The influence of clouds on radical concentrations : 1 Observations and modelling studies of HO x during the Hill 2 Cap Cloud Thüringia ( HCCT ) campaign in 2010 3

15 The potential for chemistry occurring in cloud droplets to impact atmospheric composition 16 has been known for some time. However, the lack of direct observations and uncertainty in 17 the magnitude of these reactions, led to this area being overlooked in most chemistry 18 transport models. Here we present observations from Mt. Schmücke, Germany, of the HO2 19 radical made alongside a suite of cloud measurements. HO2 concentrations were depleted in20 cloud by up to 90% with the rate of heterogeneous loss of HO2 to clouds necessary to bring 21 model and measurements into agreement demonstrating a dependence on droplet surface area 22 and pH. This provides the first observationally derived assessment for the uptake coefficient 23 of HO2 to cloud droplets and was found to be in good agreement with theoretically derived 24 parameterisations. Global model simulations, including this cloud uptake, showed impacts on 25 the oxidizing capacity of the troposphere that depended critically on whether the HO2 uptake 26 leads to production of H2O2 or H2O. 27


Introduction
Clouds occupy around 15% of the volume of the lower troposphere and can impact atmospheric composition through changes in transport, photolysis, wet deposition and incloud oxidation of sulphur.Modelling studies have shown that aqueous phase chemistry can also significantly reduce gaseous HO 2 concentrations by heterogeneous uptake and loss into cloud droplets (Jacob, 1996;Tilgner et al., 2005;Huijnen et al., 2014).This chemistry is predicted to reduce OH and O 3 concentrations also due to the reduction in the gas-phase concentration of HO 2 .This in turn, decreases the self-cleansing capacity of the atmosphere and increases the lifetime of many trace gases (Lelieveld and Crutzen, 1990) with impacts for climate and air quality.Aqueous phase models have been developed which combine multiphase chemistry with detailed microphysics (Tilgner et al., 2005), but there are limited experimental field data of gas-phase radical concentrations within clouds to corroborate model predictions of heterogeneous loss of radicals to cloud droplets.There have been a number of aircraft campaigns which have measured OH and HO 2 radical concentrations within clouds (Mauldin et al., 1997;Mauldin et al., 1998;Olson et al., 2004;Commane et al., 2010), often, however, simultaneous observations of cloud droplet number and size distributions (or other key gas-phase radical precursors) were not made during these studies, making it difficult to assess the full impact of clouds on radical concentrations.In general therefore climate and air quality models do not consider this impact of clouds on atmospheric composition.
Within the literature, a wide range of uptake coefficients of HO 2 to liquid and aerosol surfaces have been considered to reproduce observed HO 2 concentrations (e.g.(Sommariva et al., 2004;Haggerstone et al., 2005;Emmerson et al., 2007;Whalley et al., 2010)) with often large uptake coefficients (up to 1 at times) used to reconcile model over-predictions.A wide range of uptake coefficients, not wholly consistent with each other, have been reported from laboratory studies (Abbatt et al., 2012).From measurements conducted in our laboratory, uptake probabilities of HO 2 to sub-micron aerosols were found to be less than 0.02 at room temperature (George et al., 2013) for aqueous aerosols that did not contain significant transition metal ions; similarly low uptake coefficients were derived by Thornton and Abbatt (2005).In contrast, measurements by Taketani et al. (2008) suggest higher uptakes of ~0.1 with enhancements observed with increasing relative humidity.
The uptake of HO 2 to aqueous aerosols is driven by its high solubility in water owing to its high Henry's Law constant ( = 4.0×10 3 M atm -1 at 298.15 K (Hanson et al., 1992)).
Once in the aqueous phase, reaction between dissolved HO 2 and its conjugate base, O 2 -, occurs rapidly.Thornton et al. (2008) have demonstrated that the solubility and reactivity of HO 2 is temperature and pH dependent and if the well characterised aqueous phase reactions (Sect.2.3, (R1) -(R5)) alone are representative of the heterogeneous loss processes, only small uptake coefficients would be expected at room temperature, consistent with the work by George et al. (2013) and Thornton and Abbatt (2005).The enhanced uptake coefficients reported by Taketani et al. (2008) suggests that there may be additional competing mechanisms occurring, however.
Further uncertainties arise in the literature relating to the eventual gas-phase products from these aqueous-phase reactions.The general consensus, until recently, was that these reactions would ultimately produce H 2 O 2 (Jacob, 1996), but the significance of the reactions depends critically on whether this is the case or whether, instead, H 2 O is produced (Macintyre and Evans, 2011).This is significant as H 2 O 2 can photolyse to return odd hydrogen (HO x =OH+HO 2 ) to the gas phase, whilst cloud uptake of HO 2 to form H 2 O provides a terminal sink for HO x .Recent work by Mao et al. (2013) postulates that a catalytic mechanism involving the coupling of the transition metal ions Cu(I)/Cu(II) and Fe(II)/Fe(III) may rapidly convert HO 2 to H 2 O, rather than H 2 O 2 in aqueous aerosols.The concentration and availability of dissolved Fe and Cu in cloud droplets tends to be much lower than in aqueous aerosol (Jacob, 2000) with a large fraction of Cu ions present as organic complexes (Spokes et al., 1996;Nimmo and Fones, 1997) which are far less reactive towards O 2 -and HO 2 (aq) than the free ions (Jacob, 2000) and so it is uncertain whether the mechanism put forward by Mao et al. (2013) could be extended to heterogeneous processes occurring within cloud droplets.
To better understand the role of clouds and heterogeneous processes on the oxidative capacity of the troposphere, coordinated gas-phase measurements of OH and HO 2 within clouds together with aerosol-cloud microphysical measurements are needed.The Hill Cap Cloud Thüringia 2010 (HCCT-2010) campaign which took place in 2010 aimed to characterise the interaction of particulate matter and trace gases in orographic clouds.This paper presents the impact of cloud droplets on measured gas-phase OH and HO 2 and uses these observations to 2 HO H assess the proposed aqueous phase mechanisms and determine the global impact of clouds on the tropospheric oxidising capacity.

Experimental
The HCCT-2010 campaign took place at the Thüringer Wald mountain range in central Germany during September and October 2010.The radical measurements were made from the German Weather Service (DWD) and the Federal Environmental Office (UBA) research station located close to the summit of Mt.Schmücke (the highest peak in the mountain range, 937 m above sea level, 10 46'8.5"East, 50 39'16.5"North).In October, the UBA station is immersed in cloud for 25 days on average (Herrmann et al., 2005) and, hence, is highly suitable for the study of gas and aerosol interactions with orographic cloud.Two additional experimental sites, approximately 4 km upwind of the summit site at Goldlauter and approximately 3 km downwind of the summit at Gelhberg were also equipped with a number of instruments which enabled the processing of a single air parcel as it passed through a cloud to be assessed by multiphase trajectory models such as SPACCIM (SPectral Aerosol Cloud Chemistry Interaction Model (Wolke et al., 2005); see Sect.2.3).Further details of the locations may be found in Herrmann et al. (2005).

Radical measurements
OH and HO 2 measurements were made using the fluorescence assay by gas expansion technique (FAGE).Details of the instrumentation can be found in Whalley et al. (2010).A single FAGE fluorescence cell was used for sequential measurements of OH and HO 2 .This was operated from the top of a 22 m high tower to co-locate with cloud measurements and ensure that the measurements were performed in full cloud.The cell was held at 1 Torr using a roots blower backed rotary pump system which was housed in an air-conditioned shipping container at the base of the tower (Fig. 1) and was connected to the cell via 30 m of flexible hosing (5 cm OD).308 nm tuneable, pulsed laser light was used to electronically excite OH radicals, this was delivered to the cell via a 30 m fibre optic cable (Oz optics) with the laser system (a Nd:YAG pumped Ti:Sapphire, Photonic Industries) housed in the shipping container.Fluorescence was detected by a channel photo multiplier (CPM) (Perkin Elmer) and gated photon counting.Data were acquired every second (photon counts from 5000 laser shots), with a data acquisition cycle consisting of 220 seconds with the laser wavelength tuned to the OH transition (NO was injected after 110 sec to rapidly convert HO 2 to OH, to allow the quantification of HO 2 ) and 110 sec tuned away from the OH transition to determine the background signal from laser scattered light.2010) for further details); this was carried out before and after the campaign and the values agreed within 21%; the average flux was used to determine the sensitivity.The limit of detection (LOD) at a signal to noise ratio of one for one data acquisition cycle was ~6×10 5 molecule cm -3 and ~8.5×10 5 molecule cm -3 for OH and HO 2 , respectively.
A number of operational modifications (from the standard University of Leeds ground-based operations (Whalley et al., 2010)) were necessary to facilitate measurements of the gas-phase concentrations of the radicals within clouds.As tower measurements were required (schematic of the measurement set-up is provided in Fig. 1), a single, smaller (4.5 cm (ID) diameter stainless steel cylinder) FAGE fluorescence cell, based on the University of Leeds aircraft cell design (Commane et al., 2010) was used for sequential measurements of OH and HO 2 .Ambient air was drawn into the cell through a 1 mm diameter pinhole nozzle.The distance between sampling nozzle and radical detection region was 18 cm and NO (10 SCCM, BOC, 99.5%) was injected ~8 cm below the nozzle for titration of HO 2 to OH.
The fluorescence cell was orientated with the nozzle pointing horizontal to the ground in an attempt to minimise water pooling on the nozzle and being sucked into the cell during cloud events.Occasional droplets were ingested by the cell and resulted in an instantaneous large increase in the laser scattered signal.These spiked increases were discreet and short-lived; the data presented here have been filtered to remove these spikes, which were easy to identify.
Tests have been conducted post-campaign to determine the level of HO 2 interference from RO 2 radicals (Fuchs et al., 2011).Under this particular experimental set-up, an equivalent amount of ethene-derived RO 2 radicals to HO 2 were found to contribute 46 % to the total HO 2 signal (Whalley et al., 2013).The FAGE instrument was found not to be sensitive to CH 3 O 2 , and other short-chain alkane-derived RO 2 radicals but is sensitive to other alkene and aromatic derived RO 2 radicals with similar sensitivities to that for ethene-derived RO 2 .The instrument is also sensitive to longer-chain alkane-derived RO 2 radicals (>C 3 ) albeit to a smaller extent, as reported by Whalley et al. (2013).For this rural environment, at this time of year, however, the contribution of alkene and aromatic-derived RO 2 radicals to the total RO 2 budget is expected to be small as the parent VOCs for these particular RO 2 types were at low concentrations; isoprene concentrations, for example, were on average just 12.6 pptv.As a consequence of this, the resultant HO 2 interference from RO 2 radicals should also be low.

Model expression and constraints
An analytical expression has been used to predict the mean diurnal HO 2 concentrations throughout the campaign both during cloud events and outside of cloud events.This expression was originally developed by Carslaw et al. (1999) for modelling OH, HO 2 and RO 2 radicals in the marine boundary layer and was found to agree with full Master Chemical Mechanism (MCM) model predictions for OH and HO 2 to within 20% for daytime hours.It has since been extended further by Smith et al. (2006) to include additional HO 2 sinks, such as heterogeneous loss (k Loss ).The expression, given in Eq. ( 3), derives from the solution of simultaneous steady state expressions for OH and CH 3 O 2 (Eq.( 1) and Eq. ( 2) below) and includes any primary sources of HO 2 not coming from radical propagation steps such as formaldehyde photolysis: (1) (2) j(O 1 D) was measured from the top of the 22 m tower, alongside the FAGE detection cell, using a 2- filter radiometer (Bohn et al., 2008) which pointed skywards throughout the campaign.The photolysis rates of formaldehyde, j(HCHO), have been calculated using the Tropospheric Ultraviolet and Visible (TUV) radiation model (Madronich and Flocke, 1998).
The correlation between TUV calculated j(HCHO) with TUV calculated j(O 1 D) was determined allowing these photolysis rates to be scaled to the measured j(O 1 D) values to account for the presence of clouds.During cloud events, upward radiation will increase, with the magnitude of this increase dependent on the cloud optical depth (COD) and measurement height (Bohn, 2014).The contribution of upward radiation as a function of COD has been estimated using the TUV model using the methodology outlined by Bohn (2014).This estimated increase in upward radiation has been added to the in-cloud photolysis rates presented in Section 3. On average, photolysis rates are enhanced by ~17% during cloud events due to upwelling.A constant value of 1760 ppbv was assumed for CH 4 and a value of 508 ppbv was taken for H 2 .O 3 and NO x measurements were made from the top of the tower using commercial analysers which ran continuously from the 16 th September (day 3 of the field project).Details of the ancillary measurements used for comparison and model constraints are provided in Table 1.Further details of many of the measurement techniques can be found in the overview paper from an earlier hill cap cloud experiment, the Field

Investigations of Budgets and Conversions of Particle Phase Organics in Tropospheric Cloud
Processes (FEBUKA) project (Herrmann et al., 2005).
Rate coefficients are taken from the most recent recommendations in the Master Chemical Mechanism (MCMv3.2),http://mcm.leeds.ac.uk/MCM/.
A constant uptake rate for HO 2 (k Loss ) of 0.14 s -1 to cloud droplets was included during cloud events to reproduce the average HO 2 in-cloud observations.Additional model runs with no uptake during cloud events have also been run for comparison, as have model runs in which the first order loss to droplets was varied to replicate the HO 2 observations as a function of i) cloud droplet surface area and ii) pH (Sect.3.1).The equations used to calculate the theoretical increase in with increasing pH, as proposed by Thornton et al. (2008), which have been compared with determined in this work (Sect.3.1), are given by:

Trajectory model
In addition to the modelling exercises, outlined in Sect.

Global chemistry transport model
The GEOS-Chem model version 9.1.3(www.geoschem.org)has been run to assess the global impact of the uptake of HO 2 by cloud droplets.The model was run at 2x2.5 degree global resolution for two years.The first year was considered a spin-up and has been ignored.The standard model includes uptake of HO 2 onto aerosols (with an uptake coefficient of 0.2), but, the model has been updated in this work to include an uptake of HO 2 onto clouds.This is parameterized as a first order loss onto clouds in a similar way to that onto aerosols following Schwartz (1984) using the temperature dependent parameterization of Thornton et al. (2008) with a cloud pH of 5.The cloud surface area is derived from the cloud liquid water in the each model grid box (provided from the meteorological analyses) and cloud droplet radius is taken to be 6µm over continents and 10µm over oceans.Clouds below 258 K are assumed to be ice and no uptake occurs.The parameterization takes diffusional limitation in the gas phase into account but not in the cloud phase.All simulations use the same cloud liquid water fields, thus the impact of clouds on photolysis, wet deposition and transport is identical in all simulations.

Results and Discussion
Near continuous OH and HO 2 measurements were made at the Mt.Schmücke site from 13 th September to 19 th October 2010, during which 35 separate orographic cloud events were encountered which lasted as little as 24 min to more than 2 days in duration.Fig. 2 shows the time-series of OH, HO 2 , j(O 1 D), NO, O 3 and liquid water content.OH concentrations were close to or below the limit of detection (LOD) of the instrument for much of the measurement period.A clear diurnal signal was only observable when several days of data were averaged together outside of cloud events (Fig. 3).The peak OH concentration was observed at midday at ~1 x 10 6 molecule cm -3 .No clear OH diurnal profile was observed during cloud events.
HO 2 concentrations were variable depending on whether the site was in cloud or not.The average diurnal peak concentration of HO 2 was ~4 x 10 7 molecule cm -3 outside of cloud events (Fig. 3).A diurnal profile of HO 2 was also observed when sampling within clouds with peak concentrations reduced by approximately 90% on average.The measured rate of ozone photolysis, j(O 1 D), varied with time of day and cloud thickness.Daily peak photolysis rates were 8.8 x 10 -6 s -1 and 4.1 x 10 -6 s -1 outside and within clouds, respectively.Clouds thus reduced photolysis rates by ~60%.
Fig. 4 shows the dependence of measured HO 2 concentration on cloud droplet surface area for all daytime cloud events.The observed HO 2 concentration has been divided by the observed j(O 1 D) to remove the impact of the changing photolysis rates within the cloud.This ratio has then been normalized to 1 when the droplet surface area was zero and plotted against the cloud droplet surface area.The decrease in the ratio with increasing droplet surface area suggests that in addition to the reduction in HO 2 caused by a reduction in the photolysis rates within clouds, there is a further loss process of HO 2 that increases with cloud droplet surface area.A similar decrease in the ratio is also observed with increasing liquid water (not shown).From these observations it becomes apparent that a heterogeneous process must be occurring in the presence of clouds.
An insight into the mechanism by which HO 2 is lost to clouds is demonstrated by the dependence of the measured HO 2 concentration as a function of cloud water pH (Fig. 5a).
Throughout the project the pH of the cloud water was recorded every hour and ranged from 3.4 to 5.3.The lowest in-cloud HO 2 occurred in clouds with the highest cloud water pH suggesting that the solubility of HO 2 was enhanced at higher pH as might be expected given that HO 2 is a weak acid.

Determining the uptake coefficient for HO 2 to cloud droplets
The analytical expression derived by Carslaw et al.(1999), and given in Eq. ( 3), has been used to estimate HO 2 concentrations both in and out of cloud events (Fig. 6).The expression represents reasonably well the campaign mean diurnal observation of HO 2 outside of the cloud events during the daytime (red dashed line and shading).During cloud events, however, the model (black dashed line and shading) over-estimates the observed (grey line) HO 2 throughout the day.The inclusion of a first order loss process (k Loss =0.14 s -1 ) in the analytical expression is able to bring the observations and calculation into better agreement on average.The cloud droplet surface area was variable during the different cloud events encountered (1.2±0.4×10 3 cm 2 m -3 ) although no diurnal trend in this parameter was evident.
A clear anti-correlation between the observed HO 2 concentration and droplet surface area was observed and this correlation could only be reproduced by the analytical expression by increasing k Loss in the model from 2.0×10 -2 s -1 to 3.5×10 -1 s -1 as the surface area increased from 1.2×10 2 cm 2 m -3 to 1.5×10 3 cm 2 m -3 (Fig. 7).
This first order loss rate can be converted into an uptake coefficient ( ) using Eq. ( 7) (Schwartz, 1984).Using campaign mean values for cloud surface area (A) of 1.2×10 3 cm 2 m - 3 , droplet radius (r p ) of 6 µm, gas phase diffusion constant for HO 2 (D g ) of 0.25 cm 2 s -1 , and molecular speed of HO 2 (ω) of 64000 cm s -1 gives an uptake coefficient of 0.01; the uptake coefficient as a function of cloud droplet surface area is presented in the upper panel of figure 7. (7) These derived uptake coefficients are in good agreement with laboratory studies (Abbatt et al., 2012), including recent measurements in our laboratory, which ranged between 0.003 -0.02, for heterogeneous loss of HO 2 on aqueous (NH 4 ) 2 SO 4 , NaCl and NH 4 NO 3 sub-micron aerosols (George et al., 2013).This methodology provides, for the first time, a direct field assessment of the heterogeneous rate of loss of HO 2 .
Repeating this analysis but splitting the observations by cloud pH leads to values of ranging from 1.65x10 -3 at a pH of 3.7 to 8.84x10 -2 at a pH of 5.2 (Fig. 5b).These values are in good agreement with those calculated by Thornton et al. (2008) suggesting that the Thornton mechanism (which is based entirely on the known aqueous phase chemistry) is in play in real clouds and that it can be used to estimate the heterogeneous loss of HO 2 to cloud surfaces in the troposphere.
SPACCIM simulations (Wolke et al., 2005) have also been carried out, focussing on one particular cloud event which fulfilled the required meteorological and connected flow conditions for the cloud passage experiment (additional simulations relating to the other cloud events encountered during HCCT will be presented in future publications).The modelled and measured HO 2 concentrations at Mt. Schmücke during the cloud event, FCE1.1, are presented in Fig. 8. Comparisons between modelled and measured concentrations demonstrate the simulated HO 2 concentrations are in a similar range as the measurements.The mean simulated HO 2 concentrations of 3.1×10 6 molecule cm -3 for FCE1.1 are a factor of 1.4 greater than the HO 2 measurements which were, on average 2.2×10 6 molecule cm -3 during this particular cloud event.A further trajectory model simulation has been run and compared to measured HO 2 concentrations at Mt. Schmücke during a non-cloud event, NCE0.8, also.Fig. 9 reveals that the model is able to reproduce the modelled HO 2 concentrations well and tracks the temporal concentration profile throughout this event.The mean predicted HO 2 concentration is just 24% smaller than the measurements.
The agreement between the trajectory modelled and measured in-cloud HO 2 values confirms the significant reductions of radicals within clouds predicted by complex multiphase box models in the past (Lelieveld and Crutzen, 1990;Tilgner et al., 2005;Tilgner et al., 2013) and supports the findings presented above.Importantly, the results imply that the phase transfer data for HO 2 used within SPACCIM simulations, e.g. the applied mass accommodation coefficient ( = 10 -2 ), are appropriate to reproduce the reduced HO 2 concentrations for incloud conditions.These applied parameters control the uptake fluxes towards the aqueous phase and, ultimately, the aqueous phase HO X levels.Confidence in the values assumed for these parameters is essential to model in-cloud oxidation within the aqueous phase accurately, with the multiphase chemistry of other important chemical subsystems, such as the S(IV) to S(VI) conversion, the redox-cycling of transition metal ions and the processing of organic compounds all heavily dependent upon the values taken.

Global impact of the uptake of HO 2 onto cloud droplets
The GEOS-Chem Chemistry Transport Model (www.geos-chem.org) has been used to assess the impact of the uptake of HO 2 onto cloud droplets on the global oxidizing capacity using the, now field-validated, mechanism of Thornton et al. (2008).To investigate both the impact of the uptake and whether H 2 O 2 is produced three simulations are run, i) with no cloud uptake of HO 2 , ii) with cloud uptake (assumed pH of 5) of HO 2 using the Thornton mechanism to produce H 2 O 2 , and iii) with cloud uptake (assumed pH of 5) of HO 2 to produce H 2 O.All simulations include HO 2 uptake onto aerosol with of 0.2, which is the standard value used in GEOS-Chem (Martin et al., 2003;Macintyre and Evans, 2011).
Fig. 10  which are not producing O 3 .The change in the lifetime due to the HO 2 uptake onto clouds thus has little impact on O 3 production.The large surface impact of the cloud uptake primarily reflects uptake of HO 2 by clouds at the surface (see figure 12a) rather than a transported impact of cloud processes from aloft downwards.The small impact on O 3 is consistent with results of Liang and Jacob, (1997).These simulations make a variety of approximations as outlined in Sect.2.5.Given the complexity of representing cloud processes occurring over the length scale of meters to hundreds of meters in a comparatively low resolution global model (hundreds of kilometers) there are significant uncertainties as to the magnitude of these impacts.Further work in higher resolution cloud resolving models will be needed to estimate the full impact of these processes.Nevertheless, our observations show that the uptake of HO 2 onto clouds offers a substantial perturbation to the oxidising capacity on the local scale and that this perturbation may propagate into the regional and global scales.

Conclusions
We have shown here experimentally for the first time that the uptake of HO 2 onto clouds can have a significant impact on the composition of the atmosphere in a way consistent with theoretical predictions.It seems likely, however, that chemistry occurring within clouds will have other currently unknown impacts on the composition of the atmosphere.Global and regional models need to be developed further to investigate these impacts with predictive pH an especially important development.The impact of these processes may also change in the future with climate induced impacts on the hydrological cycle.Further laboratory, field studies and modelling are required to help resolve these remaining complex questions.     .This ratio has then been normalized to give a value of 1 when the droplet surface area was zero.The systematic decrease in this normalised ratio with increasing droplet surface area suggests that in addition to the reduction in HO 2 caused by a reduction in the photolysis rates within clouds, there is a further loss process that increases with cloud droplet surface area.The ratio decreases linearly with increasing droplet surface area up to 1500 cm 2 m -3 with the line of best fit being Ratio = 1 -5×10 -4 ×SA.
the fraction of O( 1 D) that reacts with H 2 O vapour to form OH, rather than being quenched to O( 3 P))Limited CO concentration data are available from the summit site during the project, owing to instrumental problems for the first two weeks of measurements.An average CO concentration of 231 ppbv was used in the analytical expression to determine HO 2 concentrations although additional model runs at + and -1σ of this average concentration (297 ppbv and 165 ppbv respectively) were also made to assess the sensitivity of the predicted HO 2 concentration to this constraint.Similarly, only discrete (non-continuous) measurements of HCHO were made during the project; an average value of 479 pptv was used as a model constraint and further model runs at + and -1σ of this average concentration (818 pptv and 139 pptv respectively) were made.
2.2 above, an up-to-date chemistry process model, SPACCIM (SPectral Aerosol Cloud Chemistry Interaction Model(Wolke et     al., 2005)) has been used to simulate the gas phase HO 2 radical concentrations along a trajectory during the mountain overflow of an air parcel passing an orographic hill cap cloud to further explore the heterogeneous loss processes occurring during the cloud events encountered.This model combines complex microphysical and detailed multiphase chemistry, permitting a detailed description of the chemical processing of gases, deliquesced particles and cloud droplets.SPACCIM incorporates the MCMv3.1-CAPRAMv4.0amechanism (Master Chemical Mechanism(Saunders et al., 2003) / Chemical Aqueous Phase RAdical Mechanism(Tilgner et al., 2013; Braeuer et al., in preparation)) with 11381 gas phase and 7125 aqueous phase reactions.The MCMv3.1-CAPRAM4.0amechanism incorporates a detailed description of the inorganic and organic multiphase chemistry including phase transfer in deliquesced particles and cloud droplets based on a timedependent size-resolved aerosol/cloud spectra.Further details about the SPACCIM model framework and the chemical mechanisms are given elsewhere in the literature(Tilgner et al., 2013;Wolke et al., 2005;Sehili et al., 2005) (and references therein).The measured meteorological data as well as the physical and chemical aerosol and gas phase data at the upwind site in the village of Goldlauter provided the basis for the time-resolved initialisation of the model.In addition, separate initial box model runs with the MCM mechanism were performed to provide a more comprehensive initialisation of the chemical gas phase composition at the simulation start.SPACCIM simulations were performed with an air parcel advected along a predefined orography-following trajectory from the upwind site (Goldlauter) through the hill cap cloud, passing Mt.Schmücke (summit site), to the downwind site (Gehlberg).Parcel simulations were performed every 20 minutes allowing a time-resolved comparison of the predicted and measured HO 2 data at the summit site.

Figure 1 .
Figure 1.Schematic of the FAGE instrument setup during the HCCT-2010 campaign.'PD' refers to photodiode, used to normalise the observed HO 2 signal to laser power.

Figure 2 .
Figure 2. Time-series showing the average liquid water content during each cloud episode (blue, horizontal lines), [OH] (purple), [HO 2 ] (red), j(O 1 D) (orange), NO (green) and O 3 (grey).All data are the average concentrations determined for each FAGE data acquisition cycle apart from OH concentrations which are hourly.

Figure 3 .
Figure 3. Average diurnal profiles of j(O 1 D), OH, HO 2 in cloud (grey) and out of cloud (coloured).The error bars represent the 1σ variability of the averaged data; only the variability in the out of cloud radical data is shown for clarity.Each data point represents 10 minute averaged data apart from the OH, for which the hourly averaged data are given.

Figure 4 .
Figure 4.The dependence of the measured HO 2 concentration as a function of cloud droplet surface area.To remove the influence of changing photolysis rates the measured HO 2 concentrations have been divided by the correspondingly observed rate of photolysis of ozone (j(O 1 D)).This ratio has then been normalized to give a value of 1 when the droplet surface

Figure 5a .Figure 6 ,
Figure 5a.Dependence of the HO 2 concentration observed in cloud as a function of cloud pH.All in-cloud HO 2 data were averaged into corresponding pH bins (0.6 pH units).The [HO 2 ] decreases exponentially with increasing pH with the line of best fit ( ) displayed by the grey line.Figure 5b.The cloud uptake coefficient estimated by optimizing the HO 2 concentration calculated from the analytic expression of Carlsaw et al. (1999) compared to the observed HO 2 concentration as a function of pH (red triangles).The theoretical expression derived by Thornton et al. (2008) (Eq.(4)) using parameters provided in Table2is shown as the black triangles with the grey line being a bestfit line for these data ().

Figure 7 ,HO
Figure 7, lower panel.The dependence of the measured HO 2 concentration (grey circles) and modelled HO 2 concentration with a variable first order loss (red squares) as a function of cloud droplet surface area.

Figure 9 .
Figure 9.Comparison of the measured (green squares) and modelled (red triangles) gas phase HO 2 concentrations at Mt. Schmücke site during the non-cloud event NCE0.8.

Figure 10 .
Figure 10.Annually average fractional change in surface HO 2 , OH, H 2 O 2 and O 3 with the inclusion of HO 2 uptake into clouds leading to a) the production of H 2 O and b) the production of H 2 O 2 assuming a cloud pH of 5 and the Thornton et al. (2008) parameterization.

Figure 11 .
Figure 11.Annually averaged fractional change in column HO 2 , OH, H 2 O 2 and O 3 with the inclusion of HO 2 uptake into clouds leading to a) the production of H 2 O and b) the production of H 2 O 2 assuming a cloud pH of 5 and the Thornton et al. (2008) parameterization.

Figure 12 .
Figure 12.Annually averaged cloud water in the GEOS5 fields as a) a column total and b) a 2 An outline of the aqueous phase reactions thought to be occurring, and which converts HO 2 shows the annual fractional change in surface HO 2 , OH, H 2 O 2 and O 3 concentrations with cloud uptake switched on, and with either H 2 O 2 being produced or not.Column changes are shown in Fig.11.Both with and without H 2 O 2 production, the impact is most evident in areas with long HO 2 lifetimes, i.e. regions with low NO x and low HO 2 concentrations, and with significant cloud water densities (see Figure12).These are concentrated in the extratropics with up to 25% and 10% reduction in surface and column concentrations respectively.

Table 1 .
Details of ancillary measurements used for comparison with radical observations and cubic model constraints.

Table 2 .
The values used for the calculation of the theoretical uptake coefficient, black triangles, Fig.5b, as a function of pH; values given at a pH = 5 here.