Characterisation of an inlet pre-injector laser-induced fluorescence instrument for the measurement of atmospheric hydroxyl radicals

Atmospheric measurements of hydroxyl radicals (OH) are challenging due to a high reactivity and consequently low concentration. The importance of OH as an atmospheric oxidant has motivated a sustained effort leading to the development of a number of highly sensitive analytical techniques. Recent work has indicated that the laser-induced fluorescence of the OH molecules method based on the fluorescence assay by gas expansion technique (LIF-FAGE) for the measurement of atmospheric OH in some environments may be influenced by artificial OH generated within the instrument, and a chemical method to remove this interference was implemented in a LIF-FAGE system by Mao et al. (2012). While it is not clear whether other LIF-FAGE instruments suffer from the same interference, we have applied this method to our LIF-FAGE HORUS (Hydroxyl Radical Measurement Unit based on fluorescence Spectroscopy) system, and developed and deployed an inlet pre-injector (IPI) to determine the chemical zero level in the instrument via scavenging the ambient OH radical. We describe and characterise this technique in addition to its application at field sites in forested locations in Finland, Spain and Germany. Ambient measurements show that OH generated within the HORUS instrument is a non-negligible fraction of the total OH signal, which can comprise 30 to 80 % during daytime and 60 to 100 % during the night. The contribution of the background OH varied greatly between measurement sites and was likely related to the type and concentration of volatile organic compounds (VOCs) present at each particular location. Two inter-comparisons in contrasting environments between the HORUS instrument and two different chemical ionisation mass spectrometers (CIMS) are described to demonstrate the efficacy of IPI and the necessity of the chemical zeroing method for our LIF-FAGE instrument in such environments.


Introduction
The hydroxyl radical, OH, plays a central role in the chemistry of the troposphere, where it acts as the main daytime oxidising agent, initiating the photochemical degradation of many chemical species emitted by natural and anthropogenic sources. In the lower troposphere the primary OH radical formation is dominated by photolysis of ozone and subsequent reaction of the excited oxygen atom with water vapour (Levy, 1971). Minor primary sources are the photolysis of nitrous acid and hydrogen peroxide and ozonolysis of unsaturated carbon compounds. Once formed, the OH radical reacts rapidly with many atmospheric trace gas species converting a large fraction of the volatile organic matter (Levy, 1974). The ambient measurement of OH is therefore a good test of proposed chemical mechanisms postulating the importance of chemical species and/or processes in the atmosphere. The OH radical is usually measured in the field with one of three techniques: differential optical absorption spectroscopy (DOAS; Perner et al., 1987) via absorption of light by OH, chemical ionisation mass spectrometry (CIMS; Eisele and Tanner, 1991) via the detection of H 2 SO 4 after the oxidation of SO 2 by atmospheric OH, and laser-induced fluorescence (LIF-FAGE;Hard et al., 1984) via the detection of OH radical fluorescence after laser excitation. Several comparison campaigns have been performed, both on ground (Hofzumahaus et al., 1998;Schlosser et al., 2007) and aircraft (Eisele et al., 2001;Ren et al., 2012), to test the consistency and performance of the different techniques and have generally shown good agreement. However, a number of recent measurements performed by LIF-FAGE instruments in VOC-rich environments have shown considerably higher values of OH radicals than can be accounted for by well-established chemical mechanisms Lelieveld et al., 2008;Ren et al., 2008;Hofzumahaus et al., 2009;Kubistin et al., 2010;Whalley et al., 2011). These model-measurement disagreements have prompted the discovery of new chemical paths Peeters et al., 2009;da Silva, 2010a, b, c;Crounse et al., 2011) and the development of alternative chemical mechanisms to account for the discrepancies (Lelieveld et al., 2008;Hofzumahaus et al., 2009;Peeters et al., 2009;Peeters and Müller, 2010;Taraborrelli et al., 2012). At least in part, the disagreement between models and measurements could be related to the measurement technique. LIF instruments can suffer from a number of well-characterised interferences, such as OH generated by the laser pulse from species like ozone, acetone, or H 2 O 2 , as well as spectral interferences from, for example, naphthalene and SO 2 . When present, these can be corrected for (Holland and Hessling, 1995;Martinez et al., 2004;Kubistin, 2009). Recent work by Mao et al. (2012) suggested that, at least in some LIF-FAGE designs and possibly depending on the characteristics of the environment, a process not currently accounted for may generate OH within the low-pressure side of the instrument. The authors report the measurement of OH radicals in a Ponderosa pine plantation in the California Sierra Nevada Mountains with the deployment of a chemical zero-level system in parallel with the traditional FAGE method.
In this paper, we describe the characterisation and application of such a chemical zero system (inlet pre-injector, IPI) to the HORUS instrument, following the design of , to address the possible role of internally formed OH in our system. In addition, we also describe how the atmospheric OH concentration is determined with the new modification of the instrument in three field measurements that include two comparisons of the Mainz IPI-LIF-FAGE instrument with CIMS measurements of OH in chemically distinct environments.

Mainz LIF-FAGE description
The LIF-FAGE instrument (HORUS) in use at the Max Planck Institute for Chemistry, Mainz, is based on the design of GTHOS (Ground Tropospheric Hydrogen Oxide Sensor) Figure 1. HORUS instrument setup scheme with IPI. The ambient air is sampled through IPI where an OH scavenger is added periodically and is then sampled by the instrument inlet through a critical orifice. In the first cell, OH is excited by a laser pulse at around 308 nm and the fluorescence is detected by an MCP. Directly in front of the second cell a mixture of NO with nitrogen is injected and HO 2 is detected after conversion into OH. The total pressure inside the instrument is maintained around 350 Pa. described by Faloona et al. (2004) and is described in detail by Martinez et al. (2010). The instrument consists of three parts: the inlet and detection system, the laser system and the vacuum system (Fig. 1). The air is drawn at ∼ 7 L min −1 through a critical orifice (1 mm diameter) and OH is selectively excited by pulsed UV light at around 308 nm on resonance with the Q 1 (2) transition line (A 2 + −X 2 , ν = 0 ν = 0). The laser pulse is directed into a multipass "White Cell" (White, 1942) crossing the detection volume 32 times to increase the sensitivity. The fluorescence signal from the excited hydroxyl radicals is detected at low pressure (∼ 300-500 Pa). As the fluorescence is detected at similar wavelengths as the excitation, a time-gated photon counting technique using micro-channel plate detectors (MCP) is used. The UV light for excitation of the hydroxyl radicals is provided by a Nd : YAG pumped, pulsed, tunable dye laser system (Wennberg et al., 1994;Martinez et al., 2010) operated at a pulse repetition frequency of 3 kHz. The instrument has two consecutive detection cells: in the first cell OH radicals are detected, and in the second cell hydroperoxyl radicals (HO 2 ) are detected via the conversion of HO 2 to OH by the addition of NO. The calibration of the instrument is achieved via production of a known amount of OH and HO 2 from the photolysis of water at 185 nm using a mercury lamp. A more detailed description of the instrument calibration is reported by Martinez et al. (2010). The fluorescence background signal of the instrument is measured by tuning the excitation laser on (online signal, Sig on ) and off (offline signal, Sig off ) resonance with the OH transition line at 308 nm (Fig. 2a). The spectra of the measured atmospheric OH is compared with the one obtained from a reference cell in order to rule out possible fluorescence signal due to species that fluoresce at similar wavelengths (such as naphthalene and SO 2 ). During all three measurement campaigns described in this paper, the HORUS instrument was equipped with a 14 cm inlet resulting in a residence time of the air between the pinhole and the detection cell of 2.5 ms . Laser power and pressure at the cell were, respectively, ∼ 4 mW and ∼ 310 Pa during the campaigns in Finland and Spain and ∼ 9 mW and 380 Pa during the campaign in Germany.

Inlet Pre-Injector (IPI)
Figure 3 shows a schematic cross-section of IPI currently in use as part of the HORUS instrument. The purpose of IPI is the addition of an OH scavenger to remove the atmospheric OH before it is sampled by the inlet to account for an OH signal generated within the instrument. Before the introduction of IPI, the atmospheric OH concentration was obtained by multiplying the OH fluorescence signal (OHF), obtained from the difference between fluorescence online and fluorescence offline signals, by the total instrument sensitivity (S). The total instrument sensitivity depends on many parameters such as laser power, efficiency of the detector, temperature and humidity , and is determined by performing calibrations on a regular basis: With IPI (Fig. 2b), the instrument is also cycled every 2 min between injection of the scavenger (background OH fluorescence signal, OHF bg ) and no injection (total OH fluorescence signal, OHF tot ). The atmospheric OH concentration (OH atm ) is then obtained by multiplying the difference between the total OH fluorescence signal and the background OH fluorescence signal by the instrument sensitivity and by a factor Figure 3. Inlet pre-injector (IPI) scheme. The injection of the scavenger is achieved via eight 0.5 mm holes (Label 1) positioned 5 cm above the pinhole of the inlet (Label 2). The scavenger is carried through IPI with ∼ 4000 sccm of synthetic air. The residence time in IPI after the injection of the scavenger is ∼ 4 ± 0.5 ms to scavenge between 80 and 95 % of the atmospheric OH, depending on the scavenger concentration. Label 3 indicates the position of a metallic grid. Label 4 shows the connection to the blower that samples the air through IPI.
(F ) accounting for scavenging efficiency and radical losses introduced by IPI: To compare the derived atmospheric OH concentration with the respective total and background signals and to describe the error in concentration of atmospheric OH that would have been made without the use of IPI, we apply the same OH calibration factor to both OH fluorescence signals, OHF bg and OHF tot , and we refer to them as the background OH signal (OH bg ) and the total OH signal (OH tot ). The units of these variables are therefore molecules cm −3 OH equivalent and no inference is drawn as to the actual concentration of the interfering species.
As mentioned above, the detection of atmospheric OH using LIF-FAGE with an additional scavenger was employed following the method described by . The setup of the OH scavenger injection and its operation differ between the two LIF-FAGE instruments. The main differences between the two injection systems are the internal shape and the amount of air sampled, leading to differences in radical wall losses and in the scavenging efficiency. The hyperbolic internal shape of IPI (max. cross-section = 35 mm; min. cross-section = 6 mm) (Fig. 3) was chosen based on Eisele et al. (1997) to sample air that has little contact with walls and high velocity where the inner diameter is smaller to provide rapid turbulent mixing of atmospheric air and added scavenger. A blower (SCL 20DH from FPZ, Italy) is directly connected to IPI (Fig. 3, label 4) pulling air with a flow rate between 150 and 280 L min −1 . The flow velocity is monitored using a differential pressure sensor calibrated against a gas metre. An aluminium perforated mesh with square holes is located around the instrument inlet between the nozzle and the connection to the blower sampling the air through IPI (Fig. 3, labels 3 and 4). As the mesh texture is very thin and located above the blower connection, the resistance to the air is strong enough to break the flow pattern that would be strongly pointing towards the direction of the pulling position of the blower, allowing a more homogeneous flow and mixing. The OH scavenger is injected via eight 0.5 mm diameter holes (Fig. 3, label 1) into the centre of the flow of air sampled by IPI, 5 cm above the pinhole of the inlet (Fig. 3, label 2). Assuming plug flow, the estimated residence time of the ambient air in IPI from the injection of the scavenger to the instrument inlet when pulling 150 L min −1 of air is ∼ 4 ms. To improve the mixing between the scavenger and the sampled air, the injection happens at the minimum cross-section of IPI, and to achieve a good penetration of the scavenger into the sample flow, the scavenger is injected into IPI with a carrier flow of synthetic air. The carrier air flow (∼ 4000 sccm) is maintained at all times to keep the conditions in IPI constant. The HORUS inlet samples approximately 7 L min −1 of air directly from the centre of the flow. There are several critical parameters involved in the deployment of this chemical scavenger methodology such as the identity and concentration of scavenger, the IPI sampling flow and therefore the residence time within IPI, and the synthetic air carrier flow. The choice of scavenger and concentration is very important. The OH scavenger must react quickly with OH but slowly with other oxidants like ozone and NO 3 , it should not be toxic and not have a high absorptivity at the laser excitation wavelength. Its concentration should be high enough to affect the removal of a known and substan-tial proportion, > 90 %, of atmospheric OH but should not be in excess to prevent the risk that excess scavenger will react with the internally generated OH. The flow rate of ambient air through IPI must be fast to minimise losses of HO x (OH and HO 2 ) onto the walls, and the residence time of OH within IPI has to be an optimal compromise between being short compared to the atmospheric lifetime of OH and allowing sufficient time for the scavenger to react. The carrier flow must be high enough to favour efficient mixing between the scavenger and the atmospheric air and to flush the lines when no injection of scavenger takes place. Figure 3 shows the schematic layout of the IPI during the HOPE 2012 campaign. The flow of the scavenger is controlled with a mass flow controller (MFC). After the MFC the scavenger line combines with the carrier gas line where it gets mixed. The mixture then reaches IPI where it is injected into the sampled atmospheric air. In both parts of the injection cycle, i.e. scavenger on and scavenger off, the same amount of air is sampled through IPI and the flow of carrier air is maintained constant. To remove residual scavenger from IPI lines when switching to a period with no injection, the lines are purged with synthetic air at a flow of approximately 5000 sccm for 5 s. The current IPI cycle results in a minimum time resolution for the measurement of atmospheric OH of one data point over 4 min consisting of cycles of 2 min of injection of the OH scavenger and 2 min with no injection of the OH scavenger. Table 1 shows IPI parameters for three measurement campaigns. Details of the three measurement sites are given in the next section. During both HUMPPA-COPEC 2010 and DOMINO HO x the prototype IPI version was in use. The main difference between the prototype version and the final design in current use, shown in Fig. 3, is the method of scavenger injection. The prototype IPI version injected the scavenger through eight 1/16 inch stainless steel tubes inserted into the centre of the IPI airstream. During these two campaigns the IPI parameters were the same. Propene (Aldrich 295663-330G, 99+ % purity) was used as an OH scavenger with a flow of 20 sccm and was carried to IPI with 4000 sccm of synthetic air (Westfalen AG). Total IPI sample flow was ∼ 280 L min −1 leading to a concentration of propene of 6.4 × 10 14 molecules cm −3 . The residence time between the injection of the scavenger and the instrument inlet was ∼ 2.5 ms, short compared to the lifetime of OH in those environments (on average ∼ 80 ms; (Nölscher et al., 2012;Sinha et al., 2012). Under these conditions, more than 95 % of the atmospheric OH was scavenged within IPI. During HOPE 2012 the current version of IPI was used (Fig. 3). The current version was designed with a simplified layout to reduce the number of connections and improve ease of use. Two OH scavengers were tested during the campaign. The main scavenger used was propane (Air Liquide 3.5, 99.95 % purity) applied with an average flow of 17 sccm, a carrier flow of synthetic air of 4000 sccm and a sample flow within IPI of ∼ 150 L min −1 (propane concentration ∼ 2.5 × 10 15 molecules cm −3 ). Pulling a smaller flow of atmospheric air through IPI led to a residence time after the injection of the scavenger of ∼ 4 ms, that was still short compared to the average OH lifetime at the site (on average ∼ 300 ms) and that allowed the use of a smaller concentration of scavenger preventing excessive titration of OH in the low-pressure side of the instrument. With this concentration of propane, the lifetime of OH was 0.3 ms and a scavenging efficiency of ∼ 90 % was achieved. Propene was also used for some measurement cycles, for purposes of comparison, every few hours.

Measurement sites
We present measurements from three measurement sites represented by various meteorological and physicochemical characteristics. The HUMPPA-COPEC 2010 (Hyytiälä United Measurements of Photochemistry and Particles in Air-Comprehensive Organic Precursor Emission and Concentration study) campaign took place during the summer of 2010 at the SMEAR II station in Hyytiälä, Finland (61 • 51 N, 24 • 17 E, 181 m a.s.l.) in a boreal forest dominated by Scots Pine (Pinus Silvestris L.). Continuous measurements of several trace gases and meteorological parameters as well as particle size distribution and composition (Junninen et al., 2009) were available. For the first part of the campaign the HORUS instrument measured side-by-side with an OH CIMS operated by the University of Helsinki (Petäjä et al., 2009). During the inter-comparison, the HORUS detection axis ( Fig. 1) was located next to a white container where the main body of the CIMS instrument was at a distance of less than a metre from the CIMS inlet and with the sampling position at the same height. Container and instruments were located in a clearing surrounded by a pine forest. More specific information about the meteorology observed during the comparison period are given in Hens et al. (2014). The instrumentation and the meteorological conditions during the campaign are described by Williams et al. (2011). The DOMINO HO x campaign took place in November 2010 in El Arenosillo, in southwestern Spain (37 • 1 N, 6 • 7 W, 40 m a.s.l.) at the same site as the DOMINO (Diel Oxidants Mechanisms In relation to Nitrogen Oxide) campaign in 2008 described in Crowley et al. (2011). The site is located in a forested area (Stone pines, Pinus pinea, 5-10 m in height) close to the South Atlantic Ocean shore and 12 km from the city of Huelva and associated petrochemical industry. The HOPE 2012 (Hohenpeißenberg Photochemistry Experiment) campaign was conducted during the summer of 2012 at the Meteorological Observatory in Hohenpeissenberg, Bavaria (47 • 48 N, 11 • 2 E). The observatory is operated by the German Weather Service (DWD) and is located at an altitude of 985 m a.s.l., about 300 m above the surrounding terrain, which consists mainly of meadows and forests. During the entire campaign the HO-RUS instrument measured side-by-side with the OH CIMS operated by the German Weather Service (DWD; Berresheim et al., 2000). HORUS detection axis and inlet ( Fig. 1) were located on the roof of the building at less than a metre distance from the inlet of the CIMS. The CIMS main body is located in the room below the roof and its inlet is at ∼ 20 cm height from the ground. Nothing was positioned around the two instruments that were both located on the same corner of the roof opening toward a pine forest. More information about the site and the routine measurements can be found in Handisides et al. (2003).
Both CIMS instruments used during HUMPPA-COPEC 2010 and HOPE 2012 campaigns are based on the instrument described by Berresheim et al. (2000) and in both systems a titration with propane is required in order to measure atmospheric OH. The accuracy of both instruments is determined by the accuracy of the calibration system that includes both instrumental uncertainty and interferences caused by ambient parameters. The accuracy of the CIMS operated by the University of Helsinki is 32 % (1σ ) (Hens et al., 2014) and the accuracy of the DWD-CIMS is 30 % (1σ ). The precision of the instruments considered counting statistics, potential wind and chemical interferences caused by NO, NO 2 , CO and hydrocarbons. The precision of the DWD-CIMS instrument is 26 % (1σ , 30 s data) while the precision for the CIMS operated by the University of Helsinki is based on 30 min average data and their variability, calculated for every single point; for a typical OH concentration of 1 × 10 6 molecules cm −3 the precision was 5 × 10 5 molecules cm −3 (Hens et al., 2014).

IPI characterisation
The addition of IPI to HORUS has a significant effect on the performance of the instrument with respect to losses of radicals in the inlet system. To account for the perturbation of the atmospheric OH measurement caused by the use of IPI, a number of tests were completed to assess the effects of changing instrumental parameters. Total radical loss on the IPI system, the variation of this loss with the atmospheric air sampled and with scavenger carrier flow, and the efficiency of OH removal by the scavenger were tested during the different campaigns in which HORUS was in use with IPI. The tests described below were performed during the HOPE 2012 campaign with the current version of IPI (Fig. 3). Titrations of atmospheric OH within IPI using different scavengers were completed with a stable source of OH radicals. The total radical loss in IPI was performed using ambient air as no artificial OH source is currently available due to the very high flow rates of zero air required. These tests were repeated and monitored through the entire campaign to obtain robust results. Tests were performed during daytime, between 10:00 and 16:00 LT, and night time after 20:00 to distinguish the effect of IPI with high radical load, i.e. during the day, from situations where interferences might dominate the total signal, i.e. during the night. Figure 4a and b show the average results of the radical loss tests checked by routinely measuring with and without IPI mounted on the inlet, conducted multiple times during the day and the night, respectively, with a sampled flow of ∼ 150 L min −1 and a carrier gas of 4000 sccm. The average total OH signal loss was 27 % during the day and 7 % during the night. The error bars for every single point represent the variability of the data (1σ ) during a single test. As the measurement characterising the losses of the total OH signal were completed during daytime when the ratio of atmospheric OH to background OH was highest, most of the variability is caused by the ambient variability of OH. Figure 4c shows the loss of the total OH signal while changing sample flow through IPI with no carrier air and scavenger. The minimum loss, 20 %, occurs at sample flows larger than 500 L min −1 while at the flow in use during the campaign, 150 L min −1 , the loss observed was ∼ 30 %. Figure 4d shows the losses of the total OH signal in IPI with the variation of the additional carrier flow used to mix the scavenger with the sampled atmospheric air while sampling 150 L min −1 of air. No dependency on the carrier flow rate between 1000 and 5000 sccm and no additional loss of OH compared to the loss due to the sampling of 150 L min −1 of air were observed, indicating that the major cause of losses is due to contact with surfaces. The measured losses during daytime, on average 27 %, are the losses of the total OH signal and therefore the sum of the losses of atmospheric OH and losses of the species causing the background OH. During night time an average loss of 7 % was measured from the total OH signal; since the atmospheric OH signal was below the limit of detection of the instrument (4 × 10 5 molecules cm −3 , 4 min data) during the tests described, the loss is assumed to be entirely due to the species causing the background OH (L bg ). The loss of atmospheric OH (L OH ) is equal to the atmospheric OH signal measured with IPI mounted on the top of the inlet (OH IPI atm ) divided by the atmospheric OH signal measured without IPI mounted on the top of the inlet (OH NoIPI atm ): The value of atmospheric OH without IPI mounted on the top of the inlet is impossible to measure but, by assuming that L bg is constant, the loss on the atmospheric OH signal can be calculated using The OH IPI tot and OH IPI bg are the total OH signal and the background OH signal, respectively, measured by the instrument with IPI mounted on the inlet and therefore affected by losses, and their difference is the atmospheric OH affected by losses, OH IPI atm . The OH NoIPI tot is the total OH signal measured during the tests without IPI on top of the inlet and therefore not affected by losses. Figure 5 shows an example of a test during which IPI was physically removed from the top of the instrument inlet (shaded area). IPI can be removed easily in less than 2 min and, on average, we measured without IPI during one period for 30 min and measurements were repeated routinely every 4 days during the HOPE 2012 campaign. Referring to Eq. (5), for one IPI on and off test, the numerator (the atmospheric OH signal affected by losses) is the result of the interpolation of the measured atmospheric OH signal 12 min before and after the removal of IPI. In the denominator, the total OH measured when IPI was not on the inlet is the average of the signal measured when IPI was not on the top of the inlet. The background OH measured with IPI is obtained from the interpolation of the measured background OH signal 12 min before and after the removal of IPI and is divided by the loss of the background OH measured during night time tests. The average value obtained for the loss of the atmospheric OH is 34 % and the data have been corrected accordingly. The variability observed for the average L OH value, ±15 % (1σ ), was taken into account for the accuracy of the HORUS instrument after the addition of IPI. The accuracy of HORUS was 34 % (2σ ) and the accuracy of IPI-HORUS becomes 42 % (2σ ). The detection limit of IPI-HORUS is obtained from the statistical significance of the difference between the total OH sig- nal and the background OH signal determined by t-test (1σ ) and decreased from 9 × 10 5 molecules cm −3 (4 min data) for the HUMPPA-COPEC 2010 and DOMINO HO x campaigns to 4 × 10 5 molecules cm −3 (4 min data) for the HOPE 2012 campaign due to higher laser power.
To optimise the OH scavenging efficiency, several titrations of OH were conducted during the campaign with various operational conditions to examine the stability of the instrument and the reproducibility of the background subtraction. The optimisation of the scavenger concentration is also important due to its potential to remove part of the internally produced OH in the low-pressure region of the instrument, which would result in the overestimation of the atmospheric OH concentration. The titration experiments were performed by producing a constant above ambient concentration of OH of about ∼ 10 9 molecules cm −3 in front of IPI using a mercury lamp and ambient air and by varying the concentration of the scavenger to measure the efficiency in the removal of the OH molecules within IPI. By sending a concentration of 2.5×10 15 molecules cm −3 of propane, 90 % of the initial OH was removed. We calculated the theoretical scavenging efficiency for each OH scavenger deployed during the campaign based on the residence time in IPI after the injection of the scavenger (∼ 4 ms) and inside the instrument in the low-pressure region (∼ 2.5 ms), and on the rate coefficients for the reactions between the respective scavenger and OH at ambient and low pressure (∼ 350 Pa) (Sander et al., 2011). Figure 6 shows, in blue, the theoretical OH titration efficiency of scavenger depending on the scavenger concentration for the case of propane. Also shown are experimental titrations performed with propane, at different carrier gas flow rates. The carrier gas flow does not have a significant impact on the dilution of the scavenger as it represents a minor percentage of the total flow sampled by IPI. What it does influence is the mixing of the scavenger with the atmospheric air, and thus the scavenging efficiency. The experimental data show a deviation from the calculated curve. At higher carrier gas flows the experimental data are closer to the model results. The deviation from the modelled data may be related to incomplete mixing between the sampled atmospheric air and the OH scavenger; by increasing the carrier gas flow, and therefore improving the mixing, we would expect to approach the theoretical titration efficiency. The setup during the HOPE 2012 campaign allowed us to have a stable carrier flow only for flows below 4500 sccm. At higher values, the MFC controlling the carrier flow showed high sensitivity to even small temperature-driven changes in the backing pressure of the supply gas. Therefore, the flow was kept stable at 4000 sccm, even though this is suboptimal with respect to mixing, and the amount of OH scavenged was checked regularly by repeated titrations. The calculated theoretical scavenging efficiency also predicts the amount of OH we would be removing in the low-pressure region of the instrument at a certain concentration of scavenger. During the HOPE 2012 campaign the concentration of propane in use was small enough to allow the removal of less than 2 % of the OH in the low-pressure region. As the calculation assumes perfect mixing, the amount of internal scavenging was tested by changing the propane flow from 5 to 35 sccm whilst keeping a constant and small concentration of atmospheric OH (night time period). Figure 7 shows a negligible variation in the background OH signal, well within the precision of the instrument, even when increasing the propane concentration by a factor of 7. Therefore, an impact from internal OH scavenging on the determination of the atmospheric OH can be excluded. The prototype IPI version was used during the HUMPPA-COPEC 2010 and the DOMINO HO x campaigns and the instrument was run with the same parameters in both campaigns. Similar tests to the ones previously described were performed during HUMPPA-COPEC 2010 but as shown in Fig. 8, most of the total OH signal measured by the instrument was due to the background OH both during day and night time. In this situation it is difficult to account for possible losses of the OH radical because its contribution to the total signal is too small. Total radical loss tests were performed but the results were not as unambiguous as for the HOPE 2012 campaign case; sometimes a small OH loss was observed on IPI but during most of the tests there was no clear indication of OH loss although any OH losses would likely be masked by the high background signal. Propene was originally selected as scavenger because of its high reaction rate with OH, allowing rapid OH scavenging even at low concentration. However, propene is known to form OH radicals after reaction with ozone with a yield of 0.34 (Atkinson et al., 2006). Hence, by mixing high concentrations of propene with the ambient air we expect formation of additional OH radicals. During HUMPPA-COPEC 2010 the residence time in IPI after the injection of the scavenger was ∼ 2.5 ms such that the concentration of ambient ozone and propene, even if reacting quickly with OH, can be assumed to be constant. The steady-state concentration of OH with the concentration of propene within IPI (∼ 6 × 10 14 molecules cm −3 ) is reached after 0.5 ms, therefore the average steady-state OH concentration produced by propene can be calculated by taking into account the average ambient ozone concentration during the campaign, the rate coefficient between ozone and propene, k 1 = 1 × 10 −17 cm 3 molecules −1 s −1 at 298 K, the OH yield, Y = 0.34, and the rate coefficient between propene and OH at 1013 hPa and 298 K, k 2 = 2.9 × 10 −11 cm 3 molecules −1 s −1 (Atkinson et al., 2006): This calculation represents an upper limit for the possible production of OH during the injection of propene as scavenger as it does not consider any physical losses for OH or ozone on IPI and also assumes perfect mixing between propene and ambient air. The OH concentration produced during the injection of propene depends only on the ozone concentration. The peak mixing ratio of ozone during the HUMPPA-COPEC 2010 campaign was 80 ppbv, which would result in a maximum OH concentration of 2.3 × 10 5 molecules cm −3 . For the average ozone value of 44 ppbv the steady-state OH concentration would be 1.2 × 10 5 molecules cm −3 . This additional OH would cause an underestimation of the atmospheric OH calculated after subtraction of the background OH signal from the total OH measured by the instrument and would increase the uncertainty on our OH measurement. During the HOPE 2012 campaign we performed tests using propene and propane as OH scavengers intermittently at an ambient concentration of ozone of ∼ 40 ppbv to estimate the production of OH due to propene ozonolysis. Figure 9 shows that there is no significant difference between the atmospheric OH concentration determined with the use of propane and the OH concentration determined using propene as the average value was (0.5 ± 1.5) ×10 5 molecules cm −3 (1σ ). This value is lower than the theoretical one calculated for the same concentration of ozone, indicating additional loss processes. Because the ambient concentrations of ozone and the propene concentration in use during the tests performed in HOPE 2012 and HUMPPA-COPEC 2010 were comparable, assuming similar mixing within the two versions of IPI, we can expect similar OH production of 0.5 × 10 5 molecules cm −3 during the HUMPPA-COPEC 2010 campaign. The value is below the precision of the OH measurement and therefore the data were not corrected for this effect.
To summarise, IPI prototype was in use during the HUMPPA-COPEC 2010 and DOMINO HOx campaigns. No correction factor, F , from Eq. (3) was applied for radical losses for the atmospheric OH data collected during those campaigns as no clear value for losses of atmospheric OH in IPI was obtained. As propene was used as scavenger aiming at the removal of more than 95 % of the atmospheric OH and no additional OH was produced while using it (Fig. 9) also no correction for the scavenger efficiency was applied. During HOPE 2012 the correction factor includes the 34 % average value of losses of atmospheric OH in IPI walls and a point-by-point correction for the scavenger efficiency. The scavenger efficiency is based on the scavenger used and its concentrations and it is obtained from the titration tests performed regularly during the campaign. Figure 8 shows the signals resulting from the first use of HO-RUS with IPI during the HUMPPA-COPEC 2010 campaign. The top time series highlights the period in which the instrument was on the ground next to a CIMS measuring OH and H 2 SO 4 (Petäjä et al., 2009), while the lower series shows the period in which the instrument was operated on a 24 m tower just above the forest canopy. During the day, within the partially shaded forest canopy, the background OH signal reaches up to 1 × 10 7 molecules cm −3 contributing 80 % to the total signal. On the tower the maximum value reached by the background OH signal is 7×10 6 molecules cm −3 contributing up to 60 % to the total OH signal measured on the majority of days. During night time the background OH concentration falls below 4 × 10 6 molecules cm −3 , but as the atmospheric OH concentration is small the fraction is almost 100 %. A side-by-side comparison with the CIMS instrument was performed for the first part of the campaign while both instruments were on the ground. Figure 10 shows the relationship between the atmospheric OH measured by the LIF and the OH measured by the CIMS. The correlation coefficient, R 2 = 0.4, is affected by the large scatter of the LIF OH data due to low laser power, a rapidly ageing detector, and the large contribution that the background signal makes to the total signal. Overall, the LIF measures higher OH values with a comparison slope of 1.30, however the difference is within the accuracy of the instruments (HORUS: 42 %, 2σ ; CIMS: 64 %, 2σ ) (Hens et al., 2014).

Atmospheric measurements using IPI
During the DOMINO HO x campaign in November 2010 (Fig. 11), the background OH signal was always below 4 × 10 6 molecules cm −3 contributing about 50 % to the total OH measured during the day and 100 % during the night. During the 3 days of measurements, two different wind sectors were sampled: air travelling from the city of Huelva and air travelling over the continent. There appears to be little difference between the contributions of the background signal in either of these wind sectors. Figure 12 shows the OH signals measured by the LIF during the HOPE 2012 campaign for a day at the beginning of the campaign (Fig. 12a) and a day at the end of the campaign (Fig. 12b). The data shown are representative of the concentrations generally observed during the campaign period. During the whole campaign the HORUS LIF was measuring side-by-side with a CIMS instrument (DWD-CIMS). The agreement between the two measurements of atmospheric Figure 11. OH signals measured by HORUS during the DOMINO HO x campaign. The blue circles represent the total OH signal measured in the absence of an OH scavenger. The red stars represent the background OH measured during the injection of an OH scavenger. The black crosses represent the atmospheric OH obtained by difference between total OH and background OH. The solid lines are 30 min averages. The two shaded areas represent the two prevailing wind directions: the blue area during wind from the continental sector and the red area from the Huelva sector. Time is in UTC.
OH is good (Fig. 13), with a correlation coefficient for the entire data set of R 2 = 0.81, a slope of 0.94 and an offset of 4.5 × 10 5 molecules cm −3 . The offset is partly caused by night time atmospheric OH observed during several nights by the HORUS instrument. When removing the night time data (between 18:30 and 07:30 UTC) from the correlation plot ( Fig. 13) it is still possible to observe a linearity in the data set with a correlation coefficient of R 2 = 0.70, a slope of 0.90 and an offset of 3×10 5 molecules cm −3 . This offset indicates that on average the HORUS instrument measured higher values of atmospheric OH radicals compared to the CIMS. No significant offset was observed during the HUMPPA-COPEC 2010 campaign, indicating that the discrepancy could originate from the DWD-CIMS instrument. The night time OH measured by HORUS is not constant: during some nights the two instruments agree, both showing an OH signal scattering around zero. Propane was used as scavenger under the same instrumental conditions during periods when HORUS did and did not measure zero atmospheric OH. No correlation between the atmospheric OH measured during night and the background OH was observed. In addition, preliminary budget calculations suggest that as the OH reactivity measured was very small during night time, the production of OH via recycling of HO 2 through NO and from ozonolysis of VOCs could lead to a concentration of ∼ 3 × 10 5 molecules cm −3 , similar to what is detected by HORUS. However, it is still unclear where the offset is generated from and further investigations are needed. The background OH signal during the campaign ranged from a minimum of 1 × 10 6 to a maximum of 7 × 10 6 molecules cm −3 on 2 days during which forest cutting was performed near the site but it was, for the most part, below 4 × 10 6 molecules cm −3 . The background signal   York et al. (2004), and yields a slope of 0.94 ± 0.01 and an offset of (4.5 ± 0.06) ×10 5 molecules cm −3 . contributed between 20 and 40 % to the total OH signal during daytime and up to 100 % of the total OH signal during night time.
The three measurement campaigns show large differences in the background OH signal and its contribution to the total OH signal measured by HORUS. The smallest contri-bution to the total OH was observed during HOPE 2012, where the total OH measured by HORUS during daytime for most of the days would have agreed with the DWD-CIMS instrument within the accuracy of the instruments even without the chemical scavenger method. At the other extreme lies the HUMPPA-COPEC 2010 campaign where the background OH signal within the forest canopy reaches 1 × 10 7 molecules cm −3 on top of a smaller atmospheric OH concentration, often below 2 × 10 6 molecules cm −3 . The relative contribution of the background signal is lower for the measurement period on the tower where the atmospheric concentration of OH is higher due to larger values of j (O) 1 D compared to the location below the canopy (Hens et al., 2014). One difference between the conditions in HOPE 2012 and HUMPPA-COPEC 2010 is the measured OH reactivity, which was relatively high during HUMPPA-COPEC 2010, on average 12 s −1 with peaks of over 40 s −1 (Nölscher et al., 2012) and often below the detection limit during HOPE 2012 (on average 3.5 ± 2 s −1 ). The average concentration of measured BVOCs (isoprene, (−)/(+) α-pinene, (−)/(+) βpinene, 3-carene, myrcene, sabinene) is similar for the two campaigns; approximately 300 pptv (Hens et al., 2014) with HUMPPA-COPEC 2010 showing higher concentrations of monoterpenes and large emissions rates of sesquiterpenes (Yassaa et al., 2012) compared to the HOPE 2012 campaign in addition to unexplained OH reactivity (Nölscher et al., 2012) that indicates the presence of unmeasured VOCs . The interpretation of the DOMINO HO x data is more complex; this campaign was at the same site of the DOMINO 2008 campaign but only a few trace gases (ozone and NO x ) and some meteorological parameters were measured in 2010. During the DOMINO 2008 campaign HORUS was in use without the injection of a chemical scavenger and, due to the observation of an interfering signal during DOMINO HO x , the OH concentration measured previously should be considered an upper limit. Because DOMINO HO x was performed during the same month as the DOMINO 2008 campaign and no sign of differences in local pollution (street work, new buildings next to the site, etc.) or unusual weather was observed, we expect a similar amount of background OH, i.e. about 50 % of the total OH measured in DOMINO HO x , for the DOMINO 2008 campaign. During DOMINO 2008 relatively high OH reactivity was measured, with an average of approximately 18 s −1 , as well as low concentrations of measured BVOCs (mainly isoprene, eucalyptol and (−)/(+) camphor) consistent with low emissions from vegetation during fall, in the range of 50 pptv  with isoprene being the most abundant BVOC measured. The concentration of anthropogenic VOCs (AVOCs) measured was on average 400 pptv consisting mainly of benzene and toluene, although only a fraction of the AVOCs was quantified on this campaign. The highest reactivity was found for air masses arriving from the continental sector due to the likely presence of reactive oxidation products formed from primary anthropogenic emissions, followed by the air coming from the Huelva sector because of the load of AVOCs . As shown in Fig. 11, there is no clear difference between the background OH contributions to the total OH signal for the air arriving from those two different wind sectors indicating a similar influence on the background OH signal and a small contribution when compared to the OH background observed during the HUMPPA-COPEC 2010 where the measured BVOC concentrations were higher.
In summary, when the instrument was located within the canopy of a monoterpene-dominated forest environment with high BVOC concentrations and high OH reactivity, the OH measurements with HORUS was strongly affected by an interference resulting in a high background OH signal. Aged air masses containing oxidation products from anthropogenic emissions and primary AVOCs such as benzene and toluene measured in high concentrations during DOMINO 2008 seem to give rise to a smaller background OH signal. The HUMPPA-COPEC 2010 campaign was an extreme case where most of the total OH measured by the instrument was due to the background OH signal, in part because of the large background, but also because the atmospheric OH concentration was low. In contrast, during DOMINO HO x the total OH signal was not completely dominated by the contribution from the background signal even though the atmospheric OH concentration is comparable to that observed during HUMPPA-COPEC 2010. During HOPE 2012, a higher OH concentration in combination with a relatively low back-ground OH signal, comparable to that in DOMINO HO x , makes the contribution of the background OH to the total OH small during daytime. It is also evident that for all three measurement campaigns performed with IPI nearly the entire nocturnal OH signal detected is due to the background OH in the instrument and not due to atmospheric OH.
The background OH has an important impact on the measurement of the atmospheric concentration of the OH radical and it is likely to be observed also in the HO 2 cell. During the HOPE 2012 campaign a background OH concentration of 1 × 10 6 molecules cm −3 was observed in the HO 2 cell when no injection of NO was performed. As the atmospheric concentration of HO 2 is of the order of ∼ 10 8 -10 7 molecules cm −3 the impact of this interference is below the precision of the instrument (1.9 × 10 7 molecules cm −3 ; Hens et al., 2014) and does not affect the HO 2 radical measurements.

Instrumental tests
The background OH measured with the HORUS instrument changes during the day and varies with different environments. As mentioned in the introduction, LIF-FAGE instruments are known to be affected by interferences (Holland and Hessling, 1995;Martinez et al., 2004); most of the known interferences are caused by laser photolysis or by fluorescence of other molecules in the vicinity of 308 nm wavelength. An interference caused by laser photolysis can be detected by observing a square dependency of the signal with laser power while a spectral interference can be eliminated by tuning the excitation laser on and off resonance with the OH transition line at 308 nm. Interferences caused by laser photolysis occur when an atmospheric trace gas is photolysed by one photon of the laser beam and produces OH directly that can then be excited by a second photon of the same laser. This could happen, for example, with HONO, HNO 3 and ROOH. Other species such as ozone and acetone can also be photolysed by the laser beam but, as the photolysis does not produce OH directly, the laser pulse of HORUS is too short (15 ns) to allow the excitation of the resultant OH to happen within the same pulse. If the transport of these OH molecules out of the detection volume is sufficiently fast, the subsequent laser pulse will not be able to excite them. If the transport time is not sufficient, the subsequent pulse will excite the formed OH. Both laser-generated interferences will show a square dependency with the laser power. Although interferences tests were performed on GTHOS , the HORUS instrument based on GTHOS also makes use of a White cell to increase the sensitivity of the HO x measurements. Test results for the species listed above  showed that the ambient concentration needed for these species to produce a detectable OH interference in the instrument has to be well above "usual" ambient conditions and higher than the values observed during the field campaigns described in this paper. Nevertheless, to test whether the origin of the background OH signal is photolytic, during the HUMPPA-COPEC 2010 and HOPE 2012 campaigns the laser power was varied by a factor of 4. The background OH signal did not show any dependency when plotted against the laser power at the OH cell (Fig. 14). Therefore, we can exclude that the background OH signal is generated by the laser within the main air flow.
Since the residence time of an air parcel in the detection area of the White cell of HORUS instrument is more than two times shorter than the time period between two consecutive laser pulses at a laser repetition rate of 3000 Hz, interferences due to double pulsing are unlikely to happen. Still, if air is collected in pockets in the White cell arms (Fig. 15) for an extended period of time there could be production of OH. To inhibit this, the arms of the White cell are constantly flushed with synthetic air that avoids deposition of particles on the mirrors and prevents the air from becoming stagnant. In addition, baffles are mounted between the arms of the White cell and the detection cell helping to reduce the scatter light of the laser and reduce the opening between the White cell arms to the detection cell letting less air to pass. To confirm that we are not affected by formation of OH in pockets in the White cell arms, the flushing flow inside the White cell was increased by a factor of 2. Figure 16 shows a test completed during HOPE 2012 in the evening, when the background OH was measured for different flushing flows in the cell. There is a general decreasing trend in the background OH due to ambient variability of the signal but no difference is observable when changing the flushing in the cell for two consecutive periods. No difference in the background OH was also observed when comparing the signal detected during night time with and without baffles and while flushing the White cell pockets with C 3 F 6 , an OH scavenger, instead of with synthetic air. This confirms that the background OH signal in the detection volume is not formed in air pockets inside the White cell.  The background OH is also not formed by a leakage/back flushing of NO into the OH cell during the injection in the HO 2 cell as no difference in the signal was observed during two consecutive periods with and without NO injection.
Finally, the possibility of formation of OH from reactions happening on walls was investigated by changing the material of the instrument inlet and by coating the inside walls with heavy water and detecting OD, but again no significant difference in the signal was observed nor was OD detected.
The results of the previously described tests show that it is unlikely that the background OH is an instrumental artefact and it is likely that the signal measured during field campaigns is caused by one or more atmospheric trace gases sampled by the instrument.

Criegee intermediates hypothesis
The pressure inside the instrument (∼ 3.5 hPa) is such that bimolecular reactions with the concentrations of trace gases in the atmosphere are negligible in the transient time in the instrument. As laser photolysis and spectral interference can also be excluded from generating an OH signal within the Figure 17. Ozonolysis test of propene in the presence of propane as an OH scavenger. The green shaded area represents the signal observed by HORUS when only ozone was injected in the flow tube. The yellow area shows the OH signal generated by photolysis of water. The first blue shaded area shows the reduction of the OH signal generated by photolysis of water after the addition of propane to the flow tube. The pink shaded area shows the signal measured while injecting ozone and propene into the flow tube. The second blue shaded area shows the small reduction observed in the signal when injecting propane into the flow tube to scavenge the OH produced.
HORUS instrument, one possible explanation for the background OH observed could be the unimolecular decomposition of an atmospheric trace gas species inside the instrument with formation of OH. Possible candidates are Criegee intermediates formed during the ozonolysis of unsaturated compounds. Criegee intermediates are known to promptly decompose and produce OH at low pressure (Criegee, 1975;Neeb and Moortgat, 1999;Donahue et al., 2011;Vereecken and Francisco, 2012). To test for this within the HORUS instrument, an ozonolysis experiment with propene was performed at ambient pressure (930 hPa) and temperature (293 K) in a flow tube connected to the inlet of HORUS. Initially, ozone (1.3 × 10 13 molecules cm −3 ), produced by passing pure oxygen in front of a mercury lamp, was injected into the flow tube (Fig. 17, green shaded area) at ambient pressure. Subsequently, a high concentration of OH was produced from photolysis of water by a mercury lamp, and introduced into the flow tube (Fig 17, yellow shaded area). Propane was then injected into the flow tube at a concentration of 2.5 × 10 16 molecules cm −3 , enough to remove 99 % of the observed OH (Fig. 17, blue shaded area). The mercury lamp was then removed and the propane injection was stopped. Only ozone and propene were injected into the flow tube (Fig. 17, orange shaded area). It is possible to observe a high OH signal generated from the ozonolysis of propene. To assess whether the OH is only formed in the flow tube at ambient pressure or partly in the low-pressure segment of the instrument, the same concentration of propane used to remove 99 % of the OH produced from the photolysis of water was injected. It is possible to observe that, despite sending the same amount of propane, there is only a small reduction in the OH signal. As the concentration of propane injected was sufficient to remove the OH generated in the flow tube at ambient pressure, the remaining OH still observed by HO-RUS has to be generated inside the low-pressure segment of the instrument. This test indicates how the HORUS instrument is sensitive to OH formed within the instrument during ozonolysis of propene. A more detailed description of the experimental setup together with modelling and a thorough description of the chemistry of Criegee intermediates and investigation of their impact on the IPI-LIF-FAGE can be found in . To understand whether the signal observed by HORUS during the ozonolysis of propene in the presence of an OH scavenger originated from walls or pockets of air in the White cell, the experiment was repeated using a single-pass setup. The dependency of the signal on the laser beam position within the detection cell was tested by moving the laser beam from the centre of the cell towards the sides. Figure 18 shows no clear dependency of the signal detected during ozonolysis of propene in the presence of an OH scavenger on the position of the laser beam. As for the background OH observed during field measurement, this confirms that the signal is not due to the White cell structure. The previous test indicates that the HORUS instrument is sensitive to OH formed within the instrument during ozonolysis of propene but more laboratory and field tests are necessary to completely ascertain whether the background OH observed during field campaigns entirely originates from the same chemical processes.

Possible influence on earlier measurements
The background OH signal depends on the type of environment and appears to be strongly related to the VOC concentration and type of VOC prevalent. It is possible that previous campaigns performed with LIF instruments without applying a chemical scavenger method have been affected by an interfering species in a similar manner as described in this manuscript. However, as underscored previously by Mao et al. (2012), the design of each particular LIF-FAGE system is likely to determine whether, and to what extent, the instrument suffers from this interference and so we will confine this discussion to the HORUS instrument.
The environment of previous campaigns may give an indication as to whether the measured OH was affected by significant inferences. The first campaign with HORUS consisted of a formal blind comparison, HOxCOMP, between several LIFs, a CIMS and a DOAS instrument both in a chamber and in ambient air. The results of the campaign are comprehensively described in Schlosser et al. (2009). During daytime, the agreement between all the instruments in the chamber was good with a regression slope between the MPI LIF and the FZJ-DOAS of one, but in ambient air the MPI LIF instrument measured higher concentrations of OH than the CIMS (Schlosser et al., 2009). As we noticed an unattributed change in the calibration factor of the instrument of 30 %, we cannot exclude a change in the same order of magnitude in the calibration source between the period in the chamber and in ambient that might explain the difference between the OH concentration measured by the MPI LIF and the one measured by the CIMS in ambient air. Night time data from HORUS are not shown either from the chamber period or in ambient air because of large unexplained measured OH signals up to 4 × 10 6 molecules cm −3 (Kubistin, 2009). Based on what we have learned since the use of IPI, it is likely that the high night time signal was due to a chemical interference in the HORUS instrumenthowever, during daytime its concentration appeared to be low enough in that specific environment as to not produce a significant OH interference (i.e. within the accuracy of the instrument). Two subsequent campaigns, using HORUS, were performed without IPI. GABRIEL, an aircraft-based campaign, took place in October 2005 over the tropical rain forest in equatorial South America . Measured OH was much higher than predicted by a traditional chemical mechanism. Further analysis indicated that OH might be recycled within the isoprene degradation scheme (Lelieveld et al., 2008;Kubistin et al., 2010;Taraborrelli et al., 2012). Although without IPI we cannot completely rule out a possible interference for our measurements, laboratory tests and quantum mechanical calculation as follow-up studies to GABRIEL provided evidence for the proposed OH recycling, which was previously not accounted for da Silva, 2010a;Peeters and Müller, 2010;Crounse et al., 2011). Even though laboratory (Crounse et al., 2011) and chamber studies (Fuchs et al., 2013) predict a smaller concentration of OH from the recycling of isoprene than reported in Lelieveld et al. (2008), measurements with a different LIF-FAGE instrument in the Borneo rainforest during the OP3 campaign (Whalley et al., 2011) also showed large discrepancies between measured and modelled OH using the traditional chemical mechanism. The elevated concentration of OH measured during OP3 was supported from co-measurements of formaldehyde and glyoxal made with a DOAS (Whalley et al., 2011). In addition, a side-byside airborne comparison between the aircraft configuration of the GTHOS LIF instrument (ATHOS -Airborne Tropospheric Hydrogen Oxides Sensor) employed without a chemical scavenger method and a CIMS , showed good agreement even at higher levels of isoprene. The aircraft campaign HOOVER, performed without IPI in the upper troposphere across Europe in September 2007 (Regelin et al., 2013), showed good agreement between the measured OH and a simple box model. This, together with the observations during other HORUS field campaigns and preliminary laboratory tests, suggests that the background OH observed by HORUS might be related to shorter-lived species, which likely do not have a significant impact on the upper troposphere due to the relatively large distance between the emission source and the measurement point.

Conclusions
An improved methodology to measure the OH radical with a LIF-FAGE instrument has been developed and deployed in three different environments. Results show that the use of the IPI-LIF-FAGE technique for HORUS results in good agreement with OH data measured with two different CIMS instruments during two campaigns. A thorough and careful characterisation of the operational parameters was necessary to find the optimum conditions to avoid inefficient mixing of the scavenger and the sampled air, excessive titration of OH in the low-pressure side of the instrument and large losses of OH on the walls of IPI. The best results were achieved when using propane as OH scavenger in a concentration of 2.5 × 10 15 molecules cm −3 with a carrier gas flow of at least 6000 sccm and a residence time after the injection of the scavenger of ∼ 4 ms. The use of a chemical scavenger method revealed the presence of a background OH signal that, using the same calibration factor as for atmospheric OH, spanned a concentration of 5×10 5 to 1×10 7 molecules cm −3 in the environments described. Without the chemical scavenger method the atmospheric OH measured during the HUMPPA-COPEC 2010, DOMINO HO x and HOPE 2012 campaigns, during day and night time, would have been overestimated. Laboratory studies with HORUS  show how the device is sensitive to OH formed within the instrument from unimolecular decomposition of Criegee intermediates, but more laboratory and field tests are necessary to clarify whether they represent the only source of the background OH observed during the field campaigns. It is, though, already clear that the background OH has a strong connection with the type of environment in which the instrument is deployed. Although it is very likely that the presence and extent of a chemical interference in different LIF-FAGE systems for the measurement of OH are dependent on the particular instrument design, our experience shows that the determination of the background OH should be a prerequisite for these systems and the ambient measurement of OH.