Measurements took place inside a botanic garden close to the small town Wangdu in China between 7 June and 8 July 2014. Wangdu is located in the densely populated North China Plain but does not have major industry itself. Major cities are located mainly in the sector northeast to southwest of Wangdu, whereas there is a mountainous area with less industry northwest of Wangdu. The closest large city is Baoding, 35 km northeast of Wangdu. The measurement site had a distance of 2 km from a road with only local traffic. The botanic garden was surrounded by agricultural fields. Trace gases from local biogenic emissions of trees, bushes and from farming can be expected.

The site was chosen because it was not directly influenced by strong close-by anthropogenic emissions or the direct outflow of a big city. However, it was expected to observe regionally transported pollution in the North China Plain. Instruments were housed in seven shipping containers, which were partly stacked up so that inlets of instruments were at a height of 7 m above the ground.

A large number of instruments characterized meteorological conditions, trace gas concentrations and aerosol properties. The measurements used for the OH reactivity analysis are listed in Table .

OH and HO2 radical concentrations were measured by a newly built instrument of Peking University (PKU) applying laser-induced fluorescence (PKU-LIF) . This instrument detects OH fluorescence by time-delayed single photon counting after excitation by short laser pulses at 308 nm in a low-pressure cell . HO2 radicals are detected as the sum of OH and HO2 (= HOx) after chemical conversion to OH in the reaction with nitric oxide (NO). In order to avoid significant simultaneous conversion of organic peroxy radicals (RO2) , the amount of NO was adjusted to yield an HO2 conversion efficiency of only 6 %. The instrument sensitivity was calibrated every 3 to 4 days by a custom-built calibration source described in detail in .

A commercial cavity ring-down instrument (Picarro model G2401) monitored CO, CH4 and H2O concentrations. Concentration measurements of ozone by two commercial UV absorption instruments (Environment S.A. model 41M; Thermo Electron model 49i) agreed well within their accuracies during the campaign. Nitrogen oxides (NO and NO2) were also detected by several instruments applying chemiluminescence (Thermo Electron model 42i NO-NO2-NOx analyzer and Eco Physics model TR 780) that were equipped with a photolytic converter. Daily calibrations were performed using a certified gas standard. The field measurements differed on average by 20 %. Measurements of the Thermo Electron instruments appeared to be more precise and are used here (see , for details). Because the reason for the disagreement could not be identified, the 20 % difference adds to the uncertainty in NO measurements here.

Nitrous acid (HONO) concentrations were simultaneously measured by several instruments applying different measurement techniques . Custom-built instruments from FZJ (Forschungszentrum Jülich) and from PKU utilized long-path absorption photometry (LOPAP). In addition, three custom-built instruments applied cavity-enhanced absorption spectroscopy (CEAS) for the detection of HONO. They were operated by the US National Oceanic and Atmospheric Administration (NOAA) , by the Anhui Institute of Optics and Fine Mechanics (AIOFM), and by the University of Shanghai for Science and Technology (USST). A gas and aerosol collector (GAC), which is based on the wet denuder/ion chromatography technique, could also detect HONO . Only measurements by the two LOPAP instruments and the CEAS by NOAA resulted in good data coverage. The agreement between these instruments was diverse. Differences were often less than 30 % but could be as high as a factor of 2 for certain periods (several hours). The reason for the disagreement during these times is not clear. For the purpose of the analysis here, measurements by the LOPAP instrument from Forschungszentrum Jülich are used because this instrument showed best data coverage and the lowest detection limit. This instrument was calibrated by using a liquid standard as described in every 10 days. The choice of the HONO data set has a rather small impact on the calculated OH reactivity, as well as on the calculated total OH production rate, which was dominated by OH recycling from HO2 during the daytime (see below).

For the analysis of the OH reactivity, measurements of organic trace gases are essential. In total, 59 organic species (C2-C11 alkanes, C2-C6 alkenes, C6-C10 aromatics and isoprene) were detected by a custom-built gas-chromatography system equipped with a flame ionization detector (FID) . Full calibrations using certified gas standards (Air Environmental Inc., Spectra Gases Inc.) were done before and after the campaign. Drifts of the sensitivity during the campaign were accounted for by measuring the instrument sensitivity for bromochloromethane, 1,4-difluorobenzene, chlorobenzene and 1-bromo-3-fluorobenzene every second day. Formaldehyde (HCHO) was detected by a commercial Hantzsch monitor (Aerolaser model AL4021) and glyoxal (CHOCHO) by a custom-built cavity-enhanced spectrometer . In addition, acetaldehyde and the sum of methyl vinyl ketone (MVK) and methacrolein (MACR) were measured by a commercial proton transfer reaction–mass spectrometry system (PTR-MS, Ionicon). Some of the species or family species were simultaneously detected by the gas-chromatography (GC) system and the PTR-MS (isoprene, benzene, toluene, styrene, C8-aromatics, C9-aromatics). Measurements during the daytime agreed well within 30 to 50 % . Calibration of the PTR-MS instrument was done every day using a certified gas standard (Air Environmental Inc.).

Photolysis frequencies were calculated from the spectral actinic photon flux density measured by a spectrometer that was calibrated against absolute irradiance standards .

The OH reactivity instrument measures directly pseudo first-order loss rate coefficients (Eq. ) of OH in the ambient air. The measurement is based on artificial OH generation by pulsed laser flash photolysis (LP) of ozone in ambient air combined with the detection of the temporal OH decay by LIF. The method was initially developed for field application by and is applied today by several other groups . The instrument deployed in this campaign is similar to the instrument described in , which was used for measurements in our two field campaigns in 2006 in China. Since then, a second instrument has been built specifically for the deployment on a Zeppelin NT airship , but it can also be operated on the ground. This instrument was deployed. Figure  gives a schematic representation of the instrument without the pump (Edwards model XDS35i) needed for the operation of the low-pressure LIF cell and without the laser that provides the 308 nm radiation for the excitation of OH. The 308 nm radiation is delivered by the dye laser system that is also used in the instrument for the OH HO2 and RO2 concentration measurements described in . This laser has three output fibers to provide laser light, one of which is used for the OH reactivity instrument.

The kOH instrument is mounted in a 19′′ rack that was placed inside one of the upper shipping containers at the field site. The inlet line (outer diameter 10 mm, length approximately 6 m) was made of stainless steel that had a SilcoNert 2000 coating. Such a sampling line has been used for OH reactivity measurements in the Jülich atmosphere simulation chamber SAPHIR for many years without notable effects on measurements. Approximately 20 Lmin-1 of ambient air is sampled through a flow tube made of anodized aluminium (length: 60 cm; inner diameter: 4 cm). Downstream of the flow tube, the flow rate is measured by a flowmeter and controlled by a blower.

The pressure inside the flow tube is 1 atm, and the temperature was the same as in the field container (between 22 and 30 ∘C). Ambient temperature was higher with up to 38 ∘C for some periods during the campaign. Differences in temperature and pressure potentially effect the measured reactivity due to changes in the reactant concentrations and in reaction rate constants . Measured reactivities were corrected for changes in the reactant concentration calculated from measured ambient and flow-tube temperature and pressure values (corrections were less than 2 %). Sensitivity studies taking either ambient temperature or flow-tube temperature for the calculation of OH reactivity from measured OH reactant concentrations (see below) indicate that the effect of temperature differences on reaction rate constants resulted in changes in the OH reactivity of typically less than 1 % (maximum values 4 %) for conditions of this campaign.

High OH concentrations on the order of 109 cm-3 are produced by flash photolysis of O3 at 266 nm, with a subsequent reaction of O1D with water vapor. The 266 nm laser pulses (pulse energy 20 to 28 mJ, repetition rate 1 Hz, pulse duration less than 10 ns) are provided by a compact, frequency quadrupled Nd:YAG laser (Quantel model Ultra 100). The laser is mounted on one side of an optical rail, on which the flow tube is mounted on the opposite side. The laser beam is widened by an optical telescope to a diameter of 3 cm and guided to the flow tube by two turning mirrors.

Water vapor, temperature and pressure in the flow tube are continuously monitored. Normally, ozone and water vapor concentrations in the sampled ambient air are sufficiently high in order to produce high OH concentrations. However, ozone can be depleted during night due to its reaction with nitric oxide and by deposition processes. Therefore, a small flow of synthetic air (0.2 Lmin-1) that has passed an ozonizer (glass tube of fused silica with a mercury lamp providing 185 nm radiation) can be added in order to increase ozone mixing ratios in the flow tube by 40–50 ppbv. The injection is controlled by a solenoid valve which is automatically opened if the ozone mixing ratio in ambient air drops below 30 ppbv.

At a distance of 48 cm from the inlet of the flow tube, 1 Lmin-1 of the total flow is sampled from the center of the flow tube through a conical nozzle into the OH detection cell. The design of the OH fluorescence cell is the same as used for OH concentration measurements .

In the cell, OH is excited by 308 nm radiation from a tunable frequency-doubled dye laser, which is operated at a pulse repetition rate of 8.5 kHz. The OH fluorescence is detected by gated photon counting and accumulated in time bins of 0.6 ms. This way, the chemical decay of OH in the flow tube is recorded for 1 s after the photolysis laser pulse. For photon detection, a gated multichannel photomultiplier (Photek, PM325) is used in combination with a multichannel counting card (Sigma Space, AMCS).

In order to achieve sufficiently precise reactivity measurements, 60 decay curves are taken for one reactivity measurement resulting in an amplitude of 50 to 100 counts of the decay curve. Because of the scanning of the laser over the absorption line of OH in order to track slow drifts in the wavelength of laser, the amplitude of the decay curve changes periodically. Therefore, 10 OH decay curves are summed up to equalize the amplitude. Six of the summed curves are then averaged to determine realistic error estimates needed for the fit procedure. A weighted single-exponential fit (Levenberg–Marquardt minimization) is then applied to derive the OH reactivity (Eq. ). Approximately the first 30 to 50 ms of the decay curve are not included in the fit because these points deviate from the single-exponential behavior that is observed at later times. The fit is started if the count rate has decreased to the 90 % level of the maximum count rate. The likely reason for an inhomogeneous initial OH distribution is that the spatial OH distribution is not perfectly homogeneous near the inlet nozzle of the OH detection cell right after the laser pulse due to inhomogeneities in the laser power across the laser beam.

Diffusion to the wall of the flow tube, where OH is lost by wall reactions, causes loss of OH even in the absence of OH reactant. This zero loss rate is regularly measured in humidified air (purity 99.999 %). Typical zero loss rates measured in laboratory characterization measurements are around 3 s-1 for this instrument. A slightly higher value of 3.8 s-1 was derived in measurements sampling synthetic air from a gas cylinder during the campaign. Analysis of the synthetic air in this gas cylinder by gas chromatography yielded contaminations with an OH reactivity of 0.7 s-1. Therefore, an instrumental zero decay value of 3.1 s-1 was subtracted from ambient OH reactivity measurements consistent with previous values for this instrument. The reactivity measured in the synthetic air is considered a potential systematic error of the OH reactivity measurements in this campaign. The accuracy of our LP-LIF technique has been tested with CO and CH4 mixtures in synthetic air. Measured kOH agreed better than 10 % with the expected, calculated OH reactivity for values up to 60 s-1, in agreement with previous studies by . At higher kOH values, the initial non-exponential part of the OH decay curve starts to influence the quality of the fitted OH decay curve, but such high kOH values were not encountered in the campaign at Wangdu (Fig. ).

Potential interferences that could be present in the OH concentration detection would not affect the measured OH reactivity because OH that would be artificially produced inside the measurement cell would only increase the background signal but not the decay time as long as it do not change on the timescale of the OH decay measurement (1 s). In any case, however, effects are expected to be negligible due to the high OH concentration inside the flow tube that are much higher compared to ambient OH concentrations, for which interferences have been recognized. This holds for the known interference from ozone photolysis by the 308 nm laser radiation but also for other potential interferences that have been reported for OH concentration measurements and which could not be fully excluded for this campaign .

If ambient NO concentrations are high enough to lead to a significant regeneration of OH from secondarily formed HO2, the shape of the decay curve changes to a bi-exponential behavior. This can be derived from reaction kinetics. The faster decay time represents approximately the OH reactivity for certain chemical conditions. As shown in , no significant effects are expected for NO mixing ratios of up to 20 ppbv for realistic OH reactant mixtures in our instrument. During the campaign in Wangdu, NO mixing ratios were generally well below 20 ppbv. No bi-exponential behavior was observed that would have been seen in the residuum of the fit. NO mixing ratios exceeded 20 ppbv only for some short periods mainly during the nighttime on 3 days, but measurements still appeared as single-exponential decays in these cases.

Measured OH reactivity values ranged between 10 and 20 s-1 during this campaign for most of the time (Fig. ). In general, values were lower during the daytime (median value 12.4 s-1) than at night (median value 15.4 s-1). During the first 2 weeks, midday OH reactivity increased from 10 s-1 on 8 June to values higher than 20 s-1 between 15 and 19 June. After 19 June, OH reactivity was generally lower and more uniform till the end of the campaign.

Maximum values were observed during the nighttime and the early morning hours, when OH reactivities show spikes with values of up to 60 s-1 for short periods of less than 1 h. The high-reactivity values were probably caused by emissions into the shallow nocturnal boundary layer. The short duration indicates that nearby local sources were responsible for these events. This happened more frequently during the first part of the campaign and only few spikes were observed after 19 June.

The overall changes in OH reactivity values from day to day were likely dominated by anthropogenic activities during this campaign. The measured OH reactivities show an increasing trend with CO, which cannot be explained by the reactivity of CO alone (Fig. ). Therefore, other reactants that were co-emitted with CO, for example in combustion processes, most likely contributed to the increase in reactivity. The correlation still holds if only reactivity from measured OH reactants other than CO, NOx and isoprene is taken into account. This further supports that also OH reactivity from organic compounds is co-emitted with CO.

Back-trajectories were calculated for this campaign using the NOAA (Nation Oceanic and Atmospheric Administration) HYSPLIT (Hybrid Single Particle Lagrangian Integrated Trajectory Model) model in order to test if measured OH reactivities are correlated with the origin of advected air masses. Twenty-four hour back-trajectories were calculated for air masses at the measurement site for each hour. During most days, back-trajectories were very similar. Therefore, trajectories between 10:00 and 19:00 LT were averaged (Fig. ). The majority of back-trajectories point to locations south of Wangdu and less often to locations east or north of the measurement site. Mountains that are west and north of the measurements site appear as barriers for air masses. Only on 3 days (8, 27, 28 June) back-trajectories indicate that air masses originated from locations in the mountains. Lowest kOH values (<10 s-1) were observed in these cases due to less emissions from industry and from other anthropogenic activities. In contrast, there is dense population east and south of the measurements site. This likely explains why OH reactivity values were highest if air masses were coming from this area. Also, the relation between kOH and CO is consistent with the assumption that OH reactivity was dominated by anthropogenic activities in this case.

The increase in OH reactivity during the first 2 weeks could be related to a change in the origin of air masses from the north (8 June) to the east (13 June) and finally to the south (15 June). However, no obvious difference between back-trajectories is seen before and after 20 June so that back-trajectories are not sufficient to explain why measured OH reactivity would be higher and spikier before 20 June.

The more likely reason for differences in OH reactivity is emissions connected with the harvesting of crop and combustion of straw and crop residuals on nearby agricultural fields in the first 2 weeks of June. On 13 June, for example, crop was harvested on the field directly next to the measurement place. Indicators of biomass burning activities were visually observed fires, reduced visibility and an increase in measured particle number concentrations. Typical daytime maximum PM2.5 concentrations ranged between 30 and 90 µgm-3 but were as high as 300 µgm-3 on one day due to local biomass burning (Fig. ). No clear connection between OH reactivity and aerosol number concentration was observed. Although a sharp drop in PM2.5 was observed on 19 June when OH reactivity also dropped, PM2.5 increased again to higher values till the end of the campaign. Elevated concentrations of acetonitrile (a marker for biomass combustion) were measured between 12 and 19 June .

OH reactivity measurements are of particular value in order to test if all important OH reactants were detected. Volatile organic compounds (VOCs) and inorganic compounds such as nitrogen oxides (NOx= NO + NO2) and carbon monoxide (CO) are typically major contributors to the total OH reactivity. However, the number of OH reactants, specifically of organic compounds is very large so that a complete measurement is not expected. Therefore, comparison of direct kOH measurements with calculations from measured reactants can reveal to which extent unmeasured reactive compounds contributed to total OH reactivity. This presents a gap in the constraints of model calculations used to test our knowledge of radical chemistry . In addition, VOCs and NOx are key species for understanding ozone and particle formation so that an incomplete knowledge of OH reactivity would lead to a systematic underprediction of ozone production by chemical models (e.g., ).

The full time series of the calculated kOH is plotted together with the measured total kOH in Fig. . The calculated reactivities were determined from measured CO, CH4, C2 to C11 alkanes, C2 to C6 alkenes, C6 to C10 aromatics, formaldehyde, glyoxal, acetaldehyde, MVK, MACR, NO, NO2 and SO2 (Table ). Reaction rate constants were taken from IUPAC recommendations (Atkinson et al., 2004, 2006) or the structure–activity relationship (SAR) as stated in the Master Chemical Model (http://mcm.leeds.ac.uk/MCM/).

During each of the two parts of the campaign (before and after 19 June), diurnal profiles of observations appear to be similar. Therefore, measured kOH and calculated reactivity from major contributors are shown as median diurnal profiles with percentiles for each period in Fig. . Median diurnal profiles of all measured contributions are summed up and compared to measured kOH in Fig. . Ambient temperature was used for the calculation of reaction rate constants, but the differences between ambient temperature and the actual temperature in the instrument does not change any of the results shown here.

The most important OH reactants were CO (on average 20 to 25 % of the total OH reactivity), nitrogen oxides (on average 12 to 22 % of the total OH reactivity) and OVOCs (on average 25 % of the total OH reactivity). The reactivity from isoprene makes a substantial contribution (often 20 %) to the total kOH in the afternoon. Reactivity from alkanes and alkenes were dominated by small alkenes, mostly ethene and propene.

The median diurnal profile of the total OH reactivity had a maximum late at night. It decreased during the day by nearly 50 % and started to increase after sunset. The accumulation of OH reactants during the night could be due to fresh emissions that are released into the shallow nocturnal boundary layer. A similar diurnal profile was also observed for contributions from NOx, alkane and alkene species. Their concentrations are typically connected to emissions from anthropogenic activities. OH reactivity from NOx was also the largest contribution to kOH during night and early morning (20 to 30 %). The diurnal profile of NOx appears as the major driver for the diurnal profile of the entire kOH, whereas nearly all other contributions exhibited a less distinct diurnal profile. A different diurnal behavior to that of NOx was observed for isoprene, which is emitted by plants. The emission strength scales with light and temperature, and, therefore, maximum mixing ratios were reached in the afternoon. Isoprene also contributed to the reactivity in the early evening, most likely because isoprene that was emitted during the daytime was only partly oxidized by OH before sunset. The diurnal profile of isoprene partly counteracted the decrease in OH reactivity due to the decrease in NOx, alkane and alkene species.

CO mixing ratios ranged between 300 and 1000 ppbv during this campaign. Therefore, reactivity from CO always made up a large fraction of the total kOH. The OH reactivity from CO showed only a weak diurnal variation with a median value of 3 s-1 and could therefore be used as an indicator of the overall origin of pollutants apart from diurnal changes. As discussed above, measured kOH scaled with CO indicating that co-emitted OH reactants such as alkenes were also important (Fig. ).

A number of oxygenated volatile organic compounds (OVOCs) were measured in this campaign (Table ). These included formaldehyde, acetaldehyde, glyoxal, methyl-vinyl ketone and methacrolein. Their reactivity made a large fraction of the total reactivity with median values between 2 and 4 s-1 over the course of 1 day. The largest contributions to the reactivity from OVOCs (more than 50 %) came from formaldehyde and acetaldehyde (20 to 25 %), while reactivity from other measured OVOCs such as acetone and glyoxal made only small contributions. These species can also originate from primary emissions. The good agreement between measured and calculated OH reactivity nevertheless indicates that these species were the most important organic oxidation products that contributed to the OH reactivity.

The reactivity of measured OVOCs shows weak diurnal variation, with a decrease by a factor of about 2 from the morning to the evening. This behavior suggests that during the daytime, dilution due to a rising boundary layer height or chemical removal had a stronger influence on the observed OVOCs than fresh production by photochemistry.

Although the general behavior of OH reactivity and OH reactants was similar during the entire campaign, there were distinct differences in the magnitude of total OH reactivity during the first (7 to 19 June) and second half (20 June to 8 July) of the campaign (Fig. ). Measured OH reactivity was on average lower after 20 June specifically during the second half of the night and early morning, when median values were higher than 25 s-1 before 20 June and 16 to 20 s-1 later. Afternoon values were only slightly less after 20 June compared to the first part of the campaign. This is reflected in a decrease in median OH reactant concentrations during the second part of the campaign. It is most prominently seen in median alkene and alkane concentrations during the nighttime (Fig. ). In contrast, isoprene concentrations increased faster in the morning and high afternoon concentrations persisted in the evening during the second part of the campaign. Air temperatures were generally a few degrees higher than during the first 2 weeks so that temperature-driven biogenic emissions could have been larger after 20 June. The largest fraction of higher OH reactivity observed in the first part of the campaign remains unexplained by OH reactant measurements. However, even during times when measured reactivity was higher than calculations from OH reactants, the gap is within the combined 2σ uncertainties: the kOH calculated from OH reactants has a 1σ uncertainty of ±10 to ±15 %, depending on the relative distributions of reactants, and the measured kOH has a maximum 1σ uncertainty of ±10 % plus +0.7 s-1 (Table ).

The good agreement between measured and calculated OH reactivity is also demonstrated by the high linear correlation coefficient (R2=0.77 for the entire data set and both subsets of data) between both values (Fig. ). For the second part of the campaign, a linear regression analysis (forced to zero) yields a slope of 1.01. As already discussed, missing reactivity was higher during the first part of the campaign so that a regression analysis yields a higher slope of 1.3.

Largest differences of 5 to 6 s-1 (approximately 20 %) between measured and calculated OH reactivity occurred during the nighttime and early morning during the first 2 weeks of the campaign, when NO concentrations were also highest. This could indicate that unmeasured OH reactants were co-emitted with nitrogen oxides in combustion processes. Unknown compounds causing the missing reactivity are the main reason for the higher observed OH reactivity in the first 2 weeks. Therefore, there is no clear further indication of the nature of missing reactivity during this period. Emissions of organic compounds from biomass burning may have not been detected during the first part of the campaign. During the nighttime nearby sources for OH reactants as indicated by the short duration of high reactivity could also have contributed to the missing reactivity. In addition, undetected products from the oxidation by the nitrate radical could have been part of missing reactivity in the night.

Exceptionally good agreement is seen at nearly all times after 20 June in the time series as well as in the median diurnal profile (Figs.  and ). The median value of missing reactivity is only 0.3 s-1. Such good agreement is not expected due to the large number of possible OH reactants in the atmosphere . Specifically the number of OVOCs that were measured in this campaign is rather small (see above) and additional reactivity from other oxidation products could be expected to contribute to the total OH reactivity.

The good agreement between measured and calculated kOH indicates that other oxidation products than those measured were not significantly contributing to the OH reactivity at the measurement site. Therefore, concentrations of oxygenated organic compounds that are produced by model calculations but that were not detected were constrained to zero in calculations presented in our accompanying paper by in order to ensure that modeled OH reactivity is consistent with measurements. One explanation could be that the photochemical age of air masses was short and, therefore, oxidation products could not accumulate. This could be the case for fresh emissions close to the measurement site. In addition, unmeasured oxidation products may still have contributed to the OH reactivity within the combined uncertainties of OH reactivity measurements and calculations from OH reactant measurements.

In our previous field campaigns in China in 2006 in the Pearl River delta (PRD; ) and Yufa close to Beijing , OH reactivity was considerably higher, but exhibited a similar diurnal profile. Maximum values were 40 to 50 s-1 in the night and early morning during the PRD and Yufa campaigns and reached minimum values of around 20 s-1 in the afternoon. Absolute contributions from CO and NOx were comparable with contributions in Wangdu 2014, with slightly higher CO concentrations in Yufa 2006. However, contributions from measured VOC were significantly higher in both previous campaigns compared to the Wangdu campaign in 2014, partly explaining the higher reactivity in these campaigns.

In both previous campaigns, measurements of OVOCs were completely missing and the measured OH reactivity was found to be about 2 times larger than the total reactivity of measured CO, NOx and hydrocarbons . The missing reactivity could be quantitatively explained by OVOCs which were simulated by a model from the photooxidation of the measured VOCs. The major modeled OVOCs were formaldehyde, acetaldehyde, MVK, MACR and some minor isoprene oxidation products, which together could explain 70 % of the missing reactivity (i.e., about one-third of the total reactivity). In the Wangdu campaign, the calculated total OH reactivity was largely in agreement with the measured kOH. This time, formaldehyde, acetaldehyde, MVK, MACR and glyoxal were directly measured and also accounted for one-third of the total reactivity. These species were also the most important OVOC species in other campaigns in anthropogenically dominated environments such as in Beijing , London and Tokyo . This confirms the high relevance of these specific carbonyl compounds as reactants for OH in the polluted boundary layer.

The OH reactivities measured at the Wangdu site in the North China Plain show diurnal profiles that are comparable to those reported for other polluted environments all over the world (see review by ). The total reactivities lie within the range of values observed during summertime at other locations that were mainly influenced by anthropogenic emissions, like Nashville , New York and Houston in the US, Tokyo in Japan , Beijing in China, Seoul in South Korea , and London in Great Britain. Also, the shapes of the diurnal profiles were similar, with peak values between 15 and 50 s-1 in the early morning and minimum values in the afternoon. Significantly higher morning values of 130 s-1 were observed in Mexico City in 2003 . Here, as well as in Wangdu and other urban sites, the diurnal shape of kOH was strongly determined by the variation in anthropogenically emitted NOx and co-emitted VOCs.

Care has to be taken if missing reactivity is compared between different campaigns because the number of measured OH reactants used to calculate the reactivity can significantly differ and references therein. For the measurements in Beijing , approximately 25 % of the measured reactivity remained unexplained, although oxygenated organic species were partly measured. Approximately 30 % of the reactivity measured in Nashville could not be explained, even if modeled organic compounds were taken into account. For the other campaigns in anthropogenically influenced areas, measured OH reactivity could be explained by either measured OH reactants alone (New York, this campaign) or if, in addition, product species from model calculations were included (Yufa, PRD, Tokyo, London).

OH reactivity measurements can be used not only to quantify the possible contribution of unmeasured OH reactants, but they also allow the quantification of the total OH production rate. Because OH is short-lived, it reaches a steady state within seconds. Thus, the total OH production rate (POH) equals the total destruction rate (DOH). DOH can be calculated as the product of kOH and the OH concentration: DOH=kOH×[OH]. This rate can be compared with the sum of production rates (POH) from known OH sources. In this campaign, OH production from HONO and O3 photolysis, ozonolysis of alkenes, and radical recycling reactions of HO2 with NO and ozone can be calculated from measurements: POH=POH(hν+O3)+POH(hν+HONO)+POH(HO2+O3)+POH(HO2+NO)+POH(O3+alkene). Potentially unknown OH sources can then be determined as the difference between DOH and POH. This was successfully applied for data from our previous field campaigns in China , revealing significant unaccounted OH sources, and in chamber studies .

The time series of calculated OH production and destruction rates are plotted in Fig. , and median diurnal profiles of quantities that are required for this calculation are shown in Fig. . Unfortunately, the data coverage of simultaneous measurements before 20 June (mostly due to missing radical measurements) is not sufficient to allow for an independent analysis of the first part of the campaign as done for the analysis of OH reactants. However, results do not change significantly, whether the first part is included in the median diurnal profiles that are discussed below or not.

Figure  shows the median diurnal profile of the OH destruction and production rates and their difference, including an estimate of the accuracy of the calculated difference. The diurnal profile of the OH production rate was mainly driven by solar radiation as expected from the photolytic nature of primary radical production, which also determines the overall abundance of HO2. During the daytime, the known OH production was dominated by the recycling reaction of HO2 with NO, reaching a maximum of about 10 ppbvh-1 shortly before noon. The relative contribution of primary OH production by either O3 or HONO photolysis to the total OH production increased during the day to reach median maximum values of 1.2 and 1.5 ppbh-1, respectively. The ozone photolysis exhibited a strong diurnal profile because both solar radiation and ozone concentration had maximum values at noon and in the early afternoon. An OH production rate from HONO photolysis of 1 to 1.5 ppbh-1 persisted into the afternoon due to relatively high HONO concentrations measured throughout the day. The budget of HONO will be discussed in a separate paper, but it is clear that HONO production from the reaction of OH with NO cannot explain the high HONO concentrations in the afternoon. Ozonolysis of alkene species made only a minor contribution to the OH production at all times. Only C2 to C6 alkene species were measured so that ozonolysis reactions of undetected alkene species (potentially monoterpenes) could have additionally contributed to the OH production. However, the good agreement between measured and calculated OH reactivity does not indicate that a large fraction of alkene species were missed.

The time series of the total OH production and destruction rates, determined by Eqs. () and (), respectively, were nearly balanced for most of the time (Fig. ). The OH destruction rate is on average only 20 % higher than the sum of OH production during the daytime. Although the difference is hardly significant with respect to the experimental accuracies (Fig. ), a systematic trend of the ratio between OH production and destruction rates with NO can be seen (Fig. ), which points to a missing OH source at low NO concentrations.

For NO mixing ratios of less than 0.3 ppbv, OH destruction was nearly twice as large as the OH production, whereas production and destruction was balanced for NO mixing ratios higher than 1 ppbv. The result of the budget analysis is consistent with the finding by that model calculations underpredict OH by up to a factor of 2 at NO mixing ratios of less than 0.3 ppbv but describe HO2 and kOH correctly under these conditions at the Wangdu site. The good description of HO2 and kOH means that the major known OH source (the reaction of HO2 and NO) and the total OH loss rate are represented well by the model. Further model tests suggest a missing process that recycles OH from RO2 and HO2 by an unknown agent that behaves like 0.1 ppbv NO . Other trace gases measured at Wangdu give no indication as to the nature of the missing source in the OH budget analysis or in the model results. A similar behavior was found in our previous field campaigns in China in 2006. However, the ratio of POH/DOH was much smaller, with a value of about 0.25 for NO mixing ratios of 0.1 to 0.2 ppbv NO in PRD . In this case, the missing OH source was highly significant with respect to the experimental uncertainties of the calculated reaction rates, whereas in Wangdu, the much weaker imbalance of the OH budget can be almost explained by the experimental errors.

In addition to the measurement uncertainties stated in Table , instrumental tests during this campaign cannot exclude that OH concentration measurements are partly affected by an artifact, as discussed in detail in . The upper limit for an instrumental interference was estimated to be equivalent to an OH concentration of 1×106 cm-3. This positive bias would also give a positive bias in the calculated OH destruction rate.

In the night, OH production from sources taken into account in this calculation is close to zero because there is no radiation. This suppresses both OH production from photolysis reactions and OH regeneration by the reaction of peroxy radicals with NO. Because of the relatively high OH reactivity OH concentrations are expected to be very small. However, median measured OH concentrations ranged between 0.5 and 1×106 cm-3 (Fig. ). A median OH production of 1 to 3 ppbvh-1 would be required to explain measured nighttime OH concentrations (Fig. ).

Potential reasons for additional OH production at night have been recently discussed by , such as OH production by ozonolysis of terpenoids or dissociation of radical reservoir species like PAN that may be transported downward in the nocturnal boundary layer. Such mechanisms may have played a role at Wangdu, but we have no suitable measured data to test these hypotheses. In order to balance the calculated OH destruction rate during the nighttime, a rather large concentration of an alkene would be required. Assuming an ozone concentration of 30 ppbv, a reaction rate constant for the ozonolysis reaction of 1.8×10-15 cm3s-1 for δ-terpene and an OH yield of 1 , the concentration would need to be around 600 pptv.

However, the impact of a potential interference in the OH concentration measurements would also be largest in the night (Fig. ) because nearly the entire OH signal could be due to interferences. As a consequence, the difference between calculated OH production and destruction during the nighttime is within this additional uncertainty. The calculated OH destruction rate is less affected during the daytime, when a potential OH interference of less than 1×106 cm-3 would only be a small fraction of the total measured OH .

In our previous field campaigns in China in 2006, the OH destruction and production rates were significantly higher than in this campaign. In PRD and Yufa, maximum mean turnover rates (OH destruction rates) of 40 and 20 ppbvh-1, respectively, were reached around noontime . These values are 1.5 to 3 times higher than median OH turnover rates in this campaign. As discussed above, the major difference is that measured OH reactivities were significantly higher in the previous campaigns. The resulting higher loss rate was only partly balanced by a higher OH production from the reaction of HO2 with NO, which was nearly a factor of 2 larger in PRD and Yufa. Therefore, the gap between calculated OH destruction and production was also clearly above the level of significance with respect to the measurement uncertainties .

Also, the distribution of primary OH sources is different in this campaign compared to our previous campaigns in China, when HONO photolysis exhibited a diurnal profile with maximum values in the morning. These values were larger compared to this campaign, but HONO mixing ratios dropped to lower values in the afternoon so that production by HONO photolysis was less in Yufa and PRD than in Wangdu in 2016. Nevertheless, total primary OH production was higher (by a factor of 2 in PRD and a factor of 1.5 in Yufa) in the previous campaigns.

HONO photolysis was also the most important primary source for OH radicals in other campaigns that were conducted in anthropogenically dominated environments for example in New York , Paris , Mexico City , Santiago and Tokyo . These campaigns took place in or very close to very large cities (the one in Paris during wintertime) and NO concentrations were often exceptionally high so that HONO formation was favored. Our measurement site in Wangdu was not directly located in an urban area, and therefore the NOx concentrations were only moderately high in the morning and rather small in the afternoon so that the importance of HONO as the largest primary source for OH was not necessarily expected. The contribution of alkene ozonolysis to the OH production in other campaigns in urban environments was partly significantly higher compared to the Wangdu site due to higher alkene concentrations.