf conductance of the corresponding trace substance.Therefore, deposition velocities in this paper should always considered as chamberspecific deposition velocities, which could be transformed to the quantities commonly used by the approach of Pape et al. (2009).
Given, that the quantities Q, A leaf , k, and j (NO 2 ) are a priori known and/or simultaneously measured with m s,i and m a,i , then, the desired quantities, v dep,i and m comp,i , are commonly determined from the linear relationship between F ex,i and m s,i , where v dep,i is the slope and m comp,i is the intersect of F ex,i with the m s,i -axis (see Rondón and Granat, 1994;Thoene et al., 1996;Weber and Rennenberg, 1996a;Sparks et al., 2001;Hereid and Monson, 2001;Geßler et al., 2002).
However, since F ex,i (see Eqs. 1.1-1.3)contains the term Q/A leaf (m a,i − m s,i ), the calculation of any form of linear regression between F ex,i and m s,i is mathematically sensu stricto not appropriate, because the dependent variable F ex,i contains the independent variable (m s,i ).This problem can be resolved by returning to the originally measured quantities, m a,i and m s,i .If we combine Eqs.(1.1)-(1.3)and Eq. ( 3) and resolve these equations for m s,NO 2 , m s,NO , and m s,O 3 , we yield three linear relationships between the measured variables m s,NO 2 and m a,NO 2 , m s,NO and m a,NO , and m s,O 3 and m a,O 3 : using the definitions: The quantities n i and b i of trace gas i cannot be determined (graphically or numerically) from single pairs of m a,i and m s,i , but from a (statistically sufficient) set of measured m a,i and m s,i (i.e.data sets classified for defined conditions of irradiation, temperature, humidity, concentrations, respectively).Therefore, n i and b i represent mean values for these data sets.Consequently, the quantities Q, A leaf , j (NO 2 ), k, m s,NO 2 , m s,NO , and m s,O 3 in Eqs.(6.1)-( 6.3), (7.1)-( 7.3), and (8.1)-(8.3)must be averaged over the same (time) period (the same data set) of m a,i and m s,i measurements from which the quantities n i and b i were derived.
The quantities n i and b i may be evaluated (graphically) as the intercept and the slope of the plot of measured m s,i versus measured m a,i .Application of different forms of linear regression analysis delivers n i and b i and bi-variate weighted linear least-squares fitting (which considers uncertainties of both, m s,i and m a,i ) provides also their standard errors s n,i and s b,i (see Sect. 3.4.6).
The linear relationships between F ex,i and m s,i are still maintained.This can be shown by resolving Eqs.(4.1)-(4.3)for m a,i and making use of Eqs.(1.1)-(1.3): Finally, the desired deposition velocities (v dep,i ) of the NO-NO 2 -O 3 triad result from Eqs. (4.1)-( 4.3), resolving for v dep,i , It has to be stated, that the use of averaged concentrations (quantities with overbars) for regression analysis (Eq.5.1 ff.) may eventually result in systematic errors for the derived quantities of compensation point concentrations and deposition velocities, particularly, if these concentrations will vary over a large range.This can basically not be ruled out.Moreover, concentration of NO, NO 2 , and O 3 may be more or less correlated, particularly during field conditions.The way we have considered these and other correlations is the application of the General Gaussian Error Propagation where those concentrations are considered explicitly (see Sect. 3.4.7).Errors (e.g.temporal variability) of averaged quantities are propagated to the final error of compensation point concentrations and deposition velocities.

Constraints of precision
Exchange flux densities F ex,i are determined from molar concentrations of the NO-NO 2 -O 3 triad, both ambient measurements (m a,i ) as well as those in the plant chamber (m s,i ) (see .These are all measured with one set of analyzers.The calculation procedure for exchange flux densities, deposition velocities as well as compensation point concentrations is based on linear regression analysis of m a,i and m s,i , which are (a) both error-prone, and (b) not very different from each other, i.e. their difference is usually (very) small.The uncertainties of these differences depend mainly on the precision of the analyzers; leading to large uncertainties in the derived quantities F ex,i , v dep,i , and m comp,i .
For the sake of simplicity we assume well defined laboratory conditions.Here, the trace gas exchange flux densities F ex,i are described by Eq. ( 2), implying that (a) only prescribed concentrations of trace gas i (= m a,i ) enter the dynamic plant chamber, (b) the enclosed leaves are only exposed to corresponding m s,i , (c) purging rate Q and leaf area A leaf are known and unchanging, and (d) sample concentrations of the other trace gases (m s,j =i ), photolysis rate j (NO 2 ) as well as wall-sorptions of trace gas i are negligible.After evaluation of the linear relationship between m a,i and m s,i , corresponding exchange flux densities F * ex,i , deposition velocities v * dep,i , and compensation point concentrations m * comp,i are given by Regarding only NO 2 , a schematic representation (using simulated data) of how the quantities defined by Eqs. ( 9)-( 11) are determined from genuine measurements of m a,NO 2 and m s,NO 2 is given in Fig. 1a.Since the "1:1"-line is equivalent to m a,NO 2 = m s,NO 2 (i.e.F ex,NO 2 = 0, see Eq. ( 2), the intersect of the linear regression line and the "1:1"-line is the NO 2 compensation point concentration, m comp,NO 2 .
Here, the difficulties associated with an experimental proof of a (highly) significant m comp,NO 2 becomes obvious.The lower m comp,NO 2 will be, the more the intersect shifts down the "1:1"-line, closer and closer to the limit of detection of the NO 2 concentration measurements (LOD (m a,NO 2 ), LOD (m s,NO 2 ); 3σ -definition).This dilemma becomes even more obvious, if we consider the schematic representation of Eq. ( 2) in Fig. 1b, where LOD (F  9)-( 11), the errors of F ex,NO 2 , v depNO 2 , and m comp,NO 2 are entirely due to the errors of n NO 2 and b NO 2 , which are in turn entirely due to the goodness of the linear relationship between m a,NO 2 and m s,NO 2 , as well as to the errors of m a,NO 2 and m s,NO 2 (s m a,NO 2 and s m s,NO 2 , see Sect. 3.4.7).This leads to the simple conclusion, that determinations of F ex,NO 2 , v depNO 2 , and m comp,NO 2 are more precise, the higher the regression coefficient R 2 (m s,NO 2 , m a,NO 2 ) and lower the standard errors s m s,NO 2 and s m a,NO 2 are.
Only one NO 2 analyzer is used for the measurements of both concentrations, m a,NO 2 and m s,NO 2 .As shown below (Sect.3.2), the standard error s m a,NO 2 (s m s,NO 2 ) was found to be a weak exponential function of m a,NO 2 (m s,NO 2 ), starting with a fixed value s m,LOD(NO 2 ) at m a,NO 2 = m s,NO 2 = 0. To demonstrate, how the goodness (R 2 (m s,NO 2 , m a,NO 2 )) of the linear relationship between m a,NO 2 and m s,NO 2 and how the magnitude of s m a,NO 2 and s m s,NO 2 impact the NO 2 exchange measurements, we consider (a) the determination of the minimum possible, but still highly significant NO 2 compensation point concentration (m comp,NO 2 ), and (b) the precision of the NO 2 exchange flux density (F ex,NO 2 ).For that we simulated data sets of m a,NO 2 and m s,NO 2 within the range LOD (m s,NO 2 ) ≤ m s,NO 2 ≤ 615 nmol m −3 (15 ppb) for pre-scribed NO 2 deposition velocities (0.1 ≤ v dep,NO 2 ≤ 0.8 mm s −1 , per leaf area) and for pre-scribed R 2 (m s,NO 2 , m a,NO 2 ) between 0.999 and 0.6.The latter was achieved by random number application to the m a,NO 2 data.Standard errors s m s,NO 2 and s m a,NO 2 were calculated from m a,NO 2 and m s,NO 2 (see Eq. 12.1, Sect.3.2), while the standard error of F ex,NO 2 (s F ex,NO 2 ) was calculated from s m s,NO 2 , s m a,NO 2 , and r (m s,NO 2 , m a,NO 2 ) = [R 2 (m s,NO 2 , m a,NO 2 )] 1/2 by application of the general form of Gaussian error propagation (see Sect. 3.4.7).
Application of bi-variate linear regression analysis to this simulated data set delivers the quantities n NO 2 and b NO 2 as well their standard errors s n,NO 2 and s b,NO 2 (which depend on s m s,NO 2 , s m a,NO 2 , and R 2 (m s,NO 2 , m a,NO 2 )).Application of the general form of Gaussian error propagation (see Sect. 3.4.7) to Eq. ( 11) delivers the standard error of the NO 2 compensation point concentration (s m comp,NO 2 ).The "detectable existence" of m comp,NO 2 (i.e.testing the hypothesis m comp,NO 2 = 0) has been statistically secured by application of the t-test to the values of m comp,NO 2 , s m comp,NO 2 and N (number of (m s,NO 2 , m a,NO 2 ) data pairs).In Fig. 2 and vdep,NO2 = 0.1, 0.2,…, 0.8 mm s -1 ).The greenish range represents simulated data of a 9 NO2 analyzer with LOD(mNO2) = 0.4 nmol m -3 (0.01 ppb), the bluish range for 10 LOD(mNO2) = 4.5 nmol m -3 (0.1 ppb), the reddish range for LOD(mNO2) = 44.6 nmol m -3 11 (1.0 ppb).12 13 Fig. 2. The dynamic plant chamber at well defined (laboratory) conditions: minimum detectable NO 2 compensation point concentrations (m comp,NO 2 at P ≥ 0.999, i.e. "highly significant") as function of NO 2 deposition velocity (v dep,NO 2 ; per leaf area) and the goodness (R 2 ) of the ambient vs. sample NO 2 concentration measurements (standard errors of NO 2 concentration measurements considered).Results are from data simulation (random number application) matching pre-scribed R 2 (m a,NO 2 , m s,NO 2 ) and prescribed v dep,NO 2 (0.999 ≤ R 2 ≤ 0.6 and v dep,NO 2 = 0.1, 0.2, ..., 0.8 mm s −1 ).The greenish range represents simulated data of a NO 2 analyzer with LOD (m NO 2 ) = 0.4 nmol m −3 (0.01 ppb), the bluish range for LOD (m NO 2 ) = 4.5 nmol m −3 (0.1 ppb), the reddish range for LOD (m NO 2 ) = 44.6 nmol m −3 (1.0 ppb).
NO 2 compensation point concentration, i.e. the lowest, but still highly significant m comp,NO 2 (P ≥ 0.999) is shown for a pre-scribed range of NO 2 deposition velocities as function of the regression coefficient R 2 (m s,NO 2 , m a,NO 2 ) and for three different values of LOD (m s,NO 2 ), namely 0.4, 4.5 and 44.6 nmol m −3 (0.01, 0.1, 1.0 ppb).These three values represent a certain "history" of NO/NO 2 chemiluminescence analyzers: LOD (m s,NO 2 ) = 44.6 nmol m −3 (1 ppb) represents the state-of-art of commercial NO 2 analyzers of 1985-1995, LOD (m s,NO 2 ) = 4.5 nmol m −3 (0.1 ppb) the best performance between 1995-2005's, while LOD (m s,NO 2 ) = 0.4 nmol m −3 (0.01 ppb) is characteristic for the most advanced NO/NO 2 analyzers which have been recently applied over the remote Southern Atlantic Ocean (Hosaynali Beygi et al., 2011).For typical ranges of laboratory measurements, i.e. 0.9 ≤ R 2 ≤ 0.99, minimum detectable NO 2 compensation point concentrations range between 17.5-99.4nmol m −3 (0.39-2.23 ppb), if NO 2 analyzers with LOD (m s,NO 2 ) = 44.6 nmol m −3 (1.0 ppb) have been used.Best performance of present-day NO 2 analyzers allow minimum detectable m comp,NO 2 between 3.6 and 21.3 nmol m −3 (0.08-0.48 ppb).Very low minimum detectable m comp,NO 2 (0.8-4.0 nmol m −3 or 0.02-0.09ppb) may be reached if the most advanced state of NO 2 analyzers is considered.It should be noted that, due to the potential goodness of the measurements, the minimum detectable m comp,NO 2 could be lower than the actual LOD (m s,NO 2 ), but statistically still highly significant.
The impact of s m s,NO 2 , s m a,NO 2 , and R 2 (m s,NO 2 , m a,NO 2 ) on the precision of the NO 2 exchange flux density (= s F ex,NO 2 /F ex,NO 2 ) is demonstrated in Fig. 3.For the sake of clarity, another data set has been simulated (random number application), namely for pre-scribed NO 2 deposition velocities (0.3 ≤ v dep,NO 2 ≤ 0.6 mm s −1 , per leaf area), a pre-scribed NO 2 compensation point concentration (m comp,NO 2 = 67 nmol m −3 (1.5 ppb)), and for 0.99 ≤ R 2 ≤ 0.9.Also shown in Fig. 3 is the precision of m s,NO 2 (= s m s,NO 2 /m s,NO 2 ; right axis) for the "history" of LOD (m s,NO 2 ) values, namely LOD (m s,NO 2 ) = 44.6,4.5, and 0.4 nmol m −3 (1.0, 0.1, 0.01 ppb).Before 1995 (LOD (m NO 2 ) = 1 ppb), a precision of m s,NO 2 better than 10 % could hardly be achieved in the lower ppb-range.Best performing present-day NO 2 chemiluminescence analyzers (LOD (m NO 2 ) = 0.1 ppb) exceed the 10 % level of m s,NO 2 precision not before m s,NO 2 falls below 14.8 nmol m −3 (0.33 ppb), while another order of magnitude can be reached with most advanced NO 2 analyzers (s m s,NO 2 /m s,NO 2 > 10 % not before m s,NO 2 < 1.5 nmol m −3 (0.03 ppb)).A review of these specifications of the NO/NO 2 analyzers are listed in Table 1.
The "history" of NO 2 analyzers is also mirrored in the precision of F ex,NO 2 (reddish, bluish, and greenish areas in Fig. 3).In any case, the precision of F ex,NO 2 (= s F ex,NO 2 /F ex,NO 2 ) reaches infinity at m s,NO 2 = m comp,NO 2 , since there the NO 2 exchange flux density equals zero.Otherwise, the precision of F ex,NO 2 rapidly falls (very) well below the 10 % level.This is a consequence of the fact, that m a,NO 2 and m s,NO 2 are the decisive quantities for the determination of F ex,NO 2 .Since m a,NO 2 and m s,NO 2 are highly correlated, the standard error of Finally, it should be emphasized, that the estimates of this subsection are made on the basis of Eqs. ( 9)-( 11) for (best) defined laboratory conditions.Under field conditions, however, the equations for the determination of F ex,NO 2 , v depNO 2 , and m comp,NO 2 will contain also average quantities of m s,NO , m s,O 3 , j (NO 2 ), and k (cf.Eqs. 6.1,7.1,8.1).It follows, that their variability (standard errors) leads to larger standard errors of n NO 2 and b NO 2 and diminish R 2 (m s,NO 2 , m a,NO 2 ).Consequently, corresponding minimum detectable NO 2 compensation point concentrations will certainly be higher and precisions of F ex,NO 2 will be lower than those given in Figs. 2 and 3. measured at the outlet of the dynamic chamber (precision ms,NO2, right axis).Results are from 6 data simulation (random number application), which considers standard errors of NO2 7 concentration measurements, and which matches pre-scribed R 2 (ma,NO2, ms,NO2) and pre-8 scribed mcomp,NO2 = 67 nmol m -3 (1.5 ppb).Black, grey, and dashed lines (= precision of 9 ms,NO2) represent data for a NO2 analyzer characterized by LOD(ms,NO2) = 44.6 nmol m -3 (1.0 10 ppb), LOD(ms,NO2) = 4.5 nmol m -3 (0.1 ppb), and LOD(ms,NO2) = 0.4 nmol m -3 (0.01 ppb), 11 respectively.Ranges of the precision of derived NO2 exchange flux densities are identified by 12 reddish, bluish, and greenish areas for LOD(ms,NO2) = 44.6 nmol m -3 (1.0 ppb), LOD(ms,NO2) = 13 4.5 nmol m -3 (0.1 ppb), and LOD(ms,NO2) = 0.4 nmol m -3 (0.01 ppb).The width of the colored 14 areas stands for all considered combinations of R 2 and vdep,NO2 (0.99  R 2  0.9 and 15 0.3  vdep,NO2  0.6 mm s -1 ).The respective upper boundary of each colored area represents the 16 combination vdep,NO2 = 0.3 mm s -1 and R 2 = 0.9, while the lower boundary represents vdep,NO2 = 17 0.6 mm s -1 and R 2 = 0.99.18 Fig. 3.The dynamic plant chamber at well defined (laboratory) conditions: precision of NO 2 concentration measurements (= s m,s NO 2 /m s,NO 2 ; right axis) and precision of derived NO 2 exchange flux densities (= s Fex NO 2 /F ex,NO 2 , left axis) as function of the NO 2 concentration measured at the outlet of the dynamic chamber (precision m s,NO 2 , right axis).Results are from data simulation (random number application), which considers standard errors of NO 2 concentration measurements, and which matches pre-scribed R 2 (m a,NO 2 , m s,NO 2 ) and prescribed m comp,NO 2 = 67 nmol m −3 (1.5 ppb).Black, grey, and dashed lines (= precision of m s,NO 2 ) represent data for a NO 2 analyzer characterized by LOD(m s,NO 2 ) = 44.6 nmol m −3 (1.0 ppb), LOD(m s,NO 2 ) = 4.5 nmol m −3 (0.1 ppb), and LOD(m s,NO 2 ) = 0.4 nmol m −3 (0.01 ppb), respectively.Ranges of the precision of derived NO 2 exchange flux densities are identified by reddish, bluish, and greenish areas for LOD(m s,NO 2 ) = 44.6 nmol m −3 (1.0 ppb), LOD(m s,NO 2 ) = 4.5 nmol m −3 (0.1 ppb), and LOD(m s,NO 2 ) = 0.4 nmol m −3 (0.01 ppb).The width of the colored areas stands for all considered combinations of R 2 and v dep,NO 2 (0.99 ≤ R 2 ≤ 0.9 and 0.3 ≤ v dep,NO 2 ≤ 0.6 mm s −1 ).The respective upper boundary of each colored area represents the combination v dep,NO 2 = 0.3 mm s −1 and R 2 = 0.9, while the lower boundary represents v dep,NO 2 = 0.6 mm s −1 and R 2 = 0.99.

Constraints of design
In addition to the demand for precise and highly sensitive measurements of NO 2 concentration, surface exchange flux measurements of NO 2 (NO, O 3 ) in a dynamic leaf chamber system require that: 1.The environment in the chamber should as closely as possible represent the surrounding (ambient) environment.4. Losses of NO 2 (NO, O 3 ) on chamber materials must be negligible (if not: must be quantified).

Enclosing
5. The chamber system should be applicable for laboratory and field measurements without substantial modifications.
6. Simultaneous measurements of surface exchange fluxes of NO 2 , O 3 , NO, CO 2 , and H 2 O should be feasible.
7. Differences of NO 2 (NO, O 3 ) concentrations between inlet and outlet of the dynamic chamber, which are expected to be small, must be resolved with statistical significance.
Furthermore, fumigation experiments to study the NO 2 surface exchange in the laboratory (NO 2 exchange under controlled conditions) demand the generation of very low (ppband sub-ppb levels) and temporally stable NO 2 concentrations in order to identify statistically significant NO 2 compensation point concentrations.These low NO 2 concentrations have to be reproducible and verifiable.

Trace gas analyzers
NO and NO 2 concentrations were measured by a gas-phase chemiluminescence NO analyzer (Model 42C, Thermo Electron Corporation, USA).In a low pressure reaction chamber, the NO of the air sample reacts with ozone (provided by the analyzer) forming electronically excited NO 2 molecules.Decaying to the ground state, the excited NO 2 molecule emits a photon (chemiluminescence) and the total light intensity in the reaction chamber, detected by a photomultiplier, is proportional to the NO concentration.NO 2 in the air sample is also measured by the NO analyzer after conversion of NO 2 to NO.In most commercial NO/NO 2 analyzers a molybdenum converter is applied (heated to 300-400 • C), where NO 2 is catalytically reduced to NO at the converter's surface.However, previous studies demonstrated that molybdenum converters are non-specific for NO 2 because other oxidized nitrogen compounds of ambient air, like gaseous nitrous acid (HONO), nitric acid (HNO 3 ), the nitrate radical (NO 3 ), dinitrogen pentoxide (N 2 O 5 ), peroxyacetyl nitrate (PAN), and other organic nitrates were found to be also converted to NO, which leads to systematic and considerable overestimation of the measured NO 2 values (Winer et al., 1974;Matthews et al., 1977;Grosjean and Harrison, 1985;Gehrig and Baumann, 1993;Steinbacher et al., 2007).During some studies hydrated, crystalline ferrous sulfate (FeSO 4 ) for the surface reduction of NO 2 to NO were used.However, FeSO 4 converter also overestimates the mixing ratio of NO and NO 2 (Ridley et al., 1988).Significant interferences of n-propyl nitrate, nitrous acid (HNO 2 ) and PAN were reported (Kelly et al., 1980;Cox et al., 1983;Fehsenfeld et al., 1987).As a consequence Fehsenfeld et al. (1987) did not recommend FeSO 4 converter for measuring NO 2 .Another frequently used analyzer to measure NO 2 is the Luminox detector (LMA-3, Scintrex/Unisearch Inc.).Its measurement principle is based on the chemiluminescent reaction of NO 2 with luminol in aqueous solution (Maeda et al., 1980;Wendel et al., 1983;Schiff et al., 1986).The luminol technique is noted for interferences by ambient O 3 and PAN, and exhibits non-linear response at low NO 2 concentrations.The interferences due to O 3 and PAN are significant especially at low NO 2 concentrations (Kelly et al., 1990).Table 2 shows an overview about commonly used NO 2 converters and their reported interferences.No interferences or any artifacts were reported for photolytic converters, where NO 2 is photolysed by ultraviolet light <420 nm (Fehsenfeld et al., 1990) or were negligible, respectively (Ryerson et al., 2000).Consequently, we used a highly NO 2 specific blue light converter (BLC) which photodissociates NO 2 into NO at a wavelength of approximately 395 nm (manufactured by Droplet Measurement Technologies Inc., Colorado, USA).To obtain a better accuracy and precision of the NO 2 (and NO) measurements at sub-ppb concentrations, the NO/NO 2 analyzer has always been operated with pure oxygen (instead with the oxygen of ambient air) for the internal generation of ozone, necessary for the reaction with NO in the low pressure reaction chamber.
Measurements of CO 2 and H 2 O concentrations were performed by infrared dual channel gas analyzer for difference measurements between the outlet of an empty reference chamber and the sample gas (LI-7000, LiCor, Lincoln, NE, USA).An additional gas analyzer (LI-6262, LiCor, Lincoln, NE, USA) monitored the absolute CO 2 and H 2 O concentrations to deliver a base signal for the LI-7000 operating in differential mode.O 3 concentration was detected using an UV-absorption analyzer (Model 49C, Thermo Electron Corporation, USA).All measured parameters are listed in Table 3.

Calibrations, limits of detection, standard errors, and precision of trace gas concentration measurements
For the calibration of the NO/NO 2 analyzer (field conditions), a NO standard (5.09 ± 0.1 ppm, Air Liquide, Germany) was applied.The standard was diluted by synthetic air, which had been additionally cleaned with activated charcoal and Purafil ® (Purafil, Inc., USA) to remove any potential NO and NO 2 contaminations.For the dilution of the NO standard a gas phase titration unit was applied (GPT, 146C Dynamic Gas Calibrator, Thermo Electron Corporation, USA).In the GPT, NO 2 calibration gas is produced by titration (see Reaction R1) of the diluted NO standard with O 3 (generated by a UV lamp in the GPT).The BLC's efficiency was determined by the ratio of measured NO 2 and the known value of NO 2 obtained by titration of NO.The O 3 analyzer was calibrated by the GPT-generated O 3 , where the exact O 3 concentration is known from the gas phase titration of the NO standard.To maintain high quality concentration measurements even under long-term field conditions, it was necessary to control and to service the system frequently.In the field, calibrations were performed once a week to ensure stability of the analyzers (quantifying potential drifts), while in the laboratory calibrations were performed just before the start of the experiment.
The determination of the limit of detection (LOD) is particularly important for the exchange measurements of NO and NO 2 , as (very) low concentrations have been encountered under both, laboratory and field conditions.According to MacDougall and Crummett (1980) the "limit of detection" is the lowest concentration level that can be determined to be statistically different from a measurement of "zero" concentration.Here we define LOD (m NO 2 ), LOD (m NO ), and LOD (m O 3 ) as three times that standard deviation (s m NO 2 ,0 , s m NO,0 , s m O 3 ,0 ), which has been obtained through a statistically significant number (laboratory: 360, field: 160-360) of zero-air measurements.In Table 3 the LOD (m i ) of the instruments are summarized.The conversion efficiency of the BLC for NO 2 was around 25 % during laboratory measurements and 32-36.5 % under field conditions.
Besides the determination and rigorous control of the LOD's, the quantification of the analyzers' reproducibility (precision) is still more necessary, as exchange fluxes of the NO-NO 2 -O 3 triad are evaluated from very small differences of concentrations measured at the inlet and the outlet of the dynamic plant chamber.We define the precision of the analyzers as the ratio of the standard errors s m,i and the corresponding concentrations m i (i = NO, NO 2 , O 3 ).The standard errors of NO and NO 2 measurements were found to be a (weak) function of the NO and NO 2 concentrations themselves: where s m NO 2 ,0 and s m NO,0 are the standard errors at m NO 2 = 0 and m NO = 0, b NO 2 and B NO (in nmol −1 m 3 ) have been derived from calibration exercises.

Design and construction
The open (flow through), dynamic chamber system was a further development of the systems operated in previous studies (Schäfer et al., 1992;Kesselmeier et al., 1996;Kuhn et al., 2002).The system was designed for measurements of trace gas exchange in the field with minimal effects on the gases.
The system has been demonstrated to work under field conditions.The design of the chambers is illustrated in Fig. 4 and details of the used materials and parts are listed in Table 4.The chambers had an inner diameter of 40 cm.The height of the chambers could be varied by extending the frame and could be adjusted for the plant specimen.The initial height was 45 cm and we used extensions of 15 cm at field measurements.The chamber frame and the lid were made of PVC and acrylic glass.
The inner walls consisted of a thin transparent Teflon film (FEP).Previous investigations of the spectral transmissivity of the FEP film have shown that photosynthetically active radiation (PAR) nearly completely transmits this film: in the spectral range of PAR (400-700 nm) transmissivity is about 95 %.In the range of λ ≤ 400 nm, the transmissivity of the FEP film is about 90 % (Schäfer et al., 1992;Pape et al., 2009).A consequence of the horizontal installation of the chamber during field measurement is that transmission of the acrylic glass parts of the chamber plays only a minor role.Furthermore, the Teflon film was reported to show no interferences with trace gases tested such as organic acids (Schäfer et al., 1992;Kesselmeier et al., 1997)  and isoprene (Kesselmeier et al., 1996(Kesselmeier et al., , 1997;;Kuhn et al., 2000), and reduced sulfur compounds (Kesselmeier et al., 1993).
The FEP film was fixed with elastic silicone straps around the outer side of the frame.The inner side of the lid was covered by the Teflon film as well.The lid was fixed to the chamber with four clamps.Several holes in the lid allowed the installation of tubes, mixing fans and the intake system of purging air.The purging air flow through the chamber was established in the field by a blowing axial inlet fan which was controlled by an air mass flow sensor installed outside the chamber frame.In the laboratory we used pressurized air for flushing the chamber.For a continuous turbulent mixing of the air inside the chamber a Teflon propeller driven by a magnetically coupled motor attached outside and two Teflon coated mixing fans were used.This design ensured that the air pumped through the chamber only came into contact with parts made of Teflon (PFA or PTFE).For the measurements several chambers were combined (Fig. 5).As in former studies on the NO 2 exchange with different plants, an extra empty ("reference") chamber was also applied.The empty chamber was used to detect basic contamination in the system, adsorption/desorption, as well as to investigate gas-phase chemical reactions within the chamber volume and at the wall surface.A central V25 microprocessor unit (PASCAL based code) controlled the power supply for the mass flow sensors, purging and mixing fans, and signal recording by a PC card.Each chamber could be controlled independently.Furthermore, the V25 operated a number of environmental sensors for air and needle temperature, photosynthetically active radiation (PAR) and relative humidity, and recorded their signals.

Implementation of concentration and flux density measurements
Exchange flux densities of the NO-NO 2 -O 3 triad as well as of CO 2 and H 2 O are determined from the difference of molar concentrations measured at the inlet and outlet of the dynamic chambers.Ideally, a total of 10 analyzers per dynamic chamber would guarantee simultaneous concentration measurements at all these positions.However, full simultaneity is usually prohibited, both for reasons of cost, and because operation of two trace gas analyzers with an agreement (in their absolute accuracy) much less of the expected difference between inlet and outlet concentration is currently not feasible.Therefore, only one set of analyzers was used operating in a mode of continuous switching between the inlet and outlet position(s) of the (different) dynamic chamber(s).For gas piping the tubes from the different positions at the chambers were combined to one insulated and heated (above ambient temperature) bundle to prevent water vapor condensation.To ensure similar conditions for all lines, all tubes were set to the same length (in this field study 37 m).The sampling air flow was maintained by Teflon membrane pumps with an air flow of 8-10 l min −1 .To avoid contamination of tubes and analyzers a PTFE particulate filter (pore size 2 µm) was installed in front of the intake line.Switching between the different intake lines was maintained by several 3-way PFA solenoid valves.The necessary quantity of valves depends on the number of dynamic chambers in operation.The sample line connected the valve block to the analyzers.Even when an individual intake line was not switched to the analyzers, the air flow through it was kept constant.A second V25 unit was used to control the solenoid valves and the cycle times and recorded the data of the trace gas analyzers.Measurement cycle times and switching (during field experiments) is shown in Fig. 12a.The shown cycling time of 4 min is a result of optimization between fast switching and the analyzers' and system's capabilities: the most important issues in this respect are the analyzers' (moving) averaging times of 30 s and the temporal response of the analyzers to switching concentrations.
Air temperature and needle surface temperatures inside the chambers were continuously recorded by Teflon covered thermocouples (0.005 , Chromega TM -Constantan, Omega, UK).PAR was detected outside the chamber with a LiCor quantum sensor (model LI-190SA, LiCor, Lincoln, NE, USA).Relative humidity was measured with a combined temperature and relative humidity probe (Model MP100A, Rotronic, Switzerland).

Laboratory set-up
For laboratory experiments the plant chambers were installed inside a thermostatted cabinet (Heraeus, Germany), which was kept under controlled temperature and humidity conditions (day: 25 • C, 60 %; night: 20 • C, 50 %) with a light/dark regime of 12/12 h.In addition to the cabinet irradiation (Osram Powerstar HQI-BT 400 W/D) we used a set of light emitting diodes with a spectral bandwidth of 400-700 nm.The total measured PAR in the middle of the chamber was about 450 µmol photons m −2 s −1 .The plant chambers were continuously flushed with purified air, obtained by passing compressed air through a gas purification system consisting of several columns in series, filled with silica gel (2-5 mm, Merck, Germany), molecular sieve (0.3 nm perlform, Merck, Germany), charcoal (0.3 mm LS-Labor Service, Germany), and glass wool (Merck, Germany).The purified air was then led through a glass tank filled with demineralized water to humidify where s m blend,NO 2 results of Gaussian error propagation applied to Eq. (13.1); concentrations (and standard errors) of m std,NO 2 , m blend,NO 2 , and m dil,NO 2 are in nmol m −3 , flow rates (and standard errors) of in Q std,NO 2 and Q dil are in m 3 s −1 .For calculation of s m blend,NO 2 it is assumed, that m std,NO 2 is constant (during the time of the laboratory experiment) and m dil is zero.The NO 2 mixture was directed into the dynamic plant chambers (without using the blowing axial inlet fan as for our field studies).For the laboratory measurements one plant chamber and one empty chamber with a volume (V ) of 57 l were used.Each chamber was flushed at a constant flow (Q) of 14 l min −1 , controlled by mass flow controllers (MKS Instruments, USA), resulting in an exchange of the entire chamber's volume every 4 min.For two minutes each, air samples were directed to the analyzers from three different intake lines (purging NO 2 mixture (upstream of the chambers), outlet of empty and plant chambers).All analyzers were placed inside a cabinet (GKPv 6522, Liebherr, Germany) thermostatted at 25 • C to minimize variations of the analyzers' signals caused by temperature fluctuations.

Field site description and set-up
The field experiment was conducted within the project EGER (ExchanGE processes in mountainous Regions).The second intensive observation period (IOP-2) of EGER took place in summer 2008 (1 June-15 July) in the "Fichtelgebirge" (northeast Bavaria, Germany), a mountainous region, covered mainly by forests and arable land (including meadows), and lakes.The research site "Weidenbrunnen" (50 • 08 31 N, 11 • 52 01 E; 774 m a.s.l.) is part of a spruce forest ecosystem, which resulted from intensive re-forestation in the last century.The plant cover is dominated by Norway Spruce (Picea abies L.).The stand-age was 56 yr (according to Alsheimer, 1997) and the mean canopy height was 23 m (Serafimovich et al., 2008).The tree density of the stand was 1007/ha (Alsheimer 1997), with a leaf area index (LAI) of 5.2 (Thomas and Foken, 2007).
For the field measurements we used two dynamic chambers to determine exchange flux densities of two spruce www.atmos-meas-tech.net/5/955/2012/Atmos.Meas.Tech., 5, 955-989, 2012 branches of two different trees.In addition, one empty chamber was operated nearby the plant chambers.The chambers were installed at a height of 13 m above ground (at a 32 m tall tower).The ambient air inlet was mounted at 16 m height.The chambers had a volume (V ) of 75 l, and a constant flow (Q) of 60 l min −1 maintained a continuous and complete air exchange in 75 s.For best performance, all analyzers were placed inside an air-conditioned container on the forest ground close to the tower.All insulated and heated (see above) intake lines were running from the individual positions of the chambers to the container and were of equal length (about 37 m).The four intake lines (ambient air; outlets of plant chamber 1, plant chamber 2, and empty chamber) were sampled consecutively for four minutes each.The measurement cycle was as follows: (1) ambient air, (2) plant chamber 1, (3) reference chamber, and (4) plant chamber 2 (see Fig. 12a).

Plant material
Laboratory experiments were performed with 3-to 4-yr old Norway Spruce trees (Picea abies L.) grown in pots in a commercial soil mixture.All specimens originated from the EGER field site and were dug out half a year before the measurements started.For the laboratory studies the aboveground parts of the whole tree were enclosed in the chamber.A typical young tree had a leaf area (A leaf ) of 0.44 m 2 in total.For the field experiments branches of adult Norway Spruces were investigated.The front part of an intact branch with older needles and new shoots, still attached to the tree, was enclosed to around 40 cm length in the chamber.Two plant chambers on different trees were used for the field studies.At the end of the studies the total enclosed leaf area (two-sided) was measured to be 0.99 m 2 (tree 1) and 1.02 m 2 (tree 2) with a dry weight of 66 g (tree 1) and 78 g (tree 2).For determination of leaf area and dry weight the leaves of the enclosed branches were harvested at the end of experiments.Leaves were scanned by a calibrated scanner system (DeskSCAN II, Hawlett-Packard, USA; area determining software SIZE, Müller, Germany).Dry leaf weight was obtained after drying for two days at 70 • C in an oven (Heraeus, Germany).During the long term field measurements spruces were producing new needles, therefore we estimated the leaf area during measurement time by linear interpolation.The needles of spruce have stomata on the entire needle surface, therefore the area of the whole surface was used.For needle surface area calculation the single surface area was multiplied by factor 2.74 according to Riederer et al. (1988).All exchange measurements started one day after enclosure in order to allow an acclimatization of the branch or plant.

Monitoring of plant-physiological processes
Working with chambers and enclosed plants (parts of plants) necessitates control of the plant living conditions.Chamber operation and design must not disturb plant metabolism.For example an insufficient purging air flow would affect the gas exchange of the plant.An increase of water vapor concentration and a drop of the CO 2 level would trigger a nonphysiological stomatal behavior.Thus, the simultaneous measurement of CO 2 mixing ratios and surface exchange fluxes (assimilation), H 2 O surface exchange fluxes (transpiration) and determination of stomatal conductance were performed to provide an indication of the plant condition.For long term field measurements further comparing measurements with non enclosed plants (or part of the plants) would be advantageous to indicate the potential effects of enclosures.Within this context, measurements of the photosynthetic capacity in response to temperature, radiation, CO 2 mixing ratio and relative humidity or analysis of the nutrient composition of enclosed and control plants are of great help.We validated the photosynthetic capacity of the enclosed needles in comparison to control needles by measurements of in-situ CO 2 and H 2 O needle gas exchange in response to temperature, radiation using a portable gas exchange system (WALZ GFS3000, Walz, Effeltrich/Germany).
After field experiment we could not identify visual differences between enclosed and not enclosed plant material.Moreover, no differences in physiological performance were detectable.Furthermore, analyses of the composition of nutrients of needles were without findings.Detailed results of these analyses will be given in a consecutive publication.

Corrections for concentration changes in long tubing
Long intake lines (mostly necessary for field experiments) may impact the trace gas concentrations (Beier and Schneewind, 1991).Trace gases may ad-or absorb on the inner walls of the tubing, and/or react with each other according Reactions (R1) and (R2) (see Appendix A).Therefore, we used opaque tubing to completely prevent photolysis of NO 2 .Hence, Reaction (R1) (NO + O 3 ) was the most important reaction to consider.For a known residence time, temperature and pressure in the tubes, the mixing ratios of NO, NO 2 , and O 3 can be corrected according to Beier and Schneewind (1991).To proceed, the residence time of the individual trace gas in the tubing as well as the characteristic chemical reaction time (τ i ; i = NO, O 3 ) must be known.The latter is calculated by Atkinson et al., 2004).Tests were carried out to check the response of analyzers to changes of concentrations when switching between intake lines with low concentration of the respective trace gas (NO, NO 2 , O 3 ) to another intake line with high trace gas concentration (after stabilization), and back to the intake line of low concentration.

Temperature dependence of analyzers
The signals of analyzers are sensitive to the surrounding temperature.These effects are of particular importance for field studies where it is more difficult to keep temperatures constant.Thus a series of tests were performed to determine the temperature dependence of all trace gas analyzers.The tests were done inside the conditioning cabinet (Heraeus, Germany) under different temperature conditions (temperature range: 18-46 • C).For each analyzer a calibration was carried out at each temperature level.We considered the correction of the analyzers' signals necessary if the observed drift with temperature exceeded the maximum signal noise measured with zero air.We did not perform a correction when the drift was below 1 % for the entire temperature range or the analyzer's noise was greater than the temperature drift.

Dynamic chamber: internal mixing, exchange rate of chamber volume, wall absorption, and transmissivity
Effective turbulent mixing and fast exchange of the plant chamber's volume are essential for the determination of exchange flux densities of reactive as well as non-reactive trace gases (cf.Meixner, 1994;Meixner et al., 1997).Particularly, the derivation of accurate NO 2 and O 3 leaf conductances from NO 2 and O 3 deposition velocities obtained by dynamic chamber measurements critically depend on the effectiveness of internal mixing and the chamber volume's exchange rate (cf.Pape et al., 2009).Fast internal mixing of the chamber's volume was assured by operation of three fans (see Fig. 4) inside the chamber.A similar procedure was chosen by Pape et al. (2009), who quantified complete mixing of the chamber volume in less than 2 s.The exchange rate of the chamber's volume is primarily determined by the volume V and the purging rate Q.However, due to delay effects of the sampling lines and due to the limited response times of the analyzers after switching between the different intakes, it is not possible to directly observe the trace gas' mixing in the plant chamber.Therefore, the time needed to equilibrate trace gas concentrations in an empty plant chamber was determined by measurements of a fast-response helium detector (Pico leak detector, MKS Instrument Inc., USA).A helium pulse was released into the purging stream of the chamber and the needed time for equilibration was determined.
Sorption effects (ad-, ab-, desorption) to and from the inner wall materials of the dynamic chamber should not modify the concentrations of (reactive) trace gases.Using the laboratory set-up, we investigated potential sorption effects to the inner walls of an empty chamber by fumigating it consecutively with different NO, NO 2 , and O 3 concentrations.There were no desorption effects observed.Wall absorption was quantified in form of "blank" deposition velocities, where v dep wall,i = Q (m a,i − m s,i )/(A wall m s,i ) (i = NO 2 , NO, O 3 ).
In the field, the transmissivity of the FEP film (the dynamic chamber's wall) for PAR and the NO 2 photolysis rate j (NO 2 ) was monitored by continuous and simultaneous measurements of corresponding radiation fluxes inside and outside the chamber.PAR was measured with a LiCor quantum sensor (model LI-190SA, LiCor, Lincoln, NE, USA) and j (NO 2 ) was determined as an omni-directional actinic UV radiation flux using a j (NO 2 )-sensor (filter radiometer, Meteorologie Consult GmbH, Königstein, Germany).

Significance of concentration differences
In the laboratory, the exchange flux density is directly proportional to m i = (m a,i − m s,i ), the difference of trace gas concentrations at the inlet and the outlet of the dynamic chamber (see Eq. 2).Even under field conditions, the major component of the exchange flux density F ex,i is Q/A leaf m i .Keeping in mind, that (a) the sign of m i determines direction of the exchange flux density, and (b) the errors of m a,i and m s,i are decisively controlling the error of m i , (and consequently that of F ex,i ), it is obvious to control the significance of m.The corresponding statistical test requires the number of individual measurements, the averages and standard errors of m s,i and m a,i .These were provided and calculated from the individual concentration measurements during one measurement cycle (laboratory: 30 min, field: 4 min).Prior to this, we identified outliers in the data sets by application of the Nalimov-test, a variant of Grubbs' test.The significance of differentiation between the two averages of m s,i and m a,i was statistically secured by application of the ttest.m with statistical significance below 99 % (α < 0.99) were correspondingly flagged and not included in subsequent calculations.

Regression analysis
Since the concentrations m a,i and m s,i are measured with identical analyzers (see above), corresponding standard errors s ms,i and s ma,i are of the same order of magnitude.Therefore, bi-variate weighted linear least-squares fitting (which considers uncertainties of both, m s,i and m a,i ) is preferred to any standard forms of linear regression analysis (which consider, at best, uncertainties in the y-values, but no uncertainties in the x-values).8.3) must be known, and all variables of each individual equation should be independent of each other.However, the latter is not the case for (at least) m s,i and m a,i (see Eqs. 1.1-1.3).Therefore, application of the generalized form of the Gaussian error propagation is preferred, which considers the mutual dependence of each pair variables (Taylor, 1982;Phillips et al., 2002).The general formulation of the standard error s y of a quantity y = f (x 1 , x 2 , x 3 , ..., x n ) reads as follows: where r (x i ; x j ) are the correlation coefficients between each pairs of all x i and x j .The individual variables x i for the quantities y = F ex,NO 2 ,  8.3).These are listed in Appendix C as well as all the corresponding derivatives necessary to calculate the standard errors of these quantities according to Eq. ( 14).

Significance of the compensation point concentrations
The bi-variate weighted linear least-squares regression analysis of m a,i and m s,i delivers the intercept n i , the slope b i , and their standard errors s n,i and s b,i .According to Eqs. (8.1)-(8.3),each of the compensation point concentrations m comp,i of the NO-NO 2 -O 3 triad can be considered as a random variable, represented by the average of m comp,i and the standard error s m,comp,i .The decision whether or not a compensation point concentration exists is equivalent to the test of the hypothesis whether or not the average of m comp,i is highly significantly (α = 0.999), significantly (α = 0.99), or likely (α = 0.95) different from m * comp,i = 0.For that, it is assumed that each of the test quantities T i matches the t-distribution with N − 1 degrees of freedom.Depending on α, the hypothesis m comp,i = m * comp,i must be rejected, if where t α;N−1 are the values of the t-distribution (N − 1) for α = 0.999, 0.99, 0.95, respectively.

Analyzers and system performance
The results for the test of temperature dependence of all analyzers (see Sect. 3.4.3)are listed in Table 5. Between 18 and 46 • C the efficiency of the BLC drifted from 37.0 % to 47.4 % over the whole temperature range.This means that for an initial concentration of 10 ppb NO 2 a drift of 2.2 ppb over the whole temperature range would be observed, which is equivalent to 3.6 nmol m −3 K −1 (0.08 ppb K −1 ).For NO the signal drift was 2.8 nmol m −3 K −1 (0.07 ppb K −1 ).The data of the CO 2 and O 3 analyzers did not need to be corrected because the signal drift was below 1 % for the entire temperature range, in contrast to the NO and NO 2 values.For the mathematical correction the slope of the regression line of the temperature tests (trace gas concentration versus temperature) was used.
In Fig. 6, the precision (s m,i /m i ) of the concentration measurements is exemplified for NO 2 during laboratory (red curve) and field experiments (green curve).The precision of m NO 2 was only approx.35 % during laboratory experiments at LOD (m NO 2 ) = 1.04 ppb (46.4 nmol m −3 ).Before the field experiment, the performance of the NO/NO 2 analyzer has been considerably improved by increasing the residence time of the air sample in the BLC cell.Consequently the precision at 1 ppb improved to nearly 10 % in the field (however, precision was still 35 % at LOD (m NO 2 ) = 0.31 ppb (13.8 nmol m −3 )).For further comparison, we consider that concentration m i , where corresponding precision curves fall short of the 10 %-precision lines.These concentrations were 161.9 nmol m −3 (3.63 ppb; laboratory conditions), 45.9 nmol m −3 (1.03 ppb; field conditions), and they would be 14.7 nmol m −3 (0.33 ppb) and 1.3 nmol m −3 (0.03 ppb), if analyzers could be applied with LOD (m NO 2 ) = 0.1 and 0.01 ppb, respectively.For the NO and O 3 analyzers applied under field conditions, corresponding NO and O 3 concentrations (<10 % precision) were 15.2 nmol m −3 (0.34 ppb; LOD (m NO ) = 0.10 ppb) and 144.5 nmol m −3 (3.24 ppb; LOD (m O 3 ) = 0.98 ppb), respectively.
The performance of the dynamic chamber system depends critically on the temporal delay of concentrations (measured by only one set of analyzers) which are caused by switching between different intake lines of considerable length and by chemical reactions inside corresponding tubing (see Sect. 3.4.1).The tubing residence time for the 36.5 m long tubes of the field experiment was ≤4.1 s under ambient temperature and pressure conditions, calculated from sample flow (8.5-10 l min −1 ), the length of the tubes, and the tubes' inner diameter (0.00435 m).Since a considerable high flow through the intake filters and the long, thin tubes caused a distinct pressure drop (approx.490 hPa), the actual residence time was consequently shorter (1.9 s).The characteristic chemical time scale (τ chem ; e-fold time) for the NO + O 3 reaction (see Reaction R1) was within 20 < τ chem < 120 s during the entire field experiment.Since τ chem was always much longer than the tubing's residence time, any effects of the NO + O 3 reaction on measured concentrations could be neglected (as well as for the NO 2 + hν Reaction R2, since opaque tubes have been used).However, the flow rate between the valve block (see Fig. 5) and the analyzers is about 1/10 of the tubing purge flow; therefore, the "response time" of the entire system for a sudden change of concentrations was tested.Results are shown in Fig. 7 for NO 2 (step change from 41 to 861 nmol m −3 ).Immediately after switching some typical pressure effects (valves) could be observed, but a temporally stable concentration was reached after 90 s.For the return switch a quite similar effect were observed, and "response times" of NO, O 3 , CO 2 , and H 2 O were comparable (data not shown).Based on these tests, the first 90 s of each concentration measurement were skipped from further data processing.
The temperature difference between inside and outside the plant chamber was 1.53 ± 0.98 K for the entire field experiment.

NO 2 blending for fumigation experiments
For laboratory NO 2 fumigation experiments very low (ppband sub-ppb levels) and temporally stable NO 2 concentrations have to be made available.That is essentially necessary to significantly identify any NO 2 compensation point whose concentrations are expected at these low concentration levels.Blended NO 2 concentrations (m blend,NO 2 ) of 13.4, 26.8, 44.6, 80.3, and 151.7 nmol m −3 (0.3, 0.6, 1.0, 1.8, 3.4 ppb) were provided by diluting an NO 2 standard into purified air (see Sect. 3.3.3).A typical course of these concentrations are shown in Fig. 8, where the vertical dashed lines indicate times where blending was changed to obtain the next NO 2 concentration.A stable signal of the new NO 2 concentration level was reached after max.60 min.Fluctuation of the blended NO 2 concentration was between 8.0 and 16.1 nmol m −3 (0.18-0.36 ppb).These fluctuations do not depend on the analyzers' temperature (see Sect. 4.1).During laboratory measurements, the temperature variation of the instrument was only ±0.5 • C, which would be equivalent to a change of m blend,NO 2 = 44.6 nmol m −3 (1 ppb) of less than 1 %.The measured fluctuations could be also due to the precision of m blend,NO 2 which depends on the precision of the applied mass flow controllers.According to the manufacturer, the precision of the mass flow controllers is ±0.8 % of full scale.Using this information, the precision of m blend,NO 2 has been calculated through Eqs.(13.1) and (13.2) and is also shown in Fig. 6.Uncertainty of the mass flow controllers may have added <20 % to the observed variation of measured the blended NO 2 concentration.Consequently application of future NO/NO 2 analyzers (lower detection limit (3σ ) < 2.2 nmol m −3 (<0.05 ppb) will be useless, unless the uncertainty of the NO 2 blending for fumigation experiments is improved significantly.

Radiation and NO 2 photolysis rate
Transmissivity of PAR through the chamber walls (FEP film) is a fundamental requirement if the plant is not to be affected by the chamber itself.Moreover, the calculation of the exchange flux density F ex,i (see Eqs. 1.1-1.3)has to consider the NO 2 + hν reaction.For this, the photolysis rate j (NO 2 ) inside the chamber volume has to be known.Therefore the transmissivity was quantified by simultaneous measurements inside and outside the chamber.While PAR was 10 % lower inside the chamber than outside, j (NO 2 ) was 30 % lower inside the chamber (Fig. 9).Therefore, 70 % of ambient j (NO 2 ) was used for the calculations of F ex,i , v dep,i , m comp,i and their standard errors.

Sorption effects and chamber volume exchange time
An empty dynamic chamber has been exposed to various concentrations of NO 2 , NO, and O 3 and "blank flux densities" have been determined according to Eq. ( 2)."Blank flux densities" for NO, NO 2 , and O 3 are listed in Table 6.They were always negative (i.e.no desorption from the chamber's inner surfaces) and revealed very low values.Expressed in corresponding "wall deposition velocities" −2.12 × 10 −3 (NO), −2.92 × 10 −3 (NO 2 ), and −1.94 × 10 −3 mm s −1 (O 3 ) were found.These values were two orders of magnitude lower than v dep,i observed under laboratory as well as under field conditions.Comparing incoming and outgoing concentrations of the NO-NO 2 -O 3 triad, a maximum of 2 % of the trace gases may have been absorbed by the inner surfaces of the plant chamber.Therefore, with regard to the mass balance of the dynamic plant chamber, neglecting of any mass fluxes to the walls of the chamber ( wall,i ) (see Appendix A) is justified.
Table 6.Parameters of sorption effects to the inner chamber walls determined by laboratory experiments.q 10 and q 90 denote the 10 % and 90 % quantiles of the entire blank flux density F wall,i data, concentration ranges represent applied fumigation concentrations during the experiment, c mean denotes the mean concentration difference of incoming and outgoing chamber air in % (range of differences in %).
F wall,i , pmol m −2 s −1 concentrations gas mean (±σ ) q 10 ...q 90 v dep wall,i , m s The chamber volume exchange time was determined from an experiment, where a short pulse of (chemically inert) helium has been added to the purging flow of the dynamic chamber (see Sect. 3.4.4).Results are shown in Fig. 10.For the time of complete exchange (i.e. a constant level of He is observed), we used the time interval to reach 98 % of the final He concentration (t 98 ).Due to the limited temporal resolution of the He detector (5 s), t 98 might have been between 80 and 85 s.This result was similar to the time (79 s) calculated from chamber volume (V = 79 l) and purging rate (Q = 60 l min −1 ).

NO 2 exchange flux density: laboratory results
Here, we confine ourselves to the results of "daytime" experiments, i.e. fumigation of the 3-to 4-yr old Norway Spruce trees with 13 < m a,NO 2 < 152 nmol m −3 (0.3-3.4 ppb), controlled temperature (25 • C), relative humidity (60 %), and PAR (450 µmol photons m −2 s −1 , for 12 h) conditions.During experiment no significant difference of m O 3 or m NO between reference and plant chamber could be detected, and the amount of j (NO 2 ) inside the chamber was negligible with respect to any measurable effects due to Reaction (R2).As shown in Sect.4.1, the performance of the NO 2 analyzer was definitely sub-optimal (LOD (m NO 2 ) = 1.04 ppb; 3 σdefinition).Therefore, we based our evaluations of F ex,NO 2 , v dep,NO 2 , and m comp,NO 2 on a 2 σ NO 2 detection limit (28.5 nmol m −3 or 0.6 ppb) for the observed concentrations (m a,NO 2 , m s,NO 2 ).A total of 51 pairs of m a,NO 2 and m s,NO 2 have been obtained during the fumigation experiments.17 data pairs passed the LOD (m NO 2 ) criterion, where another three of them had to be rejected due to the significance criterion for m NO 2 = (m a,NO 2 − m s,NO 2 ).Fourteen data pairs of m a,NO 2 and m s,NO 2 have been subjected to a bi-variate weighted regression analysis (see Sect. 3.4.6),which resulted in R 2 = 0.9706, n 1 = 1.7 ± 2.63 nmol m −3 , b 1 = 0.71 ± 0.035, v dep,NO 2 = 0.22 ± 0.013 mm s −1 , and m comp,NO 2 = 5.9 ± 9.13 nmol m −3 .The significance probability of m comp,NO 2 = 0 is 47 % ("unlikely").NO 2 exchange flux densities (F ex,NO 2 ) and their standard errors have The chamber (V = 0.079 m 3 ) was 2 operated with purging air flow rate Q = 60 L min -1 .The red lines represent start and end of the 3 helium addition, the black dashed line marks the end of equilibration.For the approximation 4 of a complete exchange we used the time interval for 98 % approximation (t98).5 Fig. 10.Results of the response time test with helium.The chamber (V = 0.079 m 3 ) was operated with purging air flow rate Q = 60 l min −1 .The red lines represent start and end of the helium addition, the black dashed line marks the end of equilibration.For the approximation of a complete exchange we used the time interval for 98 % approximation (t 98 ).
been calculated according to Eq. ( 14) and are shown in Fig. 11. Figure 11a displays results of F ex,NO 2 where the 2σ -LOD (m NO 2 )-definition, Fig. 11b where the 1σ -LOD (m NO 2 )-definition has been applied.Furthermore, in both panels F ex,NO 2 data were separated for the significance of m NO 2 (significant: blue circles, non-significant: reddish diamonds); the (F ex,NO 2 ; m s,NO 2 )-regression lines have been calculated according to Eq. ( 9) for all F ex,NO 2 data (pink line), and for those F ex,NO 2 data, where m NO 2 is significant (blue line).Corresponding NO 2 compensation point concentrations m comp,NO 2 were calculated according Eq. ( 11) and are represented by red filled circles (significant m NO 2 ) and pink hollow circles (all data).Details of statistical evaluation are listed in Table 7. Applying the simple linear least-square fitting algorithm (without considering s m a,NO 2 nor s m s,NO 2 ) m comp,NO 2 is always highly significant (regardless of which LOD (m NO 2 )-definition is applied), even if only those F ex,NO 2 data are considered where m NO 2 is significant m comp,NO 2 remains still significant.However, applying linear leastsquare fitting algorithms www.atmos-meas-tech.net/5/955/2012/(2), their standard errors according to Eq. ( 14).Blue circles identify F ex,NO2 where m s,NO2 > LOD(m s,NO2 ), white circles stand for F ex,NO2 where m s,NO2  LOD(m s,NO2 ), and reddish diamonds for those F ex,NO2 data, which have to be rejected for non-significance of m NO2 = (m a,NO2  m s,NO2 ).Blue line (considering blue circle data) and pink line (considering blue circle and reddish diamond data) are calculated according to Eq. ( 9).NO 2 compensation point concentration m comp,NO2 is calculated according to Eq. ( 11) and is represented by red filled circles (considering blue circle data) and pink hollow circles (considering blue circle and reddish diamond data).More details of statistical evaluation are listed in Table 6.(2), their standard errors according to Eq. ( 14).Blue circles identify F ex,NO 2 where m s,NO 2 > LOD (m s,NO 2 ), white circles stand for F ex,NO 2 where m s,NO 2 ≤ LOD (m s,NO 2 ), and reddish diamonds for those F ex,NO 2 data, which have to be rejected for non-significance of m NO 2 = (m a,NO 2 − m s,NO 2 ).Blue line (considering blue circle data) and pink line (considering blue circle and reddish diamond data) are calculated according to Eq. ( 9).NO 2 compensation point concentration m comp,NO 2 is calculated according to Eq. ( 11) and is represented by red filled circles (considering blue circle data) and pink hollow circles (considering blue circle and reddish diamond data).More details of statistical evaluation are listed in Table 6.which consider either s m s,NO 2 , or s m a,NO 2 and s m s,NO 2 , the existence of m comp,NO 2 becomes mostly "unlikely".With the exception of applying the 2σ NO 2 detection limit to all F ex,NO 2 data, the impact of different statistical treatments on the evaluation of NO 2 deposition velocities is small (0.19 ≤ v dep,NO 2 ≤ 0.22 mm s −1 ).

NO-NO 2 -O 3 exchange flux densities: field results
In Fig. 12 typical time series of trace gas mixing ratios are shown, measured at two different spruce branches during the EGER field campaign.The observed mixing ratio changes were due to switching between the different intakes.After switching, concentrations showed the delay effects mentioned above (see Sect. 4.1).Due to this, the first 90 s after valve switching were skipped from subsequent data processing (these first 90 s interval indicated as grey shaded vertical bars in Fig. 12).Values for CO 2 and H 2 O were measured as the difference between empty chamber and each switched intake.The temporal variation of CO 2 and H 2 O concentrations of the plant chambers versus ambient air or empty chamber represented the physiological activity of the plants, since the CO 2 exchange flux density represents the photosynthetic CO 2 assimilation and the H 2 O flux density the transpiration of the enclosed plant parts.
During the field experiment nearly 3000 pairs of m a,i and m s,i have been obtained.Applying the LOD (m i ) (3 σ -definition) and the significance criterion for m i = (m a,i − m s,i ), around 60 % of the NO 2 data pairs remained.In Table 8 the details of the data pairs selection for both trees are listed for NO, NO 2 , and O 3 .Classification according to measurements during day and night demonstrated, that during night fewer data pairs were distinguishable from each other, especially those of NO. Between the spruce branches in both sampling chambers no differences were noticeable.
After classification of all individual concentration data into different categories of leaf conductance (approx.identical to different categories of radiation conditions), bi-variate weighted regression analysis between classified pairs of m a,i and m s,i was performed (see Sect. 3.4.6).The data pairs were additionally screened for singular concentration peaks of NO, NO 2 , and O 3 , which mainly occurred due to advection of automobile exhaust gases from a busy country road (2000 cars h −1 ) in a distance of about 1-2 km from the site.The problem of advection at this field site is well known, and has been documented through profile measurements of inand above canopy concentrations, as well as through eddy covariance flux measurements of NO-NO 2 -O 3 performed simultaneously to our dynamic chamber measurements (Plake, 2009).For the analysis of dynamic chamber derived O 3 flux densities, we assumed m comp,O 3 = 0 (n O 3 = 0), since emissions of O 3 from plants are not known so far.
For O 3 the analysis resulted in R 2 (m a,O 3 , m s,O 3 ) = 0.9847, b 3 = 0.80 ± 0.005, and v dep,O 3 = 0.32 ± 0.018 mm s −1 .In Fig. 13a (Fig. 14a), results of bi-variate weighted regression analysis between m a,NO 2 and m sNO 2 (m a,O 3 and m sO 3 ) Table 8.Percentage of data m i (i = NO, NO 2 , O 3 ) above limit of detection (LOD (m i ), 3 σ -definition) and significant differences m i = (m a,NO 2 − m s,NO 2 ) of tree 1 and 2 for field measurements.tree 1 tree 2 m i > LOD + significant m i % of total m i > LOD + significant m i % of total (number of total) (number of total) all ( 2988) day ( 1885) night( 1103) all ( 2993) day ( 1887  and reddish diamonds for those data pairs, which have to be rejected for non-significance of mNO2 = (ma,NO2  ms,NO2).Blue line (considering 17 blue circle data) is calculated according to bi-variate weighted linear least-squares fitting regression analysis (see Sect. 3.4.6).(b) NO2 exchange 18 flux density (Fex,NO2) vs. NO2 concentration measured at the outlet of the dynamic plant chamber (ms,NO2).Fex,NO2 data were calculated according 19 Eq. (1.1), their standard errors according to Eq.( 14).Reddish diamonds stand for those Fex,NO2 data, which have to be rejected for non-20 significance of mNO2 = (ma,NO2  ms,NO2).Blue line (considering blue circle data) and pink line (considering blue circle and reddish diamond 21 data) are calculated according to Eq. ( 6 ) and reddish diamonds for those data pairs, which have to be rejected for non-significance of m NO 2 = (m a,NO 2 − m s,NO 2 ).Blue line (considering blue circle data) is calculated according to bi-variate weighted linear least-squares fitting regression analysis (see Sect. 3.4.6).(b) NO 2 exchange flux density (F ex,NO 2 ) vs. NO 2 concentration measured at the outlet of the dynamic plant chamber (m s,NO 2 ).F ex,NO 2 data were calculated according Eq. (1.1), their standard errors according to Eq.( 14).Reddish diamonds stand for those F ex,NO 2 data, which have to be rejected for non-significance of m NO 2 = (m a,NO 2 − m s,NO 2 ).Blue line (considering blue circle data) and pink line (considering blue circle and reddish diamond data) are calculated according to Eq. (6.1).are shown, while in Fig. 13b (Fig. 14b) those of F ex,NO 2 (F ex,O 3 ) versus m s,NO 2 (m s,O 3 ).In Fig. 13a and b, data can be individually identified for their significance of m NO 2 by corresponding color coding.For O 3 , there is no corresponding color coding, since all m O 3 were significant (see Table 8).Linear relationships between F ex,NO 2 and m s,NO 2 were calculated by Eq. (6.1) for data pairs owing significant m NO 2 and for all data pairs.In Table 9 all results of statistical analysis of F ex,NO 2 and F ex,O 3 data are listed.Results of bi-variate weighted regression analysis for NO are shown in Fig. 15.A large part of m NO was lower than LOD (m NO ) (grey diamonds) or corresponding data pairs were nonsignificant with respect to m NO = (m a,NO − m s,NO ) (reddish diamonds).The regression coefficient R 2 (m a,NO , m s,NO ) was only 0.5355.Therefore, consecutive analyses are biased: probabilities of significant m comp,NO and v dep,NO becomes unlikely (51.7 and 22.4 %, respectively).Hence, there were no further evaluations for F ex,NO , v dep,NO , and m comp,NO .

Effects on enclosed plants
Enclosing plants or parts of plants in a dynamic chamber requires the control of plant conditions in order to be sure that observations and data are not created under artificial conditions and consequently transferable to the normal environment.It is important to make sure that the plant is not affected by the chamber, especially for long-term studies.Consequently, we checked the status of the plants after field experiment.In most chamber studies plant conditions were monitored just by measuring the CO 2 and H 2 O exchange of the plant(s) and these values were used to calculate  vs. O3 concentration measured at the outlet of the dynamic plant chamber (ms,O3).Fex,O3 data were calculated according Eq. (1.3), their standard errors according to Eq. ( 14).Dark red line is calculated according to Eq. (6.3).(1.3), their standard errors according to Eq. ( 14).Dark red line is calculated according to Eq. (6.3).
corresponding leaf conductances (e.g.Thoene et al., 1996;Sparks et al., 2001;Geßler et al., 2002).These measurements allow quantification of the actual photosynthesis and transpiration rates of the enclosed plants.However, to check for a potential effect of the enclosure on the plant control measurements (e.g.photosynthesis and transpiration rates, nutrient content) on enclosed and comparable non-enclosed parts of the plant are necessary.Some elemental analyses of the needles were previously done by Rennenberg et al. (1998), but rather to secure a sufficient initial nutrient supply of the plants than to control effects of the chamber on the nutrient conditions during the experiments.

Overview of previous NO 2 exchange flux measurements using dynamic plant chambers
Table 10 shows a list of past dynamic chamber studies that have focused on NO 2 exchange between different plant species and the atmosphere.Most of these measurements were made with NO 2 converters which were not specific for NO 2 detection.Some authors used heated molybdenum converters (Thoene et al., 1991(Thoene et al., , 1996;;Teklemariam and Sparks, 2006;Raivonen et al., 2009), heated ferrous sulphate converters (Rondón et al., 1993;Rondón and Granat, 1994), or a detector based on chemiluminescence on liquid surfaces (Hanson et al., 1989;Hereid and Monson, 2001;Sparks et al., 2001).All these converters overestimate NO 2 concentrations because of interferences with other (oxidized) nitrogen compounds (see Sect.Fig. 15.Field measurements: NO concentration measured at the outlet of the dynamic plant chamber (m s,NO ) vs. NO concentration measured at the inlet of the dynamic plant chamber (m a,NO ).Light green circles identify data pairs for significance of m NO = (m a,NO − m s,NO ), reddish diamonds stand for those data pairs, which have to be rejected for non-significance of m O 3 and grey diamonds for data pairs where m NO ≤ LOD (m NO ). Green line (considering green circle data) is calculated according to bivariate weighted linear least-squares fitting regression analysis (see Sect. 3.4.6).
converter guarantees the interference-free determination of low NO 2 concentrations.
During most of the field studies filtered air was used for purging the dynamic chambers.In most cases, this air was free of O 3 and NO x , and known NO 2 concentrations were delivered to the dynamic chamber by diluting standard mixtures of NO 2 from a cylinder (Geßler et al., 2000(Geßler et al., , 2002;;Sparks et al., 2001;Hereid and Monson, 2001).Some studies additionally controlled the CO 2 and water vapor concentrations of the purging air, the irradiance and temperature conditions inside the chamber (Hereid and Monson, 2001;Sparks et al., 2001).Filtered and/or synthetic air (i.e.home-made H 2 O and CO 2 concentrations, free of non target reactive trace gases) hardly represents ambient air.Therefore, a potential influence on the physiological behavior of the plant cannot entirely be excluded.
For field measurements of the NO-NO 2 -O 3 triad under ambient conditions, fast gas phase reactions inside the chambers must be considered.Therefore, NO, NO 2 , and O 3 concentrations have to be measured simultaneously, even if only one of the trace gases is of interest (Pape et al., 2009).All previous field studies described corrections of the calculated exchange flux densities not in detail.Rondón et al. (1993) specified some corrections for measured NO concentrations only, although O 3 and UV radiation were present in their dynamic chamber.In those cases where measurements of exchange flux densities were performed applying a simultaneously operated empty chamber (as "reference" chamber), corresponding flux densities were calculated from the concentration differences m NO 2 between the outlet of the plant and empty chambers, respectively.This allowed a certain correction for chamber specific wall absorption and/or desorption processes (Geßler et al., 2000(Geßler et al., , 2002;;Raivonen et al., 2009).However, this procedure may not rule out adverse effects of fast gas-phase reactions on the evaluated flux densities, deposition velocities, and compensation point concentrations (see below).

Precision and data quality
As shown in Sect.4.1, the precision of NO 2 concentration measurements of our NO 2 analyzer improves from 35 % (at its limits of detection) rapidly to <10 % at 162 nmol m −3 (3.63 ppb; laboratory) and 46 nmol m −3 (1.03 ppb; field).In Sect.2.1 we presented the expected precision of the NO 2 exchange flux density for NO 2 concentrations up to 200 nmol m −3 , for pre-scribed m comp,NO 2 = 67 nmol m −3 (1.5 ppb), pre-scribed NO 2 deposition velocities (0.3-0.6 mm s −1 ), and typical R 2 (m a,NO 2 , m s,NO 2 ) ranging from 0.99 to 0.9 (see Fig. 3).Since F ex,NO 2 approaches zero at m s,NO 2 = m comp,NO 2 , the exchange flux density's precision (σ Fex,NO 2 /F ex,NO 2 ) will become indefinite there.Consequently, the uncertainty of F ex,NO 2 will become as higher as closer m s,NO 2 approaches m comp,NO 2 (from either side).Analogously to the results shown in Fig. 3, we determined which NO 2 concentration difference, ± m s,NO 2 − m comp,NO 2 , will be necessary to keep the NO 2 exchange flux density's precision for our NO 2 analyzer under 10 %.For laboratory conditions (LOD (m NO 2 ) = 45 nmol m −3 or 1.01 ppb), this difference was ±13.8 nmol m −3 or ±0.31 ppb (v dep,NO 2 = 0.6 mm s −1 ; R 2 (m a,NO 2 , m s,NO 2 ) = 0.99), and ±91 nmol m −3 or ±2.05 ppb (v dep,NO 2 = 0.3 mm s −1 ; R 2 (m a,NO 2 , m s,NO 2 ) = 0.9).During the EGER field experiment (LOD (m NO 2 ) = 13.8 nmol m −3 or 0.31 ppb) corresponding values were ±4.5 and ±8.5 nmol m −3 (0.1 and ±0.19 ppb), respectively.A serious consequence of these calculations is, that, for a given detection limit, there is a well defined limit of m comp,NO 2 where the NO 2 compensation point concentration can be inferred from flux density data (σ Fex,NO 2 /F ex,NO 2 ≤ 10 %) by interpolation of data measured on both sides of m comp,NO 2 .Below that limit, due to the obvious conflict of the requested m s,NO 2 − m comp,NO 2 and LOD (m NO 2 ), m comp,NO 2 can only be inferred from flux density data at m s,NO 2 > m comp,NO 2 by extrapolation, owing the risk of (much) higher uncertainties.These limits were for our NO 2 analyzer 33.5 and 133.8 nmol m −3 (0.75 and 3.0 ppb; laboratory) and 13.4 and 44.6 nmol m −3 (0.3 and 1.0 ppb; field) for the above mentioned combinations of v dep,NO 2 and R 2 (m a,NO 2 , m s,NO 2 ).
In previous studies the NO 2 sensitivity (a proxy for precision) of corresponding NO x or NO 2 analyzers has   Thoene et al. (1991Thoene et al. ( , 1996) ) 10).Neubert et al. (1993) and Geßler et al. (2000), who used analyzers equipped with photolytic NO 2 converters mentioned a LOD (m NO 2 ) of 4.5 nmol m −3 (0.1 ppb); however, the corresponding definition of LOD (1σ , 2σ or 3σ of σ NO 2 ,0 ) is not reported.Based on the manufacturer's data of the analyzers and on our experience, we assume that the reported values correspond to the 1σ -definition (P = 0.68).This assumption is in agreement with the values of Rondón and Granat (1994), who have used the same NO 2 analyzer model, namely with LOD (m NO 2 ) = 8.9 nmol m −3 (0.2 ppb; 2σ -definition).Using the same LOD-definition (2σ ), Rondón and Granat (1994) reported a four times lower LOD for NO of 2.2 nmol m −3 (0.05 ppb).Weber and Rennenberg (1996a,b) using also a photolytic NO 2 converter, have not reported any specifications about their instrument's sensitivity; therefore, we assumed that, based on the manufacturer's information about the applied NO/NO 2 analyzer, the LOD for NO was 3.3 nmol m −3 (0.075 ppb; 3σ -definition).According to Rondón and Granat (1994), and based on our experience the corresponding LOD for NO 2 can be assumed to have not been better than 10 nmol m −3 (0.225 ppb; 3 × LOD (m NO )).Using the results of our simulation of the minimum detectable NO 2 compensation point concentration (see Sect. 2.2), we can state that NO 2 compensation point concentrations ≥44.6 nmol m −3 (≥1 ppb) can be detected with high significance, if NO 2 analyzers with LOD (m NO 2 ) ≈ 13.4 nmol m −3 (0.3 ppb) were used (as Weber and Rennenberg, 1996a;Geßler et al., 2002) and R 2 (m a,NO 2 , m s,NO 2 ) was in a typical range (0.9-0.99) of laboratory measurements.Using NO 2 analyzers with LOD (m NO 2 ) ≈ 44.6 nmol m −3 (≈1 ppb; e.g.analyzers with molybdenum converters) the significant detection of m comp,NO 2 > 44.6 nmol m −3 (1 ppb) would already be difficult, if the v dep,NO 2 is very small (<0.3 mm s −1 ).For example, Thoene et al. (1996) reported m comp,NO 2 = 73.1 nmol m −3 (1.64 ppb) which has most likely be detected with high significance, because they reported v dep,NO 2 = 0.8 mm s −1 .On the other hand, the detection of m comp,NO 2 = 13.4-31.2nmol m −3 (0.3-0.7 ppb; Rondón et al., 1993) at v dep,NO 2 = 0.8 mm s −1 seems now, from a statistical point of view, to be unlikely.
The data quality of exchange flux densities requires the control of quantifiable parameters of the measurement technique.To these belong the results of regular calibrations of the applied analyzers, their detection limits and those parameters which quantify the dependence of the analyzers' signals from other external factors like the ambient temperature.Our studies showed that the temperature dependence of the applied chemiluminescence NO/NO 2 analyzer cannot be neglected (0.08 ppb K −1 ).Hence, constant ambient temperature is definitely necessary to operate the analyzers at the requested level of precision.For our laboratory experiments we solved this problem with a commercial thermostat housing for the analyzers.During field experiments this may be not always feasible.There, we used an air conditioning system for the entire instruments' shelter (container).Since the still remaining fluctuations of temperature were large enough to affect the precision of the NO/NO 2 analyzer, we corrected the analyzer's signals (see Sect. 4.1) It should be stated, that all mentioned previous studies on NO 2 exchange flux densities have even not mentioned this problem.
Laboratory measurements at very low concentrations demand low and stable blended NO 2 concentrations for fumigation of the plants.During our experiments we observed substantial fluctuations of the blended NO 2 concentration which entered the dynamic plant chamber.These fluctuations were due to the blending procedure (and the limited sensitivity of the NO/NO 2 analyzer).As shown in Fig. 6 (blue line), the noise of NO 2 concentrations caused by the blending procedure itself will substantially affect the precision of the NO 2 concentration measurements (and consequently those of NO 2 flux density), particularly if the detection limit of future NO 2 analyzers will be improved to be better than 10 nmol m −3 (0.25 ppb).Then, the improved precision of the NO 2 concentration measurements will fall short of the noise of the blended NO 2 concentration at the inlet of the dynamic chamber (see Fig. 6) and the improvement of the blending procedure (e.g. by application of more precise flow controllers) will become necessary.

Significance of concentration differences
The error of NO 2 exchange flux density measurements by the dynamic chamber method mainly depends on the error of trace gas concentration differences, m i , between the inlet and the outlet of the dynamic plant chamber.In contrast to laboratory conditions, NO 2 concentrations in the field were relative high and rarely conflicted LOD (m NO 2 ).However, during field measurements about 30 to 40 % of daytime m NO 2 data were found to be not significantly different from zero (Table 8) and had to be rejected from further analysis.This rather high percentage of rejected data was mostly due to the temporal variation of ambient NO 2 concentration (m a,NO 2 ) during the 4 min measurement interval, rather than due to the precision or to LOD (m NO 2 ).Ambient NO 2 mixing ratio can rapidly change due to the spatially and temporally varying sources within area surrounding the site of measurements (nearby country roads).In our laboratory studies the percentage of non-significant m NO 2 "daytime" data was 37 % for m a,NO 2 < 44.6 nmol m −3 (1 ppb) and vanished for m a,NO 2 ≥ 71.4 nmol m −3 (1.6 ppb).
In some of the previous studies means or data sets were compared for significant differences by analysis of variance (e.g.Weber and Rennenberg, 1996a,b;Hereid and Monson, 2001;Sparks et al., 2001).However, actual numbers on significant m NO 2 were not reported.We like to emphasize, that (1) our approach to apply a significance test on the measured concentrations directly is rather novel, and (2) the control of the significance of m NO 2 is one of the fundamental quality control criteria for highly significant NO 2 exchange flux densities, NO 2 deposition velocities, and above all the detection of highly significant NO 2 compensation point concentrations.When using data without significance control of m NO 2 , NO 2 compensation point concentrations will be overestimated.

Photo-chemical reactions in the dynamic plant chamber: impact on net exchange flux densities, deposition velocities, and compensation point concentrations
In the previous studies mentioned above, the impact of photochemical reactions was for the most part not considered, neither for the calculation of v dep,NO 2 nor for that of m comp,NO 2 .Not all components of the NO-NO 2 -O 3 triad were always measured.This was related to the fact that most field studies have not used ambient air as purging air.Instead, ambient air was filtered to remove reactive trace gases, particularly O 3 and NO x .Afterwards, the desired NO 2 concentration was blended (e.g.Geßler et al., 2000).Use of filtered air, free of NO and O 3 , allows Reaction (R1) to be neglected, but photolysis of NO 2 (Reaction R2) will still occur, as soon as appreciable amounts of j (NO 2 ) are present in the plant chamber.Consideration of photo-chemical reactions, like the NO 2 loss by Reaction (R2) and the formation of NO 2 by Reaction (R1) were mentioned by Neubert et al. (1993), the production and destruction of NO by Rondón et al. (1993).
With the framework of equations developed in Sects.2.1 and 2.2, we provide a straightforward tool to examine the impact of photo-chemical reactions on the determination of exchange flux densities, deposition velocities, and compensation point concentrations.While actual F ex,i , v dep,i , and m comp,i are described by Eqs.(6.1)-(6.3),(7.1)-(7.3),and (8.1)-(8.3), the quantities F * ex,i , v * dep,i , and m * comp,i are given by Eqs. ( 9)-( 11).The latter are the quantities, which would have been observed if no photo-chemical reactions had taken place (e.g. for NO 2 during our laboratory experiments, see Sect.4.4.1).
In previous experiments, where photo-chemical reactions have not been considered, the actual exchange flux densities F ex,i have been substituted by F * ex,i alone.During some of the more recent experiments photo-chemical reactions were either (partially) excluded by corresponding set-ups or were taken into consideration by application of the "empty chamber (reference chamber) approach" (Rondón et al., 1993;Geßler et al., 2000Geßler et al., , 2001;;Hereid and Monson, 2001;Sparks et al., 2001;Raivonen et al., 2009).However, photo-chemical reactions within the latter chamber will be definitely different from those in the dynamic plant chamber, simply for the fact, that neither j (NO 2 ), nor m s,NO 2 , m s,NO , or m s,O 3 are identical in both chambers.In order to examine potential under/overestimation of simple "chamber flux densities" F * ex,i , by neglecting NO-NO 2 -O 3 gas-phase production and Fex,i for NO (green diamond), NO2 (blue diamond) and O3 (orange diamond).Results are from 3 the field experiment, restricted to one selected data category (see Sect. 4.4.2).The apexes of 4 the diamonds represented the upper (75 %) and the lower (25 %) quantile and the black dash 5 in the middle of the diamonds the median.Fgas,NO and Fgas,NO2 were applied to the left y-axis 6 and Fgas,O3 to the right y-axis.7 8 Fig. 16.Relative contribution of gas-phase flux densities F gas,i to the exchange flux densities F ex,i for NO (green diamond), NO 2 (blue diamond) and O 3 (orange diamond).Results are from the field experiment, restricted to one selected data category (see Sect. 4.4.2).The apexes of the diamonds represented the upper (75 %) and the lower (25 %) quantile and the black dash in the middle of the diamonds the median.F gas,NO and F gas,NO 2 were applied to the left y-axis and F gas,O 3 to the right y-axis.
destruction fluxes, we combine the mentioned equations to obtain: Whether actual exchange flux densities F ex,i are higher, equal or lower than corresponding F * ex,i depends whether the difference of the corresponding gas-phase destruction and production fluxes (second term, right hand side of Eqs.17.1-17.3)is positive, negative and different from zero.
If we differentiate our calculated exchange flux densities F ex,i of the field experiment into the (chamber) flux densities F * ex,i and the gas-phase flux densities F gas,i , which comprised the gas-phase production and destruction of NO-NO 2 -O 3 , we can identify the fraction of F gas,i , of each F ex,i .For the selected leaf conductance category (see Sect. 4.4.2), the percentage of F gas,i is displayed in Fig. 16 for NO, NO 2 and O 3 .The fraction of F gas,O 3 at the exchange flux density of O 3 is very small (±1 %); therefore, it can be neglected.For the NO 2 exchange flux density the fraction of F gas,NO 2 becomes much more important.The median contribution of F gas,NO 2 to F ex,NO 2 was just +8 %, but in particular cases it could be +22 % or −12 %, respectively.Quite clear becomes the impact of the gas-phase reactions for the NO exchange where the quantities with the superscript " * " are those which be derived from using "chamber flux densities" F * ex,i instead of actual exchange flux densities F ex,i .The actual deposition velocities v dep,i are in any case lower than v * dep,i with the exception m s,O 3 = 0, m s,NO = 0, and j (NO 2 ) = 0 (i.e. during nighttime).To examine how much the gas-phase reactions will affect v dep,i , we split our calculated deposition velocity v dep,i for the field data into v * dep,i and the complementary part caused by gas-phase reactions.The contribution of photolysis (see Eq. 18.1) to v dep,NO 2 was 80 %, that of Reaction (R1) on v dep,O 3 only 3 %.Corresponding estimates on v dep,NO were not performed, since NO deposition velocities were not significant during the EGER field experiment.For their experimental conditions, Neubert et al. (1993) identified an error of about 20 % for their v dep,NO 2 determination, if they would neglect photolysis of NO 2 .However, our results should be compared to those of previous studies with caution: in most of the previous studies it is not clear whether the photolysis of NO 2 was correctly taken into account.Nevertheless, we tried to estimate the potential impact of NO 2 photolysis on these, previously reported v dep,NO 2 .For that, the quantities A leaf , V , j (NO 2 ), v dep,NO 2 , and transmissivity of used chamber material have to be a priori known or they must be derived from other (accompanying) data.We made an educated guess of A leaf on available accompany data.The transmissivity for the wavelength range of j (NO 2 ) was estimated on basis of available material information.Thoene et al. (1991Thoene et al. ( , 1996) ) and Geßler et al. (2002) used borosilicate glass (Schott Glaswerke, Mainz, Germany) with an estimated material transmissivity of 60 % and Rondón et al. (1993) used FEP-Teflon film with estimated material transmissivity of 70 %.The j (NO 2 ) values ranged between 6.02 × 10 −4 and 3.48 × 10 −3 s −1 .In summary: if actual NO 2 photolysis would not have been considered at all, v dep,NO 2 values would have potentially been overestimated by 20 up to more than 100 % (according to Eq. 18.1).However, applying an empty ("reference") chamber (see Sect. 5.2), the impact on NO 2 photolysis on the reported v dep,NO 2 values might be smaller if the underlying assumption would be correct, that the effect of NO 2 photolysis is identical in the plant and in the empty chamber.The results of field measurements by Sparks et al. (2001) and Hereid and Monson (2001) most likely have not been affected by NO 2 photolysis because they used a leaf chamber system with red light-emitting diodes which produce no appreciable radiation in the wavelength range of j (NO 2 ).
The corresponding relations for the compensation point concentrations m comp,i are obtained by combining Eqs.(8.1)-(8.3)with Eqs.(11): Here, the value of the fraction (right hand side of Eqs.19.1-19.3)determines whether the actual compensation point concentrations m comp,i are higher, equal, or lower than m * comp,i .For our experimental conditions, m comp,NO 2 would be overestimated by 10 %, if the gas-phase reactions would not have been considered (i.e.assuming m comp,NO 2 = m * comp,NO 2 ).For the compensation point concentration of O 3 the overestimation would be only 1 %.The NO 2 compensation point concentration values reported in previous studies (Thoene et al., 1991(Thoene et al., , 1996;;Rondón et al., 1993;Geßler et al., 2002) would be overestimated between 3 and 17 %, if the photolysis of NO 2 was not considered.
When the value of the fractions on the right hand side of Eqs.(19.1)-(19.3)are examined for being greater, equal, or lower than unity, the following relations are obtained: 3) The relevance of these relations consists in their potential for simply checking, whether or not the correct evaluation of compensation point concentrations has to consider photochemical reactions.Having evaluated measured concentrations m a,i and m s,i by bi-variate weighted linear regression (which delivers n i and b i ), the quantities m * comp,i are determined.Using the simultaneously measured averages of k, j (NO 2 ), m s,NO 2 , m s,NO , and m s,O 3 , the right hand fractions of relations Eqs.(20.1)-(20.3)can be calculated, which provide the necessary quantities to test whether or not m * comp,i have to be corrected for photo-chemical reactions in the dynamic plant chamber (by . It should be noted that interaction of VOCs with the NO-NO 2 -O 3 triad inside the plant enclosure affecting exchange fluxes cannot be excluded completely.However, though reactions of VOCs with O 3 would be fast enough, we may consider them to be negligible, because Picea abies is a species emitting low amounts of VOCs (Kesselmeier and Staudt, 1999).Furthermore, the residence time of air in the dynamic chamber of 75 seconds can be regarded as short enough to exclude direct reactions of NO 2 and VOCs.Reactions of VOCs with NO are only relevant if peroxides derived from the reactions of VOCs and OH are generated inside the chamber.As we may exclude a flux of OH into the chamber due to losses within the tubes, we expected maximal OH concentrations deriving from reactions of ozone with monoterpene species to range around 10 5 molecules cm −3 .As typical monoterpene species we took into account α-pinene and β-pinene, which are the main monoterpenes emitted by spruces.This would result in a maximal concentration of peroxides around 50 × 10 5 molecules cm −3 .For a potential impact of the reaction of NO with these peroxides on the determination of exchange flux densities we assumed a mean NO concentration of 0.16 ppb, as measured in course of our field experiment.Under these conditions the contribution by the reaction of NO with peroxides to the NO 2 exchange flux density was found to be smaller than 1 %.Therefore, from our point of view, these chemical reactions can be neglected.

Bi-variate weighted linear regression
The determination of deposition velocities v dep,i , as well as compensation point concentrations m comp,i is based on linear regression of the measured concentration of trace gas i in ambient air and within the dynamic plant chamber.Therefore, it was necessary to consider errors of both variables in the determination of v dep,i and m comp,i .For our laboratory results (see Sect. 4.4.1)we have shown the effect of applying simple linear regression (no errors considered at all), linear regression (y-errors considered), and bivariate weighted linear regression (y-and x-errors considered) on the significance of derived v dep,NO 2 and m comp,NO 2 data (see Table 7).Generally speaking, applying a simple linear least-square fitting algorithm, the probability of m comp,i = 0 can be highly significant, while applying the bi-variate weighted linear leastsquare fitting algorithm the probability for the existence of m comp,i could easily become "likely" or even "unlikely".In a few cases previous authors have applied the bivariate algorithm (e.g.Geßler et al., 2000Geßler et al., , 2002)).Finally, it should be stated that in all previous studies values of v dep,NO 2 and m comp,NO 2 have been derived from linear relationships between F ex,NO 2 and m s,NO 2 which is statistically problematic, since the dependent variable F ex,NO 2 contains the independent variable m s,NO 2 (see Sect. 2.1).

Conclusions
In this paper we presented a dynamic chamber system for surface exchange flux measurements of reactive and nonreactive trace gases on plants under field and laboratory conditions.We conclude our findings as follows: 1.One of the most important characteristics of our dynamic chamber system is the minimal disturbance of plant physiology and growth.The check of the plant status after long-term field experiment resulted in no detectable differences of photosynthetic capacity between enclosed and not enclosed plant material.
2. According to our "blank" measurements, the wall material of our plant chamber can be considered as chemically inert.We emphasize, that mass fluxes to the walls of the chamber can basically not be neglected and must be considered in the mass flux balance of the dynamic plant chamber, if there are any appreciable effects of ador desorption.
3. The performance of the dynamic chamber system must be controlled and, if necessary, suitable parameterized correction algorithms applied to maintain/improve the precision of NO 2 concentration and exchange flux density measurements.The sensitivity of the NO/NO 2 analyzer to changes in ambient temperature is one of these key parameters.The drift in our analyzer was 0.07 ppb K −1 (NO) and 0.08 ppb K −1 (NO 2 ).The precision of the NO 2 exchange flux densities is almost entirely determined by the precision of the NO 2 concentration measurements, which in turn depends on the sensitivity (limit of detection) of the NO 2 analyzer.At best a flux density precision of ≤10 % may be reached, as long as NO 2 concentrations in the plant chamber differ by more than 0.1 ppb from the expected NO 2 compensation point concentration.the outlet of the dynamic chamber) is an important quality criterion for the determination of high quality exchange flux densities and deposition velocities, and also has a considerable impact on the resulting compensation point concentrations.Especially under field measurements, the percentage of non-significant m i can be rather high due to the temporal variation of ambient concentrations during the measurement interval.
6. Photo-chemical reactions in the dynamic plant chamber's volume must be considered (or be excluded by corresponding set-ups).Otherwise, particularly the exchange of the NO-NO 2 -O 3 triad with the plants could be seriously over-or underestimated.This is particularly important for the determination of the NO 2 deposition velocity.Under our experimental conditions in the field, the overestimation of the NO 2 deposition velocity had reached about 80 % if photolysis of NO 2 has been neglected.Excluding the chemical reaction of NO with O 3 by corresponding experimental design (e.g. using NO and O 3 free purging air), effects of NO 2 photolysis would still be present, as long as there is appreciable illumination of the plants.This is unavoidable because for plant physiological studies the presence of photosynthetically active radiation is essential.The only way out would be to use a chamber wall material where the transmissivity for PAR is high, and in the wavelength range of j (NO 2 ) negligible.For laboratory studies, the application of light-emitting diodes which do not emit in the wavelength range of j (NO 2 ) is promising.
7. While the application of an empty ("reference") chamber for the exchange of non-reactive trace gases may be not problematic, it becomes difficult for reactive trace gases.Considering photo-chemical reactions (which might compensate each other) implies that NO 2photolysis and NO 2 , NO, and O 3 concentrations of the empty and the plant chambers are identical; however, this is not the case, neither under laboratory nor under field conditions.
8. In a mathematical stricter sense, deposition velocities and compensation point concentrations should be derived from linear relationships between the originally measured quantities, namely the NO, NO 2 , and O 3 concentrations at the inlet and the outlet of the dynamic chamber.A straight-forward and thorough statistical treatment of measured data will result in high-quality and reliable data of exchange flux densities, deposition velocities, and compensation point concentrations, if solid characterization and quantification of trace gas concentration errors as well as errors of all other quantities (necessary for calculation of the exchange flux densities) is achieved and general Gaussian error propagation as well as bi-variate weighted linear least-squares fitting regression analysis is applied.9.It is recommended, that results from previous studies on NO 2 exchange flux densities, NO 2 deposition velocities, and NO 2 compensation point concentrations which have been obtained by dynamic plant chambers should be handled with care owing to neglecting (at least) the effects of NO 2 photolysis in the plant chamber's volume and insufficient characterization of the specifity and precision of the NO 2 analyzers.A re-evaluation would be helpful.
and m s,i of the NO-NO 2 -O 3 triad have been measured in cycles of 4 minutes.During this time period, it has been shown, that the error of the purging rate Q is negligible.The volume V of the chambers is a-priori known, its error is considered to be zero.Standard errors of m a,i and m s,i are known for each data pair of measurements.Averages and standard errors of A leaf , j (NO 2 ), k and conjugated concentrations m s,j (j = i) have to be calculated individually from each data set which is used for the determination of F ex,i , v dep,i , and m comp,i .Therefore, according to Eq. (1.1), the mass exchange flux density F ex,NO 2 is a function of 7 error-prone variables, namely x 1 = m a,NO 2 , x 2 = m s,NO 2 , x 3 = j (NO 2 ), x 4 = k, x 5 = m s,NO , x 6 = m s,O 3 , and x 7 = A leaf .Analogously to F ex,NO 2 , the 7 variables for F ex,NO (F ex,O 3 ) in Eq. (1.2) (Eq.1.3) are x 1 = m a,NO (m a,O 3 ), x 2 = m s,NO (m s,O 3 ), x 3 = j (NO 2 ), x 4 = k, x 5 = m s,NO 2 , x 6 = m s,O 3 (m s,NO ), and x 7 = A leaf .Considering Eq. (7.1), the deposition velocity v dep,NO 2 is a function of 3 error-prone variables, x 1 = b NO 2 , x 2 = j (NO 2 ), and x 3 = A leaf , while the deposition velocity  agation (Eq. 14), one have to calculate all the derivatives of y i = F ex,i , y i = v dep,i , and y i = m comp,i , (i = NO 2 , NO, O 3 ) with respect to the corresponding variables x 1 , x 2 , ..., x n mentioned above.

List of symbols and abbreviations
A leaf leaf area m 2 b i slope of regression nmol m −3 analysis of gas i F ex,i exchange flux density nmol m −2 s −1 of gas i F * ex,i exchange flux density nmol m −2 s −1 of gas i, "nonreactive" case j (NO 2 ) photolysis rate of NO 2 s −1 (λ ≤ 420 nm) k rate constant for cm 3 molecule 7.3) and the desired compensation point concentrations (m comp,i ) of the NO-NO 2 -O 3 triad result from combining Eqs.(5.1)-(5.3)and Eqs.(7.1)-(7.3):

Fig. 1 .
Fig. 1.Schematic representation of the determination of bi-directional NO 2 exchange flux density (F ex,NO 2 ), NO 2 deposition velocity (v dep,NO 2 ), and NO 2 compensation point concentration (m comp,NO 2 ) from measurements of NO 2 concentrations at the plant chamber's inlet (m a,NO 2 ) and outlet (m s,NO 2 ) under laboratory conditions (m a,NO = m a,O 3 = j (NO 2 ) ≈ 0).(a): by linear regression of m s,NO 2 with m a,NO 2 .(b): by plotting F ex,NO 2 vs. m s,NO 2 .Dashed lines represent the limits of detection (3 σ -definition) for NO 2 concentration measurements -(a) and (b) panel -and the determination of the NO 2 exchange flux density (b), which are both defined by the sensitivity of the applied NO 2 analyzer (note: LOD (m a,NO 2 ) = LOD (m s,NO 2 )).Data points and error bars of NO 2 concentrations have been simulated to match R 2 (m a,NO 2 , m s,NO 2 ) = 0.9925, error bars of NO 2 exchange flux have been calculated by Gaussian error propagation (cf.Eq. 2).Filled circles identify data points > LODs, hollow circles those ≤LODs.

Figure 6 .
Figure 6.Precision (sm,NO2/mNO2) of the applied NO/NO2 analyzer during laboratory (red curve) and field experiments (green curve).For comparison, curves for precisions of hypothetical analyzers with 0.01  LOD(mNO2)  2 ppb are also shown (numbers on black and grey curves).The blue curve is the precision of the blended NO2 concentration used for fumigation of the young spruce trees in the laboratory.

Fig. 6 .
Fig.6.Precision (s m,NO 2 /m NO 2 ) of the applied NO/NO 2 analyzer during laboratory (red curve) and field experiments (green curve).For comparison, curves for precisions of hypothetical analyzers with 0.01 ≤ LOD (m NO 2 ) ≤ 2 ppb are also shown (numbers on black and grey curves).The blue curve is the precision of the blended NO 2 concentration used for fumigation of the young spruce trees in the laboratory.

Figure 7 . 6 Fig. 7 .
Figure 7. Response test for step changes between two different NO2 concentrations (mNO2).3 The red dashed line marks the switching point.Note: negative values resulted in delayed 4 internal compensation process for temperature and pressure of the analyzer after switching.5 6 Fig. 7. Response test for step changes between two different NO 2 concentrations (m NO 2 ).The red dashed line marks the switching point.Note: negative values resulted in delayed internal compensation process for temperature and pressure of the analyzer after switching.

Fig
Fig. 8. Temporal course of blended NO 2 concentrations(12.3,24.6,  41.0, 73.8, and 139.4 nmol m −3 (0.3, 0.6, 1.0, 1.8, 3.4 ppb)) used for fumigation of young spruce trees during the laboratory experiments.NO 2 concentrations were provided by diluting a NO 2 standard into purified air.Red dashed lines indicate times where blending was changed to obtain the next NO 2 concentration.

Figure 10 .
Results of the response time test with helium.

Fig. 11 .
Fig. 11.Laboratory NO 2 fumigation of 3-4 yr old Norway Spruce trees (Picea abies L.) under controlled conditions (25 • C, 60 %, 450 µmol photons m −2 s −1 ): NO 2 exchange flux density (F ex,NO 2 ) vs. NO 2 concentration measured at the outlet of the dynamic plant chamber (m s,NO 2 ) for application of 2σ -LOD (m s,NO 2 )-definition (a) and 1σ -LOD (m s,NO 2 )-definition (b).F ex,NO 2 data were calculated according Eq. (2), their standard errors according to Eq. (14).Blue circles identify F ex,NO 2 where m s,NO 2 > LOD (m s,NO 2 ), white circles stand for F ex,NO 2 where m s,NO 2 ≤ LOD (m s,NO 2 ), and reddish diamonds for those F ex,NO 2 data, which have to be rejected for non-significance of m NO 2 = (m a,NO 2 − m s,NO 2 ).Blue line (considering blue circle data) and pink line (considering blue circle and reddish diamond data) are calculated according to Eq. (9).NO 2 compensation point concentration m comp,NO 2 is calculated according to Eq. (11) and is represented by red filled circles (considering blue circle data) and pink hollow circles (considering blue circle and reddish diamond data).More details of statistical evaluation are listed in Table6.

Figure 12 .Fig. 12 .
Fig. 12. Switching scheme and time series of trace gas mixing ratios over two full measurement cycles during EGER field experiment.Data were corrected for calibration factors, temperature dependency and offset of analyzers.(a) Control scheme indicating periods of skipped data (first 90 s) for data processing (grey bars), sampling/analysis of ambient air (yellow bars), sampling/analysis of plant chamber 1 (green bars), sampling/analysis of reference chamber (red bars) and sampling/analysis of plant chamber 2 (blue bars).(b-c) Time series of CO 2 and H 2 O mixing ratios measured as difference between reference chamber and respectively switched intake.(d-f) Time series of O 3 , NO 2 and NO mixing ratios.(g) Photosynthetic active radiation (PAR).

Figure 13 .
Figure 13.Field measurements: (a) NO2 concentration measured at the outlet of the dynamic plant chamber (ms,NO2) vs. NO2 concentration 15 measured at the inlet of the dynamic plant chamber (ma,NO2).Light blue circles identify data pairs for significance of mNO2 = (ma,NO2  ms,NO2) 16

Fig. 13 .
Fig.13.Field measurements: (a) NO 2 concentration measured at the outlet of the dynamic plant chamber (m s,NO 2 ) vs. NO 2 concentration measured at the inlet of the dynamic plant chamber (m a,NO 2 ).Light blue circles identify data pairs for significance of m NO 2 = (m a,NO 2 − m s,NO 2 ) and reddish diamonds for those data pairs, which have to be rejected for non-significance of m NO 2 = (m a,NO 2 − m s,NO 2 ).Blue line (considering blue circle data) is calculated according to bi-variate weighted linear least-squares fitting regression analysis (seeSect.3.4.6).(b) NO 2 exchange flux density (F ex,NO 2 ) vs. NO 2 concentration measured at the outlet of the dynamic plant chamber (m s,NO 2 ).F ex,NO 2 data were calculated according Eq. (1.1), their standard errors according to Eq.(14).Reddish diamonds stand for those F ex,NO 2 data, which have to be rejected for non-significance of m NO 2 = (m a,NO 2 − m s,NO 2 ).Blue line (considering blue circle data) and pink line (considering blue circle and reddish diamond data) are calculated according to Eq. (6.1).

Fig. 14 .
Fig. 14.Field measurements: (a) O 3 concentration measured at the outlet of the dynamic plant chamber (m s,O 3 ) vs. O 3 concentration measured at the inlet of the dynamic plant chamber (m a,O 3 ).Orange circles identify data pairs for significance of m O 3 = (m a,O 3 − m s,O 3 ).Orange line is calculated according to bi-variate weighted linear least-squares fitting regression analysis (seeSect.3.4.6).(b) O 3 exchange flux density (F ex,O 3 ) vs. O 3 concentration measured at the outlet of the dynamic plant chamber (m s,O 3 ).F ex,O 3 data were calculated according Eq. (1.3), their standard errors according to Eq. (14).Dark red line is calculated according to Eq. (6.3).

Figure 15 .
Figure15.Field measurements: NO concentration measured at the outlet of the dynamic plant chamber (ms,NO) vs. NO concentration measured at the inlet of the dynamic plant chamber (ma,NO).Light green circles identify data pairs for significance of mNO= = (ma,NO  ms,NO), reddish diamonds stand for those data pairs, which have to be rejected for non-significance of mO3 and grey diamonds for data pairs where mNO ≤ LOD(mNO).Green line (considering green circle data) is calculated according to bi-variate weighted linear leastsquares fitting regression analysis (seeSect.3.4.6).
glass; BG = borosilicate glass; FEP, PFA, PTFE = Teflon materials; PMMA = polymethylmethacrylate (Plexiglas); L = dynamic leaf chamber of gas exchange system Model LI-6400, LiCor, Lincoln, Nebraska, USA 2 w air humidified; pure air = air from a pure air generator; zero air = reactive trace gases removed with filters (converter; PLC = photolytic converter; FeSO been specified through their detection limit only (see Table

Fig. 16 .
Fig. 16.Relative contribution of gas-phase flux densities Fgas,i to the exchange flux densities 2

4.
Determination of NO 2 concentrations at sub-ppb level and of NO 2 exchange flux densities at the thousandths (hundredths) of nmol m −2 s −1 level definitely require (a) a NO 2 specific converter (photolytic converter) and (b) a highly sensitive NO/NO 2 analyzer (lower detection limit (3σ ) of at least 13 nmol m −3 (0.3 ppb), preferably 4.5 nmol m −3 (0.1 ppb)).5.The significance of concentration differences m i (between trace gas concentrations measured at the inlet and www.atmos-meas-tech.net/5 v dep,NO (v dep,O 3 ) depends on 4 error-prone variables, namely x 1 = b NO (b O 3 ), x 2 = k, x 3 = m s,O 3 (m s,NO ), and x 4 = A leaf .The compensation point concentrations m comp,NO 2 (m comp,NO , m comp,O 3 ) are each functions of 6 error-prone variables (see Eqs. 8.1-8.3).These are x 1 = n NO 2 (n NO , n O 3 ), x 2 = b NO 2 (b NO , b O 3 ), x 3 = j (NO 2 ), x 4 = k, x 5 = m s,NO (m s,NO 2 , m s,NO 2 ), and x 6 = m s,O 3 (m s,O 3 , m s,NO).Bi-variate weighted linear least-squares fitting regression analysis of measured m s,i versus m a,i (which considers both, s ma,i and s ms,i ) delivers the quantities n NO 2 , n NO , n O 3 and b NO 2 , b NO , b O 3 as well as their standard errors s nNO 2 , s nNO , s nO 3 , and s bNO 2 , s bNO , s bO 3 .To calculate the standard errors s Fex,NO 2 , s Fex,NO , s Fex,O 3 , s v,dep NO 2 , s v,dep NO , s v,dep O 3 , s m,comp NO 2 , s m,comp NO , and s m,comp O 3 by application of the general Gaussian error prop-

989, 2012 960 C. Breuninger et al.: The dynamic chamber method to study trace gas exchange vs
. m s,NO 2 (which is statistically problematic, see above).For high NO 2 compensation point concentrations (as in Fig.1), m comp,NO 2 can still be evaluated by interpolation from significant data pairs (i.e.data pairs, where >LOD (m NO 2 ), ≥+LOD (F ex,NO 2 ), or ≤−LOD (F ex,NO 2 ), respectively).If m comp,NO 2 falls below LOD (m s,NO 2 ) and F 0 is consequently below +LOD (F ex,NO 2 ), m comp,NO 2 may only be determined by extrapolation from significant data pairs.According to Eqs. ( ex,NO 2 ) has been calculated from corresponding s m s,NO 2 and s m a,NO 2 by Gaussian error propagation.Here, m comp,NO 2 (F ex,NO 2 = 0) is the intersect of the m s,NO 2 -axis with the best-fit line of F ex,NO 2 www.atmos-meas-tech.net/5/955/2012/Atmos.Meas.Tech., 5, 955-

www.atmos-meas-tech.net/5/955/2012/ Atmos. Meas. Tech., 5, 955-989, 2012Table 1 .
Review of the specifications of NO/NO 2 analyzers under well defined (laboratory) conditions.Results are from data simulations (random number application), for details of simulation conditions see text (Sect.2.2).Ranges of minimum detectable NO 2 compensation point concentrations (m comp,NO 2 ), correspond to ranges of NO 2 deposition velocity and the goodness (R 2 ) of relation between the ambient vs. sample NO 2 concentration measurements.Sample NO 2 concentrations (m s,NO 2 ), where the precision of NO 2 concentration measurements (= s m,s NO 2 /m s,NO 2 ) exeeds the 10 % level, are also given.

Table 2 .
Interferences of chemiluminescence NO-NO 2 -NO x analyzers using different NO 2 converters.

Table 3 .
Measured parameters and instrument specifications.Limit of detection (LOD (m i ), 3 σ -definition) for the gas concentrations were determined under field and laboratory conditions.

Table 4 .
Manufacturer details for parts of the dynamic chamber system.
• C, heat output 15 W m −1 the air.Different NO 2 concentrations (between 0.3 and 4 ppb) were generated by mixing NO 2 from a pressurized standard cylinder (m std,NO 2 = 41 151 ± 2049 nmol m −3 (1.004 ± 0.050 ppm) NO 2 in N 2 ; Air Liquide, Germany) into the purified air stream.Mixing was performed by adjustment of two mass flow controllers (MKS Instruments, USA), one to keep the flow of NO 2 standard gas (Q std,NO 2 ), the other the flow of the purified air stream (Q dil ) constant.The blended NO 2 concentration (m blend,NO 2 ) and its standard error (s m blend,NO 2 ) are given by

7 Standard errors of exchange flux densities, deposition velocities, and compensation point concentrations
York et al. (2004)rithm delivers corresponding values of intersect (n i ) and slope (b statistical quantities, like the standard errors of n i and b i (s n,i , s b,i ), as well as correlation and regression coefficients, r(m s,i , m a,i ) and R 2 (m s,i , m a,i ).York et al. (2004)presented the original set of equations for bi-variate weighted linear least-squares fitting regression analysis, where the slope b i has to be solved iteratively (see Appendix B).Standard errors of exchange flux densities F ex,i , deposition velocities v dep,i , and compensation point concentrations m comp,i of the NO-NO 2 -O 3 triad may be derived by applying standard Gaussian error propagation.The standard errors of all variables on the right hand side of Eqs.

Table 5 .
Results of the temperature dependence tests of the analyzers used in this work.Stated temperatures are internal temperatures of the analyzers.The drift specifies the signal change over the whole temperature range.The signal noise is the maximum noise (3σ ) detected with zero air during the test.

Table 7 .
Parameters of NO 2 laboratory measurements of simple (no errors considered), simple (standard error of m s,NO 2 considered) and bi-variate weighted (standard error of m s,NO 2 and m a,NO 2 considered) linear least-squares fitting regression analysis.Data were separated for all data of m NO 2 = (m a,NO 2 − m s,NO 2 ) and for only significant data of m NO 2 .Limit of detection (LOD) of 2 σ , 1 σ and no LOD was applied to the data.

Table 9 .
Parameters of bi-variate weighted linear least-squares fitting regression analysis for field measurements.NO 2 data were separated for all data of m NO 2 = (m a,NO 2 − m s,NO 2 ) and for only significant data of m NO 2 .Data of O 3 were always significant for m O 3 = (m a,O 3 − m s,O 3 ).3 σ detection limit was applied to the data.NO 2 − m s,NO 2 ) (m a,NO 2 − m s,NO 2 ) (m a,O 3 − m s,O 3 )

Table 10 .
Overview of studies which have performed dynamic chamber NO 2 flux measurements on different plant species.