Large amounts of biogenic and anthropogenic trace gases are emitted annually into the atmosphere. Recent estimates (Guenther et al., 2012) suggest that about 1000 Tg of biogenic volatile organic compounds (VOCs), especially isoprene (contributing 50 %) and monoterpenes (15 %), are emitted annually by vegetation. The global burden of anthropogenic emission is dominated by CO2, CO, N2O, CH4, SO2, NO2 and organic carbon, the latter contributing about 11 Tg (Huang et al., 2015). In particular, nitrogen oxides from combustion and microbial activity in soils have a major impact on the chemistry of the natural atmosphere (Crutzen, 1973). Most VOCs are oxidized efficiently in the Earth's boundary layer, the oxidizing capacity of which represents 15 % of that of the entire atmosphere (Lelieveld et al., 2016). Biogenic and anthropogenic VOCs have a significant impact on air quality and human health and knowing and understanding their lifetimes, which are determined by the oxidizing capacity of the atmosphere, is a prerequisite to predicting future atmospheric composition and related climate phenomena (Lelieveld et al., 2008).

During daytime, photochemically formed OH radicals represent the dominant contribution to the oxidative capacity of the atmosphere. As OH levels are vastly reduced in the absence of sunlight, the NO3 radical (formed by reaction of NO2 with O3; Reaction R1) is the major oxidizing agent at elevated NOx for many biogenic terpenoids and other unsaturated compounds at nighttime (Wayne et al., 1991; Atkinson, 2000; Atkinson and Arey, 2003a, b; Brown and Stutz, 2012; Ng et al., 2017). NO2+O3→NO3+O2 NO3 reacts rapidly with NO (Reaction R2, rate constant 2.6×10-11 cm3 molecule-1 s-2 at 298 K; Atkinson et al., 2004) and undergoes rapid photolysis (Reactions R5, R6) so that its lifetime is usually of the order of seconds during the day and its concentration too low for it to be considered an important daytime oxidant.

At night, NO3 can react with NO2, forming N2O5, which thermally decomposes to set up a thermal equilibrium between NO2, NO3 and N2O5 (Reactions R3, R4) with N2O5 formation favored by lower temperatures. As both NO3 and N2O5 are formed from NOx (NOx= NO + NO2) the loss of either NO3 via gas-phase reactions or N2O5 via heterogeneous uptake to particles or deposition implies a reduction in NOx, and thus a reduction in the rate of photochemical O3 formation (Dentener and Crutzen, 1993). In addition, heterogeneous loss of N2O5 can also result in release of ClNO2 from chloride-containing particles (Reaction R7) (Osthoff et al., 2008; Thornton et al., 2010; Mielke et al., 2011; Phillips et al., 2012; Riedel et al., 2012). The main loss processes of NO3 are summarized in Fig. 1. NO3+NO→2NO2NO2+NO3+M→N2O5+MN2O5+M→NO3+NO2+MNO3+hν→NO2+ONO3+hν→NO+O2N2O5+surface→NO3-(and/or ClNO2) In rural and forested areas, reaction with biogenic VOCs can dominate the loss of NO3 (Mogensen et al., 2015). Especially terpenoids like limonene (k=1.2×10-11 cm3 molecule-1 s-1), α-pinene (k=6.2×10-12 cm3 molecule-1 s-1) and isoprene (k=6.5×10-13 cm3 molecule-1 s-1) have high rate constants for reaction with NO3 (IUPAC, 2016; Ng et al., 2017). In such environments, when NOx levels are low, NO3 mixing ratios may be sub-pptv and below the detection limit for most instruments (Rinne et al., 2012).

The reaction of NO3 with trace gases containing unsaturated C=C bonds proceeds via addition to form nitrooxyalkyl radicals that undergo rapid reaction with O2 to form nitrooxyalkyl peroxy radicals. The peroxy radicals react further (with HO2, NO, NO2 or NO3,) to form multifunctional organic nitrates, which can contribute to generation and growth of secondary organic aerosols (Fry et al., 2014; Ng et al., 2017) or be lost by deposition.

The role of NO3 as an oxidizing agent may be assessed via its total reactivity (or inverse lifetime). Whereas for OH, experimental methods for measuring total reactivity in ambient air exist (Kovacs and Brune, 2001; Sinha et al., 2008), NO3 reactivity has not yet been directly measured. Stationary-state approximations have often been used to calculate NO3 lifetimes from its mixing ratio and production rate, the latter being given by k1[NO2][O3] (Heintz et al., 1996; Geyer and Platt, 2002; Brown et al., 2007a, b, 2009; Sobanski et al., 2016b). Thus the stationary-state turnover lifetime, τss, can be calculated according to Eq. (1), where k1 is the rate constant for Reaction (R1). kss=1τss=k1O3NO2NO3 This method is applicable when the chemical lifetime of NO3 is sufficiently short so that stationary state can be achieved within transport time from emission to measurement location (Brown et al., 2003). Formally it is achieved when the production and loss of NO3 and N2O5 are balanced (Brown et al., 2003; Crowley et al., 2011). The time to acquire stationary state depends on production and loss rates for NO3 and N2O5 and can take several hours. This approach can break down under conditions of moderate to high NO2 levels, strong sinks, low temperatures or very clean air masses in which the sinks for NO3 and N2O5 become small (Brown et al., 2003). Indeed, Sobanski et al. (2016b) observed much lower stationary-state loss rates compared to those calculated from measured VOC mixing ratios during the PARADE 2011 campaign and concluded that this was mainly the result of sampling from a low-lying residual layer with VOC emissions that were too close for NO3 concentrations to achieve stationary state. They also considered the possibility that NO3 may be formed by the oxidation of NO2 by Criegee intermediates, which would bias calculations of its reactivity.

Summarizing, NO3 reactivity with respect to gas-phase losses is an indication of nighttime oxidation rates of VOCs, with direct impacts on NOx levels by forming long-lived reservoir species (alkyl nitrates), some of which will partition to the particle phase. Via modification of N2O5 concentrations (formed in an association reaction of NO3 with NO2, Reaction R3), the NO3 reactivity indirectly controls heterogeneous NOx losses and ClNO2 formation rates.

In this paper we describe a newly developed instrument that enables point measurements of NO3 reactivity in ambient air. After introducing the methodology in Sect. 2, we show the results of extensive laboratory characterization of the instrument along with discussion of the uncertainties associated with those measurements in Sects. 3 to 5. In Sect. 6 we present a data set of ambient NO3 reactivity obtained at a forested/urban location in southwestern Germany.

Our experiments to measure NO3 reactivity involve comparison of loss rates of synthetically generated NO3 in zero air and in ambient air introduced into a flow-tube reactor. In zero air, the loss of NO3 is due to its reaction with NO2 (present as a necessary component in the generation of NO3, see below) and losses on surfaces of the flow tube. When zero air is replaced by ambient air, NO3 is additionally removed by reaction with reactive gases present and its mixing ratio reduced accordingly. An analysis of the change in signal for a fixed reaction time enables the NO3 reactivity to be derived once certain corrections have been applied (see below).

Figure 2 displays a schematic diagram of the experimental setup. The three central components are a dark reactor for generation of NO3, the flow tube in which NO3 reacts with trace gases in ambient air samples and the detection system for NO3.

We first consider the passage of NO3 through the flow tube in a flow of zero air. If NO3 is lost in one or more pseudo-first-order processes, its decay should be exponential and its concentration, [NO3]tZA after a reaction time t, is given by Eq. (8). [NO3]tZA=[NO3]0ZAexp⁡(-kZAt), where ZA refers to use of zero air. As NO3 is lost only via reaction with NO2 and to the wall, kZA=kwall+kNO2, where kw is the first-order loss rate constant for wall loss and kNO2 is the first-order loss rate constant for reaction with NO2 and is equal to k3[NO2]. When zero air is switched for ambient air containing reactive trace gases, we have [NO3]tAmb=[NO3]0Ambexp⁡(-kAmbt), where kAmb=kw+kNO2+kRTG and kRTG is the first-order loss rate constant for reaction of NO3 with trace gases present in ambient air other than NO2.

If [NO3]0ZA and [NO3]0Amb are equivalent, Eq. 10 is obtained. [NO3]tZAexp⁡-kZAt=[NO3]tAmbexp⁡-kAmbt Rearranging and substituting for kZA and kAmb leads to kRTG=ln⁡[NO3]tZA[NO3]tAmbt=1τ, where τ is the NO3 lifetime. In principal, it should thus be possible to calculate the reactivity of NO3 in ambient air by measuring [NO3]tZA and [NO3]tAmb and knowing the reaction time t. Later we discuss the applicability of this expression and show that corrections are necessary to take the reformation of NO3 into account, especially when dealing with air masses with high NO2 content. This is similar to the laboratory experiments described above and required numerical simulation, which we present below.

The concentration of NO3 in zero air measured when the injector is positioned for maximum reaction time, [NO3]tZA, was measured by flushing the inlet with 3000 sccm zero air creating an overflow of ≈ 100 sccm. When switching to ambient measurements, the zero-air overflow was redirected via a flow controller (F3 in Fig. 2) that connected the zero-air overflow line to the exhaust and which was set to 3500 sccm. This setup has the advantage of enabling dynamic dilution of ambient air. If the reactivity is so high that the NO3 levels approached the detection limit, F3 does not withdraw the entire 3500 sccm overflow but allows, e.g., 2000 sccm to be added to the inlet, resulting in sampling 900 sccm of ambient air plus 2000 sccm of zero air, a dilution factor of 2900/900 which is slightly increased by the 400 sccm flow from the darkened reactor. A five-point dynamic dilution with zero air is implemented in the software, which changes the set point for F3 and dilutes the ambient air with zero air if [NO3]tAmb decreases below 10 pptv for a time period of 30 s. Conversely, the dilution can be decreased again if [NO3]tAmb becomes ≥[NO3]tZA-10 ppt. Dilution factors (Di) were determined using a Gilibrator flow meter (Gilian Gilibrator-2) and were D1=1.14 for the measurement of pure ambient air (here the small dilution effect is caused by the 400 sccm zero air used in the production of NO3), D2=1.74, D3=3.71, D4=8.98 and D5=14.07 when diluting ambient air. With increasing dilution, errors in the measurement will increase as well (see Sect. 5).

The analytical expression given above to derive the NO3 reactivity is an ideal case in which NO3 is lost by a number of first-order processes and is not formed in the flow tube to a significant extent. However, as we already demonstrated in the laboratory experiments to examine the effects of varying NO, NO2 and NO3 concentrations, the formation of N2O5 in the reaction of NO3 with NO2 (Reaction R3) and its thermal decomposition back to NO3 can impact the NO3 concentration as NO2 is present both in the mixture used to generate N2O5 and NO3 and in ambient air. While the formation of N2O5 from NO2 and NO3 (Reaction R2) is, to a good approximation, independent of temperature between about 280 and 305 K, the rate constant for thermal decomposition of N2O5 (Reaction R3) varies by a factor of 26 over the same temperature range. The simple, analytical approach outlined above thus fails at temperatures where the decomposition of N2O5 is important and when sufficient NO2 is present to account for a significant fraction of the loss of NO3. This is illustrated in Fig. 7a, in which simulations of the NO3 concentration at a reaction time of 10.5 s and at different temperatures and amounts of NO2 (as reactant) are displayed and compared with the simple exponential behavior (black data points) calculated from Eq. (11). The simulations show that the dependence of the NO3 concentration on NO2 is non-exponential, indicating that regeneration of NO3 from the N2O5 formed is significant, especially at higher temperatures. Figure 7b plots the ratio of the true reactivity (i.e., that used as input into the numerical simulation) versus that obtained by analyzing the simultaneous change in NO3 concentration using Eq. (11). It is evident that the use of this expression generally results in underestimation of the true reactivity due to the formation and decomposition of N2O5. The bias will be largest when sampling polluted air where the reactivity has a large component due to NO2 and small under conditions of low NO2 and high kRTG typical for remote, forested areas. However, as previously mentioned, the decomposition of N2O5 is strongly temperature dependent so that the bias will increase with rising temperature and decrease with sinking flow-tube temperature.

Apart from the formation and thermal dissociation of N2O5, the reaction of NO2 with O3 may, under some conditions, represent a further potential source of NO3 in the flow tube despite the low rate constant for Reaction (R1). Due to the in situ method of production of N2O5 and NO3 in the dark reactor, NO2 (0.6–3 ppbv) and O3 (40–50 ppbv) are always present in the flow tube. NO3 generated in the flow tube was therefore simulated for different amounts of O3 and NO2 corresponding to the minimum and maximum mixing ratios used in our experiments. Figure 8 indicates that with 50 ppbv of O3 and 2 ppbv NO2, < 0.5 pptv of NO3 is formed in the 10.5 s available for reaction in the flow tube, which would not strongly impact the results if the analytical expressions above were used to derive the NO3 reactivity. Under highly polluted conditions (e.g., 100 ppbv O3 and 20 ppbv NO2) the effect is, however, measurable (> 2 pptv).

The discussion above indicates that the use of Eq. (11) can, under certain circumstances (e.g., low NOx, high NO3 reactivity to VOCs), give a reasonable representation of the NO3 reactivity. However, in order to be able to derive NO3 reactivities from any air mass we use numerical simulation take NO3 reformation into account and enable extraction of accurate values in any conditions.

To validate our experimental and analytical procedure, we performed reactivity measurements on a bottled isoprene standard (0.933±0.09 ppmv, Westfalen), diluted in zero air. Isoprene was chosen as it is an important biogenic reactant for NO3 in the troposphere and also because the rate coefficient, kisoprene, for its reaction with NO3 has been studied on many occasions (Atkinson et al., 2006; IUPAC, 2016) and therefore has a low associated uncertainty (kisoprene=6.5±0.15×10-13 cm3 molecule-1 s-1 at 298 K).

Experiments were carried out at various isoprene and NO2 mixing ratios and the results are summarized in Fig. 10, which indicates excellent agreement between the measured reactivity and that calculated from the isoprene mixing ratio and rate coefficient, the slope of an unweighted fit being 1.00±0.03. The error bars on the calculated reactivity represent total uncertainty in the isoprene and NO2 mixing ratio, the reaction time and the rate coefficient. These results confirm that the instrument and data analysis procedure measure accurate values of NO3 reactivity in the presence of NO2 and organic reactants.

While the overall uncertainty associated with absolute NO3 concentration measurement are influenced by factors such as uncertainty in the cross section as well as in the measurement of the laser emission spectrum, the fractional change in concentration used to derive the NO3 reactivity is not impacted. The detection limit for measuring NO3 reactivity is defined by the minimal detectable change (MDCNO3) in the NO3 mixing ratio. This depends on noise levels and drift in ring-down time, i.e., on the precision of the NO3 signal and also on the stability of the synthetically generated NO3. The instrumental noise on the NO3 signal was reduced by averaging over ≈ 3 s per data point (≈ 1800 ring-down events) to give a noise-limited detection limit (1σ) of ∼ 0.2 pptv. Precision is limited by the stability of the CRDS setup where changes in the mirror reflectivity induced by thermal or mechanical stress can lead to a drift in the ring-down time. The precision can be estimated from the standard deviation of the signal from one zeroing period to the next over the measurement period. Under typical laboratory conditions this was normally ≈ 0.7 pptv.

Since [NO3]0ZA is interpolated onto the measured [NO3]tAmb time series to calculate the reactivity, the stability of the NO3 source is of great importance. Changes in the amount of synthetically generated NO3 are caused by fluctuations in the temperature or pressure of the dark reactor, the flow of NO2 and changes in the intensity of light from the O3 generator. In general, the poorer the stability of the NO3 source chemistry, the more frequently the NO3 mixing ratio in zero air has to be measured. In laboratory conditions, changes of ±1 pptv within 1 h were typical, making [NO3]tZA measurements every 1200 s more than sufficient. In field conditions, where the instrument housing may be subject to larger temperature fluctuations, more frequent determination of [NO3]tZA may be necessary. The NO3 source stability was obtained from the standard deviation of the averaged [NO3]tZA concentrations and propagating this with the standard deviation of two consecutive [NO3]tZA measurements, for which typical values in laboratory conditions were ≈ 1 ppt. To define an overall, minimal detectable change in NO3 (MDCNO3), the noise- and drift-limited precision was combined with the NO3 source stability to result in MDCNO3=2.5 pptv.

An MDCNO3 of 2.5 pptv results in a lower limit for the measurement of NO3 reactivity of 0.005 s-1 (obtained from Eq. 11 with [NO3]tZA=50 pptv and [NO3]tAmb=[NO3]tZA-MDCNO3=47.5 pptv, at the lowest dilution factor of 1.14). An upper limit for the measurable reactivity is 45 s-1, largely defined by the uncertainty of the dilution factor. Dilution factors were obtained by measurements of the actual flows going into the flow tube using a Gilibrator flow meter (Gilian Gilibrator-2, stated accuracy ±1 %). The total uncertainty in the dilution factor is defined by the accuracy of the measurement of the dilution flows as well as by the accuracy of the flow controllers used for flow regulation (±2 %) and was calculated to be 2.5 %. The error in the calculated reactivity is lowest for the lowest dilution but if the [NO3]tAmb gets close to the detection limit this will also have a strong influence on the calculated reactivities making a higher dilution factor favorable. Dilution factors were chosen to keep the instrument operating in a region (10 pptv < NO3 < 40 pptv) where both effects are minimized.

A minimum detectable change in NO3 of 2.5 pptv leads to an uncertainty of ≈ 15 %, when NO3 varies between ≈ 10 and 30 pptv (starting from 50 pptv in zero air). The uncertainty increases dramatically when NO3 levels are close to 50 pptv (i.e., very low reactivity) or less than 5 pptv (very high reactivity without dilution). This is illustrated in Fig. S1 of the Supplement. As mentioned in Sect. 2.3 the uncertainty in the reaction time (10 %) also contributes to the overall uncertainty.

To assess the uncertainty associated with derivation of the NO3 reactivity from numerical simulation, uncertainties associated with the input parameters have to be considered. As previously demonstrated (Groß et al., 2014) this is best assessed in a Monte Carlo approach in which the key parameters are varied within a range reflecting their uncertainty limits. The parameters that most sensitively influence the derived value of NO3 reactivity are the NO2 mixing ratio and the rate coefficients for N2O5 formation (k3=1.2±0.1×10-12 cm3 molecule-1 s-1) and decomposition (k4=4.4±0.4×10-2 cm3 molecule-1 s-1). The rate coefficients listed are for 1 bar and room temperature as appropriate for the experimental conditions, the uncertainties quoted (≈ 10 %) are based on assessment of kinetic data (Burkholder et al., 2016). The Monte Carlo simulations were initiated with a NO3 mixing ratio (in zero air) of 50 pptv, decreasing to 20 pptv upon reaction with air. In total, six sets of ≈ 1200 simulations were carried with variation of the initial NO2 mixing ratio between 1 and 5 ppbv and the associated error in NO2 mixing ratio was taken as 8 %. For any given simulation, the output value of the NO3 reactivity (kRTG) was stored. The 2σ uncertainty was derived from the Gaussian fits to histograms of kRTG (insets at NO2=1.0, 3.0 and 5.0 ppbv) and is plotted (as a percent of kRTG) versus kRTG/NO2. The latter may be considered a measure of whether NO3 reacts predominantly with NO2 to form N2O5 (k3) or with reactive trace gases. Figure 11 shows that the uncertainty associated with the simulations is very sensitive to ambient NO2 levels, varying between > 100 % (at 5 ppbv NO2 and a reactivity of 0.017 s-1) and 3.1 % (at 1 ppbv NO2 and a reactivity of 0.092 s-1) of the extracted kRTG. Clearly, the extraction of kRTG is most accurate in conditions of low NOx and when NO3 lifetimes are short (e.g., forested regions far from anthropogenic activity).

Another potential bias in the measurement is the temperature dependence of the rate constant of the reactions of trace gases with NO3. Measurements were normally conducted at 20 ∘C in the flow tube whilst outside temperature can differ from this. However, (unlike OH) the NO3 reactions which dominate its reactivity involve addition to double bonds (e.g., of terpenes) and are only weakly temperature dependent. Therefore, to a good approximation, this error can generally be neglected. To illustrate this, we consider the reaction between NO3 and the usually most abundant monoterpene, α-pinene. The rate constant at the flow-tube temperature (20 ∘C) is 6.4×10-12, increasing to 7.0×10-12 cm3 molecule-1 s-1 at 5 ∘C and decreasing to 5.9×10-12 cm3 molecule-1 s-1 at 35 ∘C, which are changes of < 10 %. Note also that for many monoterpenes, the temperature dependence of the rate constant is not known but expected to be weak (IUPAC, 2016).

Under circumstances where the reactivity is known to be driven by reaction with reactive trace gases for which NO3 has large temperature dependence this error has to be taken into consideration.

We now examine the potential bias caused by use of NO3 concentrations as large as 50 pptv, which may change the reactivity of the air either by removing a significant fraction of gas-phase reactants or via formation of peroxy radicals (RO2), which may also react with NO3. In a first scenario, we assume that the reactivity is caused by a single species, namely the generally dominant terpene, α-pinene and consider both low (kRTG=0.005 s-1) and high reactivity regimes (kRTG=0.1 s-1). A value of kRTG=0.005 s-1 would result if 34 pptv of α-pinene were available for reaction. In a first approximation, assuming first-order kinetics, we calculate that 2.5 pptv of the initially available 50 pptv of NO3 is lost in the 10.5 s reaction time, and consequently a change in α-pinene of 2.5 pptv would also occur. This is only 7 % of the initial concentration, indicating an upper limit to a negative bias of 7 %. This is an upper limit, as the assumption of first-order kinetics is not entirely appropriate. As NO3 reacts with α-pinene in air to form a nitrooxyperoxy radical (RO2) we also consider a positive bias due to reaction of NO3 with this RO2. To do this we assume a rate constant of 1.2×10-12 cm3 molecule-1 s-1 for the reaction as observed for NO3+ CH3O2 (Atkinson et al., 2006) and assume that this rate constant is approximately independent of the nature of the organic fragment (R) as is the case for reactions of RO2 with NO. The 2.5 pptv RO2 thus generated results in an incremental NO3 reactivity of 7×10-5 s-1, a positive bias of 1.5 %. Again, this is an upper limit, as the calculation assumes that this concentration of RO2 is constant and available for the whole 10.5 s of reaction time. For higher reactivity (0.1 s-1) a similar calculation shows that the 670 pptv required would reduce the NO3 concentration to 17.5 pptv, itself being diminished to ∼ 640 pptv, a change of just 5 %. The 32.5 pptv RO2 generated would result in a loss rate constant for NO3 of ∼9×10-4 s-1, a positive bias of ∼ 1%. In conclusion, for reactive systems in which a large concentration of reactive trace gases with moderate reactivity towards NO3 are encountered, we expect no significant bias. The only scenario in which a large bias can ensue is when a low reactivity is caused by a very low concentration of an extremely reactive traces gas. Taking the example of 1 pptv of a highly reactive terpenoid (k=2×10-10 cm3 molecule-1 s-1) it is easy to show that it would be reduced to just a few percent of its initial concentration when mixed with 50 pptv of NO3 for 10.5 s. In this case a large negative bias would result. In the real atmosphere, this situation is however unlikely to occur as such reactive species are usually substantially reduced in concentration compared to the generally dominant biogenics such as α-pinene.

The overall uncertainty thus derives from a combination of measurement errors (cavity instability, drift in NO3 source, etc.) and the need to correct for NO3 reactions with NO2. Under ideal conditions (e.g., as described above for laboratory operation) the former can be reduced to ≈ 16 %. For a scenario in which biogenic VOCs dominate NO3 reactivity in a low NOx (< 1 ppbv) environment an additional uncertainty of ≈ 6–10 % from the numerical simulations results in a total uncertainty of ≈ 17–20 %. In a high-NOx environment, the total uncertainty will be dominated by that associated with the simulations. For example, at 5 ppbv NO2 and a reactivity of 0.03 s-1 the total error would be close to 45–50 %.

The NO3 reactivity setup described above was deployed for the first time in the field during the NOTOMO campaign in the Taunus mountains (southwestern Germany) in 2016. The site, previously described in detail (Crowley et al., 2010; Sobanski et al., 2016b), is situated on top of the “Kleiner Feldberg” mountain (850 m above sea level) in a forested area with urban influence. The site is impacted by biogenic emissions from forested regions (mainly in the northwest) and by anthropogenic emissions from the local urban centers of Frankfurt, Mainz and Wiesbaden in the southeast to southwest.

We present the first instrument for measurement of NO3 reactivity in ambient air. The flow-tube-based instrument utilizes the depletion of synthetically generated NO3 when mixed with ambient air and has a dynamic range of 0.005 to 45 s-1. Following intensive laboratory characterization to determine the effective reaction time, the wall loss constant of NO3 and the effect of NO3 formation and reformation in the flow tube, it was successfully tested against an isoprene standard. The overall uncertainty depends on the relative rate of reaction of NO3 with NO2 or with other traces gases (e.g., VOCs or NO) that do not generate N2O5 and which, under ideal conditions, is close to 15 %. The instrument is thus best suited for measurement of NO3 reactivity in regions with high biogenic activity and relatively low direct anthropogenic emissions of NOx, i.e., regions where the measurement of NO3 concentrations is difficult owing to low production rates and a high loss term.

First deployment of the instrument was during the NOTOMO observational experiment in summer 2015 at a forested, mountain site with urban influence. The measured NO3 reactivity ranged from 0.006 to 0.1 to s-1 at nighttime and reached values as high as 1.4 s-1 during daytime. As expected, daytime reactivity was dominated by reaction with NO while nighttime reactivity involved other (presumably organic) trace gases. A comparison with stationary-state calculations of the NO3 reactivity revealed occasional poor agreement, presumably related to very local emissions causing a breakdown of the stationary-state assumption.

Improvements to the dynamic range of the instrument require further stabilization of the NO3 source and cavity optics to reduce the minimal detectable change in NO3 (presently MDCNO3=2.5 pptv). This could also be achieved by the use of larger volume flow tubes. Reduction in the initial NO3 concentration used would also reduce any potential bias caused by depletion of reactants or secondary chemistry. Future deployment with simultaneous measurements of NO3, NO2, O3 and VOCs will be conducted to compare direct measurements of NO3 reactivity with those obtained from the stationary-state approach and also those calculated from summing losses due to individual VOCs.