Observations of OH and HO 2 radicals over West Africa

Observations of OH and HO2 radicals over West Africa R. Commane, C. F. A. Floquet, T. Ingham, D. Stone, M. J. Evans, and D. E. Heard School of Chemistry, University of Leeds, Leeds, UK National Centre for Atmospheric Science, University of Leeds, Leeds, UK Institute for Climate And Atmospheric Science, School of Earth & Environment, University of Leeds, Leeds, UK now at: School of Engineering & Applied Sciences, Harvard University, Cambridge, USA now at: National Oceanography Centre, University of Southampton, Southampton, UK Received: 11 February 2010 – Accepted: 11 March 2010 – Published: 18 March 2010 Correspondence to: D. E. Heard (d.e.heard@leeds.ac.uk) Published by Copernicus Publications on behalf of the European Geosciences Union.


Introduction
The concentration of the hydroxyl radical (OH) determines the daytime oxidative capacity of the atmosphere.Reaction with OH is the major removal pathway for trace gases such as methane (CH 4 ), carbon monoxide (CO), Volatile Organic Compounds (VOCs) and some halocarbons, often as the first and rate-determining step in their oxidation.A deeper knowledge of the distribution of the concentration of OH and HO 2 (collectively known as HO x ) helps our understanding of atmospheric oxidation and fast photochemical processes.
Usually the primary production pathway of OH in the troposphere, P (OH), is through the photolysis of ozone (O 3 ) and the subsequent reaction of O( 1 D) with water vapour (Ehhalt, 1999): From this, the primary rate of OH production can then be calculated as: OH is quickly converted to the hydroperoxy radical (HO 2 ) by reaction with CO (Weinstock, 1969): The reaction of OH with VOCs produces peroxy radicals (RO 2 ), which, in the presence of NO (nitric oxide), leads to the formation of HO 2 .NO is also important as it rapidly recycles HO 2 to form OH, allowing OH and HO 2 to be considered as a combined HO x family.The ultimate fate of the HO x radicals is the formation of water soluble species such as H 2 O 2 (through HO 2 self reaction) or HNO 3 (through the reaction of NO 2 with OH under high NO x conditions), both of which can be subject to wet or dry deposition.
Tropical regions with high levels of solar irradiation, and often high humidity, have the highest production rates of OH (Bloss et al., 2005).The presence of high concentrations of OH and large surface areas allow tropical regions to dominate global oxidation of long-lived species.The rate of oxidation of CH 4 by OH is also highly dependent on temperature and 80% of this important greenhouse gas is oxidised in the tropical troposphere (Bloss et al., 2005).The tropics also contain large regions of forest, which are a large source of global biogenic VOCs (Fehsenfeld et al., 1992;Guenther et al., 1995), with isoprene (C 5 H 8 ) contributing about 40% of global VOC emissions (Guenther et al., 2006;Kesselmeier, 1999).Most biogenic emissions are controlled by solar irradiation and temperature (Kesselmeier, 1999;Fuentes et al., 2000).Global models calculate that over regions with high concentrations of biogenic VOCs, OH will be depleted compared to concentrations at similar latitudes (Poisson et al., 2000;von Kuhlmann et al., 2004;Lelieveld et al., 2002;Karl et al., 2007).While there have been some observations of HO x in the mid-latitudes (e.g.Heard and Pilling, 2003, and references therein), observations in the tropics have been less extensive, with a distinct focus on the Pacific ocean and American continents.In tropical marine air, OH mixing ratios up to 0.8 pptv (parts-per-trillion, 10 12 mol mol 1 ) and HO 2 mixing ratios between 2 and 45 pptv were observed during the PEM-Tropics-A (Mauldin et al., 1999), PEM-Tropics-B (Tan et al., 2001) and INTEX-B (Mao et al., 2009) aircraft campaigns over the Pacific, while maximum OH and HO 2 mixing ratios of 0.37 pptv and 24 pptv respectively were observed at a surface site in the North Atlantic ocean (Whalley et al., 2009).Lelieveld et al. (2008) found that mixing ratios of OH and HO 2 observed in the presence of high concentrations of isoprene over the rainforests of Suriname were not depleted to the extent expected.OH mixing ratios up to 0.8 pptv and HO 2 mixing ratios up to 80 pptv were observed (Martinez et al., 2010), much higher than calculated in modelling studies (Kubistin et al., 2008).
This paper presents airborne observations of OH and HO 2 radicals above the understudied tropical West African region.The observations were made using a Fluorescence Assay by Gas Expansion (FAGE) instrument flown aboard the BAe-146 as part of the AMMA (African Monsoon Multidisciplinary Analyses) Special Observation Period 2 (SOP-2) intensive in July and August 2006.The instrument and the associated calibration procedure are described in some detail (Sect.2).The variation of HO x with altitude is presented in Sect. 3 and is interpreted in terms of the changing chemical environment, detailed previously in Stewart et al. (2008), Capes et al. (2009) and Reeves et al. (2010).OH mixing ratios are compared to the primary production rate, while HO 2 mixing ratios are compared to a simple steady-state calculation.From these comparisons, a number of interesting cases were identified for further examination.These include observations of HO x in the boundary layer over forested regions (Sect. 4.3), HO x observations in a biomass burning plume (Sect.4.4), HO 2 observed in and around cloud (Sect.4.1) and HO 2 at night (Sect.4.2).In an accompanying paper, Stone et al. (2010) compare the HO x observations presented here with a comprehensive box model.A inter-comparison of HO 2 and peroxy radicals observed on two aircraft flying close together during AMMA is discussed in Andres-Hernandez et al. (2010).

Experimental
The Leeds Aircraft FAGE instrument was designed to be flown aboard the BAe-146 research aircraft operated by FAAM (Facility for Airborne Atmospheric Measurements).
Fluorescence Assay by Gas Expansion (FAGE) is an onresonance low pressure laser-induced fluorescence technique for the measurement of OH and HO 2 (Hard et al., 1984).The strong A 2 6 + (⌫ 0 =0) X 2 5 i (⌫ 00 =0) Q 1 (2) transition of an OH molecule is excited by radiation at ⇡308 nm and the subsequent fluorescence is also detected at ⇡308 nm.The fluorescence cell is maintained at low pressure (ranging from 1.5 Torr at 9 km to 2.5 Torr at sea level) to ensure the fluorescence from the OH molecule lasts longer than the laser excitation pulse, which allows for temporal gating of the fluorescence detection.OH and HO 2 are detected simultaneously with two excitation axes in series.After the OH detection axis, NO is added to the air flow to titrate HO 2 in the sample gas to OH, which is detected in the HO 2 axis.
The sensitivity of the instrument to OH and HO 2 must be determined by calibration.Figure 1 shows the instrument setup.The instrumentation is installed on two 19 inch aircraft racks, located behind the fluorescence cell sampling point with pumps located below the fluorescence cell.

HO x radical sampling
Ambient air from outside the aircraft is sampled through an inlet (adapted from Eisele et al., 1997) mounted on a window blank on the starboard side of the BAe-146, close to the front of the aircraft.Using two concentric tubes with restrictions, the inlet slows the air prior to sampling, while minimising turbulence, and hence wall contact, and the associated loss of OH radicals.
Air is drawn in through a pinhole (0.75 mm diameter) to a fluorescence cell maintained at reduced pressure by a rotary pump (Leybold Trivac D25B) coupled with a supercharger (Eaton M90).In order to extend outside the aircraft skin and deliver OH to the excitation axis, a tube made of anodised aluminium (internal diameter = 5 cm; length = 50 cm) was added to the fluorescence cell.Characterisation tests on the cell have shown significant OH loss on contact with the internal tube walls, thus reducing the instrument sensitivity to OH.No solar scattered light has been detected at either the OH or HO 2 detection axes when the instrument was operated on the aircraft.A fluorescence cell for the detection of RO 2 is shown in Fig. 1 but is not yet operational.

Laser excitation of OH
Laser light at ⇡308 nm was generated by a laser system consisting of a Nd:YAG (Neodymium-ion doped Yttrium Aluminium Garnet (Y 3 Al 5 O 12 ) DS20-532, Photonics Industries) pumped Titanium Sapphire (Ti:Sapp) laser (TU-UV 308 nm, Photonics Industries).The YAG laser typically produces 9-10 W of light at =532 nm at a Pulse Repetition Frequency (PRF) of 5 kHz, which is used to pump the Ti:Sapp laser.The Ti:Sapp output wavelength of ⇡924 nm is selected through the use of a diffraction grating and is passed through the first of two non-linear harmonic generation stages consisting of a Cerium Lithium Borate (CLBO) crystal, generating the second harmonic at ⇡462 nm.A half wave plate corrects the polarisation of the light to allow the second crystal to perform sum-frequency mixing of the second harmonic ( ⇡462 nm) with the fundamental ( ⇡924 nm), producing light at ⇡308 nm.
Under optimal conditions, ⇠100 mW of UV radiation is produced with a pulse width of 35 ns (Full Width Half Maximum, FWHM) and a laser linewidth of 0.065 cm 1 .However, during the observations presented here, typical powers observed after adjustment varied from 30 mW to 50 mW at ⇡308 nm.Dielectric coated beam splitting mirrors split the UV beam before launching the light into fibre optic cables (Oz Optics) in the ratio 60%:36%:4% for use within the OH and HO 2 axes and the reference cell respectively.Upon exiting the fibre optic cable, the laser beam entering each detection cell is collimated and baffled before intersecting the air sample.After traversing the excitation region, the beam exits the cell through another baffled arm and is directed (via a 45 mirror) into a calibrated UV photodiode (UDT -555UV, Optoelectronics), allowing real-time measurement of the laser power.The measured laser power is used to normalise the fluorescence signal for fluctuations in laser power.

Fluorescence detection
In the fluorescence cell (Fig. 2), the UV fluorescence is collected and passed through an interference filter (Barr Associates ⇠50% transmission at =308 nm, FWHM ca. 5 nm centered at =308 nm).A back reflecting mirror (CVI Optics) improves the signal collection efficiency.The fluorescence signal is then focused onto the Channel Photomultiplier (CPM) (Perkin-Elmer 943P, ⇠5⇥10 8 gain), where the electron pulses generated from a single photon are counted with a photon-counting card (PMS 400, Becker and Hickl).To prevent saturation, the CPM is turned off by a high voltage switching circuit ("gating box", designed and built in-house) from the time the laser is triggered until ⇠60 ns after the laser pulse (FWHM=35 ns).The use of a temporal gating allows sensitive measurement of the fluorescence emitted from the electronically excited OH and discrimination against laser scattered light.

Reference cell
The wavelength of the laser is changed by a stepper motor, which controls the angle of the diffraction grating within the Ti:Sapp laser cavity.The absolute wavelength of the second harmonic of the Ti: Sapp ( ⇡462 nm) is monitored by a wavemeter (Wavemaster 33-2650, Coherent).However, the resolution of this wavemeter (precision ±0.001 nm, accuracy ±0.005 nm) is insufficient to locate the OH Q 1 (2) rotational line ( =307.995 nm).The reference cell facilitates scanning over and identification of the Q 1 (2) peak.With the reference cell maintained at low pressure (⇠3 Torr), humidified cabin air is passed over an electrically heated 80:20 Ni:Chrome filament, producing OH radicals by pyrolysis of water vapour (Stevens et al., 1994;Wennberg et al., 1994).The reference cell receives 2-3 mW (⇠4%) of the total UV radiation.The OH fluorescence signal is filtered and focussed onto a CPM perpendicular to the laser beam.The resulting signal is recorded using a photon-counting card.Because of the high concentration of OH radicals produced in the reference cell, discriminating the OH fluorescence signal from the background laser scatter does not require gating of the CPM.
Figure 3 shows the signal observed in the OH, HO 2 and reference cells during a measurement cycle on a flight on August 8 th , 2006.The OH signal in the reference cell is typically 12 000 cts s 1 above the laser scatter background of approximately 4000 cts s 1 .The first 25 points on the figure show the OH fluorescence while scanning over the Q 1 (2) rotational line of OH to locate the wavelength of the peak fluorescence (finding line: from =307.990 nm to =308.000 nm).Points ⇠25-375 in OH and reference cells show the online fluorescence signal (and background) as the laser wavelength is tuned to the peak of fluorescence.In the HO 2 cell, no fluorescence signal is observed until the injection of NO to titrate HO 2 to OH (points 85-375).Points ⇠375-450 show the fluorescence as the laser wavelength is tuned to a wavelength which does not induce fluorescence in order to measure the background signal (offline).
The signal in the reference cell varies with aircraft cabin pressure due to changes in the laser wavelength.The variation in the wavelength generated by the laser is due to the changing refractive index of the cabin air with pressure compared with the material of the laser grating.This change in wavelength away from the maximum of the OH transition results in the signal due to OH fluorescence varying with cabin pressure.Independent of this laser wavelength change with pressure, fluctuations in cell pressure will also alter the instrument sensitivity.While the effect of changing cell pressure on sensitivity can be determined by calibration, it cannot be decoupled from changes in wavelength during AMMA as the Ti:Sapp laser cavity was not kept at constant pressure, so only data recorded on level flights (with constant laser wavelength) are considered.2) rotational line is scanned over ("finding line", points 1-25, blue).Online (points ⇠25-375, red) represents the online measurement period.Offline (points 375-450, black) represents the background signal to be removed in the analysis.A 2.2% (1 ) variation in reference signal is observed online.

Data analysis
The laser is operated at a pulse repetition frequency of 5 kHz and the contribution of the solar and dark counts (CPM thermal noise) to the observed signal is determined for each laser pulse.The fluorescence signal is collected for 1 µs (Gate A).After a 5 µs delay the non-laser background signal is recorded for 20 µs (Gate B) and the signal solely due to OH fluorescence and laser background, Sig fl , is calculated: where SigA (cts) is the number of counts in the 1 µs collection Gate A (cts s 1 ) and SigB is the number of counts in the 20 µs collection gate (typically 0 or 1 cts s 1 ).The total signal, Sig fl (cts), is then recorded at 1 Hz and normalised for laser power (P, mW).For OH, the normalised offline signal OH Sig offline is subtracted from the normalised online signal OH Sig online , leaving the signal solely due to OH (OH Sig (cts s 1 mW 1 )): As shown in Fig. 3, HO 2 Sig is the difference in normalised online signal observed with NO injected (HO 2 +OH) and without NO added (OH only).The concentration of OH is then calculated using the experimentally determined instrument sensitivity (C OH , cts s 1 mW 1 molecule 1 cm 3 ): The concentration of HO 2 is calculated using the experimentally determined instrument sensitivity HO 2 (C HO 2 , cts s 1 mW 1 molecule 1 cm 3 ):

Calibration
Before and after deployment, the instrument sensitivity was determined over the range of cell pressures experienced during the AMMA campaign (between 1.5 and 2.1 Torr, equivalent to 2-2.8 hPa) using a method similar to that employed by Faloona et al. (2004) and Martinez et al. (2010).Pinholes with diameters between 0.65 and 0.85 mm were used to obtain cell pressures between 1.5 and 2.7 Torr (2-3.6 hPa) respectively.
By varying the current supplied to a low-pressure mercury lamp, a range of concentrations of OH and HO 2 were produced from the photolysis of water in humidified air (constant [H 2 O]) at =184.9 nm: where H 2 O is the absorption cross-section of water at =184.9 nm (7.14±0.2⇥10 20molecule 1 cm 2 ; Cantrell et al., 1997a), OH and HO 2 are the photodissociation quantum yields of OH and HO 2 respectively (=1), F is the photon flux of the mercury lamp (photon cm 2 s 1 ) at =184.9 nm and t is the residence time in the photolysis region.
Figure 4 shows the experimental set-up of the calibration system.A known flow (ca.50 slm: L min 1 at standard temperature and pressure) of ultra high-purity air (BTCA 178, BOC Special Gases) is introduced into the system, with a variable fraction of the flow diverted (through bypass valves) to a bubbler containing distilled water in order to humidify the air.After mixing, most of the calibration gas passes through a 30 cm long black-anodised square (1.27 cm⇥1.27cm) aluminium tube, where the gas is photolysed by the =184.9nm emission of a mercury lamp of known actinic flux (see below) to produce equal concentrations of OH and HO 2 before delivery to the FAGE sampling point.Immediately prior to the air entering the calibration tube, a small flow is diverted to a dew-point hygrometer www.atmos-chem-phys.net/10/8783/2010/Atmos.Chem.Phys., 10, 8783-8801, 2010 (CR4, Buck Research Instruments) to measure the concentration of water vapour in the flow.There are three 3.81 cm Suprasil ® windows separated by equal distance down the length of the calibration tube.At high flow rates through the calibration tube, the gas was found to have a uniform velocity profile consistent with turbulent flow.The mercury lamp is housed in a nitrogen-purged aluminium casing, and usually mounted over the window closest to the exit of the calibration tube.This purged housing is heated to maintain a stable lamp temperature (±1 C), while the nitrogen flow prevents both ozone formation within the housing and absorption of radiation by ambient water vapour and oxygen prior to the light entering the calibration tube.The lamp is collimated using thin walled tubes (3 mm diameter, 8 mm length) packed together in front of the lamp.These tubes are used to create a more uniform actinic flux in the photolysis region.
Prior to calibration, the lamp flux at =184.9 nm was determined by N 2 O actinometry through the generation of a quantifiable mixing ratio of NO from the photolysis of 10% N 2 O in air.The theory and method of this technique has been described extensively in previous publications (e.g.Edwards et al., 2003;Faloona et al., 2004;Glowacki et al., 2007;Whalley et al., 2007).Two corrections were made in the calculation of the flux, F .In the NO x analyser (Thermo Electro 42C; detection limit/precision ⇠50 pptv) used to measure the NO produced, the introduction of high concentrations of N 2 O results in the reporting of artifically low NO mixing ratios.This is due to increased quenching of the electronically excited NO 2 fluorescence by N 2 O within the analyser.The effect of this quenching was characterised and the corrected NO mixing ratios were calculated.The second correc-  During AMMA, the cell pressure varied between 1.5 Torr (9 km altitude) and 2.1 Torr (sea level).The dashed line indicates the mean sensitivity used for the AMMA dataset (C OH = 0.63 (±0.095) ⇥10 7 cts s 1 mW 1 molecule 1 cm 3 and C HO 2 = 1.17 (±0.23)⇥10 7 cts s 1 mW 1 molecule 1 cm 3 ).
tion accounted for absorption of the light flux by N 2 O across the width of the photolysis area (Edwards et al., 2003).A linear relationship between the lamp current supplied to the lamp and the flux was observed after corrections, allowing the lamp current to be varied to generate various concentrations of HO x .
Calibrations were conducted over a range of water vapour mixing ratios from 330 ppmv to 10 000 ppmv (parts-permillion, 10 6 mol mol 1 ). Figure 5 shows a typical calibration for the lowest water vapour used (332 ppmv).The instrument sensitivity to OH is a function of water vapour in the gas flow (Smith et al., 2006) due to quenching of electronically excited OH by the water vapour.Using a constant concentration of water vapour during calibrations ensures a linear relationship between the observed signal and concentration of OH or HO 2 .
Over the pressure range observed in the fluorescence cell during AMMA, the sensitivity of the instrument did not vary significantly (Fig. 6) and the mean sensitivities of the instrument to OH and HO 2 were determined (C OH =0.63 (±0.095) ⇥10 7 cts s 1 mW 1 molecule 1 cm 3 and C HO 2 =1.17 (±0.23)⇥10 7 cts s 1 mW 1 molecule 1 cm 3 ).The calibration coefficients for a given pressure before and after the campaign were found to be within 5%, well within the calculated uncertainty.While in-flight calibrations were not possible during the AMMA campaign, instrumentation to allow calibration during flight is currently being developed.The instrument sensitivity to OH is half that of HO 2 and is due to a higher loss of OH compared to HO 2 on the surface of the cell between sampling and detection.

Limit of detection
It is possible to calculate the instrumental Limit Of Detection (LOD) for OH and HO 2 by carrying out a t-test between an observed concentration and zero concentration (0 molecule cm 3 ), assuming the variance in the background is constant throughout the measurement cycle and is represented by the variance in the offline measurement.For unequal sample sizes (online and offline), the minimum detectable HO x , Sig min , for a given confidence interval can be calculated: where T CI is the T value for a given confidence interval (T 69% =1), n and m are the sample sizes online and offline respectively defined by the sampling frequency and bg is the standard deviation of the background, which is assumed to be representative of the standard deviation of the background signal during the online sample.Previous studies have assumed the background signal obeys Poisson statistics (Stevens et al., 1994;Faloona et al., 2004;Martinez et al., 2010) and thus where S lb is the signal due to laser scatter, S sb is the signal due to solar background and S dc are the dark counts of the CPM.Sig min is converted into a minimum detectable concentration of OH (and HO 2 ) by: where C OH and C HO 2 are the instrument sensitivity to OH and HO x (cts s 1 mW 1 molecule 1 cm 3 ) respectively and P is the laser power in the cell (mW).Using a confidence interval of 69% (T value=1), a limit of detection is calculated for each measurement cycle (⇠5 min).
Table 1 shows the OH and HO 2 LODs calculated for a typical measurement cycle during the the final flight of the AMMA aircraft campaign (August 2006).During the campaign the mean OH LOD was 7.1⇥10 5 molecule cm 3 and the mean HO 2 LOD was 7.5⇥10 5 molecule cm 3 for a one minute integration time.In order to improve sensitivity to HO 2 , the time between laser trigger and fluorescence collection is shortened, compared to that set for OH, which increases the background signal (and therefore bg ) and hence limit of detection but will also increase the sensitivity to HO 2 .This shortened delay time is suitable for HO 2 observations, as a typical concentration of HO 2 in the atmosphere is around two orders of magnitude greater than the calculated LOD.During AMMA, the performance of the YAG laser was degraded during flight due to an increase in cabin temperature, which often increased from 15 C to 35 C during a flight.The reduced power output from the YAG resulted in the Ti:Sapphire laser taking longer to reach the lasing threshold.The time delay between the laser trigger and the laser pulse was found to increase by several tens of ns during a given flight, an increase large enough to increase the laser background signal and affect both the sensitivity of the instrument to HO x and the instrumental limit of detection.As the background is recorded at the end of the measurement cycle, it is possible that an increase in background signal throughout a measurement cycle could also lead to a slight underestimation of the OH concentration.
Since the AMMA campaign, a diagnostic system coupling a fast photodiode (Hamamatsu, S6468) to an oscilloscope (Waverunner LT372, Le Croy) has been developed to monitor the timing of the CPM gate trigger relative to the laser pulse and not the laser trigger.This technique allows the time between the laser pulse itself and the CPM gate trigger to be adjusted automatically in flight every 10 s, thus maintaining a constant time delay.A similar system is now used in the Leeds ground-based FAGE instrument and is described in detail in Whalley et al. (2009).Using this system, the sensitivity of the instrument to the time interval between the laser pulse and the fluorescence collection was investigated.Figure 7 shows how the background signal increases dramatically as the time between the laser pulse and fluorescence collection decreases and more of the laser pulse is detected.The limit of detection is also shown to worsen with at shorter time intervals (Fig. 7).
Under normal operating conditions (delay of 60 ns), the instrumental LOD for OH (⇠7⇥10 5 molecule cm 3 ) is below peak concentrations of OH observed in the tropics (e.g.[OH]=9⇥10 6 molecule cm 3 observed at a tropical marine site in summer, Whalley et al., 2009).However, as the delay time decreases, the LOD becomes comparable to ambient concentrations of OH.Ambient concentrations of HO 2 are much greater than the HO 2 LOD (e.g. and the relative increase of HO 2 LOD compared to ambient concentrations has far less of an impact than for OH.

Instrumental uncertainty
During AMMA, water vapour ranged from 300 ppmv at altitudes above 4 km to almost 30 000 ppmv on occasion in the boundary layer.In the laboratory during ground calibrations, the impact of humidity on the calibration coefficient of the aircraft FAGE instrument was determined.For water vapour mixing ratios between 500 ppmv and 10 000 ppmv, changes of 11% and 9% were observed for C OH and C HO 2 respectively.Although the mixing ratio of water vapour was observed on occasion to be higher than this during AMMA, calibrations of other configurations of FAGE instruments up to 20 000 ppmv in Leeds have not observed a significant further change in the calibration sensitivity beyond 10 000 ppmv.
The influence of temperature on the calibration coefficient has not been systematically characterised.However, the distance between the sampling pinhole and the fluorescence cell inside the aircraft (50 cm for the OH cell) should allow sufficient time for enough collisions to ensure equilibration to the temperature within the aircraft.In an earlier configuration of a ground-based FAGE instrument (similar nozzle diameter), the rotational temperature of the hydroxyl radical was measured as a function of distance between the sampling pinhole and the laser-excitation axis (Creasey et al., 1997).Following substantial cooling due to expansion within a supersonic jet, the temperature rose to close to ambient at the Mach Disk, followed by a slight cooling before reaching ambient again at ⇠20 cm.The first fluorescence cell (OH) is 50 cm from the inlet, and a thermocouple placed in the gas flow between the OH and HO 2 fluorescence cells observed room temperature when the instrument was sampling laboratory air.Therefore, it is expected that the temperature within the fluorescence cells will have reached a temperature close to that within the aircraft, although this has not been explicitly verified.
The instrumental uncertainty is calculated as the rootmean-square of the uncertainty in the determination of the flux (13% and includes N 2 O calibration, rate constants, lamp current, etc.), the hygrometer (1%), absorption cross-section of H 2 O (2.8%), flow controller error (1%), excitation peak location (5%), laser power reading (3%), the fit to the multipoint calibration (9% for OH and 6% for HO 2 ), the standard error in the fit in Fig. 5 (15% for OH and 20% for HO 2 ) and the uncertainty due to effects of humidity (12% for OH and 9% for HO 2 ), leading to an overall uncertainty of 25% for OH and 26% for HO 2 .

Ancillary measurements
J (O 1 D) data was provided by the University of Leicester.J (O 1 D) is the combination of the upward and downward welling observations.During the campaign the downward pointing J (O 1 D) actinometer failed and subsequent observations were scaled to the combined J (NO 2 ) observations, as described in Stone et al. (2010).Core aircraft data such as temperature, pressure, CO (Aerolaser AL5002 instrument), O 3 (TECO 49UV photometery instrument) and liquid water content (Johnson Williams Liquid Water Content probe) were provided by the Facility of Airborne Atmospheric Measurements (FAAM) (Reeves et al., 2010), PAN data were provided by the University of York (Stewart et al., 2008) and NO x , NO y (Stewart et al., 2008), acetonitrile and isoprene (Proton Transfer Mass Spectrometry -PTr-MS; Murphy et al., 2010) were provided by the University of East Anglia.

Results: observations of OH and HO 2 radicals
The Leeds aircraft FAGE instrument observed OH and HO 2 over a range of altitudes (500 m-9 km) and through various airmasses over West Africa.The observed signal was processed and the concentration of OH and HO 2 calculated using the instrument sensitivity determined by calibration, as described in Sect.2.6.The 1 Hz mixing ratios of OH and HO 2 were calculated using the temperature and pressure recorded on the aircraft and averaged over 60 s.The 1 Hz HO 2 data were used directly in the study of HO 2 in clouds (Sect.4.1).
HO x observations were made on 13 flights during July and August 2006, with flights both during day and at night. Figure 8 shows the spatial distribution of HO 2 observed over West Africa during the AMMA campaign.The aircraft was based in Niamey, Niger and HO x observations were made over barren soil in the north (Sahel region of northern Niger and Mali), over regions of forest (Benin) and out over the Gulf of Guinea to the south.An overview of the aircraft measurements can be found in Reeves et al. (2010) and an overview of the range of HO x observations observed on various flights is shown in Table 2. Figure 9 is an example of a typical OH and HO 2 time series for a flight on 17 August 2006.The aircraft sampled between 500 m and 6 km over the forested region to the north of Benin (7 -13 N).The O 3 mixing ratios were found to increase with altitude from 20 ppbv (parts-per-billion, 10 9 mol mol 1 ) in the boundary layer to over 60 ppbv above 6 km.OH mixing ratios ranged from below the limit of detection to 0.6 pptv.HO 2 mixing ratios were found to be more variable from 5 pptv to 50 pptv on a 10 s time scale.The highest mixing ratio of both OH and HO 2 was observed at 4 km over the forested region.At the level run at 1.5 km, light cloud was encountered and HO 2 mixing ratios were found to rapidly decrease.This effect will be studied in more detail in Sect.4.1.Overall during the campaign, the mixing ratios of OH and HO 2 were variable.Figure 10 shows the probability density function of the complete dataset of OH and HO 2 mixing ratios, with the median of the limit of detection.OH mixing ratios (Fig. 10a) from 0.27 pptv to 1.34 pptv were observed (but with very few points above 1 pptv), with a median mixing ratio of 0.17 pptv.The most frequent OH mixing ratio was observed at 0.15 pptv.HO 2 mixing ratios (Fig. 10b) varied from 0.32 pptv to 42.65 pptv, with a median mixing ratio of 7.97 pptv.1.5 pptv was the most frequent HO 2 mixing www.atmos-chem-phys.net/10/8783/2010/Atmos.Chem.Phys., 10, 8783-8801, 2010 ratio observed.For significant periods, the OH observations were close to the limit of detection of the instrument.This was due to problems with the instability of the timing of the fluorescence detection (as discussed in Sect.2.7).However, as most HO 2 is well above the limit of detection, conclusions can be drawn with more confidence.

Diurnal variation
Although complex mechanisms determine the in situ concentrations of OH and HO 2 , daytime OH production is generally dominated by the photolysis of ozone and the subsequent reaction of O( 1 D) with water and often exhibits a strong diurnal profile (Rohrer and Berresheim, 2006).Figure 11 shows the diurnal profile of the entire dataset of OH and HO 2 radicals as a function of local solar time.Both OH and HO 2 were highly variable throughout the day, with the maximum OH mixing ratio (⇠1 pptv) coinciding with solar noon, while maximum HO 2 mixing ratio (⇠40 pptv) occurred shortly after solar noon.Mixing ratios of HO 2 up to 9 pptv with a mean of 3 pptv were observed at night.The observations of HO 2 at night will be discussed in detail in Sect.4.2.

HO x observations as a function of altitude
Figure 12 shows the daytime altitude profiles of (a) OH, (b) HO 2 and (c) the HO 2 /OH ratio for data between 09:00 and 15:00 LT (Local Time), with the median value for a given 500 m altitude bin also shown.
OH mixing ratios (binned for 500 m altitude bins up to 6 km) are highly variable but the median OH is relatively constant with altitude at about 0.2 pptv, except for a large increase at the top of the boundary layer at 2 km to 0.7 pptv and distinct decrease at 5 km to less than 0.05 pptv.These observations of OH fall within the range of previous OH mea- surements.Observations over the mid-latitude and tropical Pacific Ocean found that median OH was relatively constant with altitude at ⇠0.1-0.2 pptv, increasing from ⇠0.1 pptv below 2 km to ⇠0.2 pptv at 7 km (INTEX-B and PEM-Tropics-B campaigns, Mao et al., 2009;Tan et al., 2001, respectively).Over continental mid-latitudes, median OH was again relatively constant at all altitudes but at a slightly higher mixing ratio at ⇠0.2-0.3 pptv (INTEX-A campaign, Ren et al., 2008).However, boundary layer OH mixing ratios varied significantly from 0.01-0.6pptv.Observations over a tropical forested region (GABRIEL campaign, Martinez et al., 2010) found much higher OH mixing ratios with median OH increasing from ⇠0.4 pptv in the boundary layer, peaking at 0.75 pptv at 3-4 km and dropping to ⇠0.25 pptv at 8 km.
HO 2 mixing ratios were also variable at all altitudes and median HO 2 varied between 5 and 30 pptv for 500 m altitude bins up to 7 km.The lowest observed mixing ratios of HO 2 were found within 500 m of the forest surface, but the highest HO 2 mixing ratios were observed between 500 m and 1 km above the forest.At altitudes above 6 km, individual HO 2 data points varied from below the limit of detection to over 30 pptv HO 2 , with a median value of ⇠5 pptv.Previous observations of HO 2 generally show a decrease with altitude in all locations.Over the mid-latitude North Pacific HO 2 steadily decreased from ⇠20 pptv at 2 km to ⇠7 pptv at 10 km (INTEX-B campaign, Mao et al., 2009), while over continental mid-latitudes, median HO 2 steadily decreased from ⇠30 pptv near the surface (ranging from ⇠2-60 pptv) to ⇠10 pptv at 10 km (INTEX-A campaign, Ren et al., 2008)  found to decrease from ⇠40 pptv in the boundary layer to ⇠10 pptv at 8 km (GABRIEL campaign, Martinez et al., 2010).However, observations over the tropical south Pacific Ocean show a slightly different profile; HO 2 increased from ⇠10 pptv at the surface to ⇠20 pptv at 3 km before steadily decreasing again with altitude to ⇠5 pptv at 11 km (PEM-Tropics-B campaign, Tan et al., 2001).The HO 2 /OH ratio was highly variable at all altitudes, ranging from 5 to 5 000 with a median value of 95.While HO x production is dominated by O 3 photolysis, the cycling between OH and HO 2 is closely related to the abundance of CO and NO present (Ehhalt, 1999).At the altitudes considered here, overall O 3 varied between 10 ppbv and 100 ppbv and median values gradually increased with altitude to 7 km.CO varied between 50 ppbv and 450 ppbv, but overall decreased slightly with altitude (Reeves et al., 2010).Median NO varied from 100 pptv in the boundary layer to 10 pptv up to 5 km, while total NO x was relatively constant with altitude between 100 pptv and 300 pptv (Stewart et al., 2008).

Steady state calculations
As discussed earlier, the photolysis of ozone in the presence of water vapour is the main production pathway of OH in the unpolluted troposphere (Eq.1). Figure 13 shows the correlation between OH and P (OH).OH mixing ratios were highly variable with only ⇠ 11% of this variability explained by the variability in P (OH).In contrast to this work, previ- ous observations at ground sites with relatively constant OH sinks, found that the observed variation in OH was largely explained by corresponding variations in P (OH) (Rohrer and Berresheim, 2006).
In a clean airmass, the production of HO 2 is closely linked to P (OH) and the loss is dominated by the HO 2 self reaction: Assuming this production of HO 2 (⇠P (OH)) and loss of HO 2 through self-reaction (Reaction R7) are equal, a steady state concentration, HO 2 ss , can be calculated: Figure 14 shows that HO 2 ss reproduces 33% of the variability in the observed HO 2 and the slope of 0.62±0.02indicates that overall HO 2 ss is overestimating HO 2 , suggesting missing sinks of HO 2 .This is not unexpected, given that the rapid loss of HO 2 by reaction with species such as RO 2 or NO 2 or other sinks of HO x radicals such as heterogeneous chemistry.From Fig. 14, a number of distinct groups have been identified.HO 2 is generally reproduced by HO 2 ss in areas of high mixing ratios of isoprene (green) (Sect.4.3).However, the HO 2 /HO 2 ss ratio is much less for observations in a biomass burning plume, suggesting the presence of significant HO 2 sinks (Sect.4.4).Stone et al. (2010) describe how the observed photolysis rates were generally slightly lower than the clear sky J (O 1 D) calculated from the Tropospheric Ultraviolet and Visible (TUV) Radiation model.However, on part of one particular flight (B228, shown in white in Fig. 14) the observations exceeded the calculated J (O 1 D) by a factor of 1.5.Thus, it is possible that the P (OH) calculated at this time is significantly overestimated.This simple analysis is suitable for the examination of the overall HO x behaviour but detailed interpretation requires a more comprehensive chemical modelling study.The detailed chemical model employed by Stone et al. (2010) was found to reproduce 55% of the variability in the observed HO 2 , compared to 33% for the study here.However, the chemical model relies on the availability of ancillary data (e.g.NO) so less data points were available for comparison than used in the steady state calculation of HO x used in this study.Due to large uncertainties and lack of coverage in the observations of OH and other key species (e.g.NO, photolysis rates), Stone et al. (2010) could not confidently compare the model to OH.

Discussion
A number of case studies have been identified in Sect. 3 and will now be examined in detail.A reduction in HO 2 was observed when sampling in cloud (Sect.4.1) and HO 2 was observed at night (Sect.4.2).Compared with P (OH) and HO 2 ss , both OH and HO 2 are enhanced in the presence of high isoprene (Sect.4.3), but are depleted within a biomass burning plume (with acetonitrile higher than 250 pptv) (Sect.4.4).

HO 2 in cloud
When sampling in and around clouds, HO 2 was found to rapidly decrease for short episodes, before returning to the previous higher levels again (Fig. 15).The background signal did not change when sampling around clouds, suggesting that this effect was not an instrumental artefact.Studies of a similar FAGE instrument have shown no response of the instrument sensitivity to aerosol loading (Whalley et al., 2010).Modelling studies have shown that gaseous HO 2 concentrations can be significantly reduced by aqueous phase chemistry, specifically through the efficient uptake of Fig. 15.Time Series of HO 2 variation in Cloud.HO 2 (1 s, pptv, red), HO 2 ss (1 s, pptv, black), scaled Nevzorov probe liquid water content (g m 3 ⇥ 50) (blue).Data were taken between 13:00 and 13:30 UTC at a constant altitude of 1.5 km over northern Benin (10.9 -12.5 N).
HO 2 onto cloud droplets (Tilgner et al., 2005).These aqueous phase models combine detailed microphysics and multiphase chemistry, but few experimental observations exist to validate the model calculated depletion of radical species in clouds.Figure 15 shows the 1 s time series of the observed HO 2 , steady state calculated HO 2 ss mixing ratios and the Liquid Water Content (LWC, g m 3 ) observed between 13:00 and 13:30 UTC at a constant altitude of 1.5 km over northern Benin (10.9-12.5 N) (Flight B235A, 17 August 2006).The timing of the short-lived reduction in HO 2 observed when sampling in cloud is reproduced by HO 2 ss but, as seen earlier, HO 2 ss is consistently greater than HO 2 .Much of the variability in HO 2 ss is driven by changes in J (O 1 D), which varied greatly in and out of cloud, depending on the cloud thickness.O 3 was generally found to decrease in cloud, while the fraction of O( 1 D) reacting with H 2 O vapour to form OH (F OH ) generally increased quickly due to higher water vapour, counteracting the decrease in ozone and J (O 1 D) to an extent.Overall these fast changes in P (OH) resulted in reductions in HO 2 ss .However, the observed reduction in HO 2 was much greater than that seen in HO 2 ss .Thus the relative change in the HO 2 /HO 2 ss ratio is considered for each cloud encounter, as this removes any variation due to changes in the HO x production rates.Figure 16a shows the 1 Hz time series of the HO 2 /HO 2 ss ratio and the simultaneously observed liquid water content.The magnitude of the short-lived decreases in HO 2 inside the cloud are not reproduced in the calculated HO 2 ss , resulting in a decrease in the HO 2 /HO 2 ss ratio.These rapid decreases in the HO 2 /HO 2 ss ratio generally coincide with increases in liquid water and may be due to the heterogeneous uptake of HO 2 onto the cloud aerosol surface.Figure 16b shows the HO 2 /HO 2 ss ratio as a function of liquid water content.While the values of HO 2 /HO 2 ss ratio are highly variable, the highest ratios (median of 0.47) are found for low liquid water (below 0.2 g m 3 ), and the lowest ratios (median of 0.25) are found for higher liquid water content (greater than 0.2 g m 3 ).The reduction in HO 2 in clouds has been observed previously using both FAGE and CIMS instruments (e.g.Mauldin et al., 1999;Olson et al., 2004), and this loss of HO 2 has been attributed to heterogeneous uptake of HO 2 onto the cloud aerosol.During the TRACE-P (Transport and Chemical Evolution over the Pacific) campaign, Olson et al. (2004) found that the observed HO 2 was much lower than that calculated by a comprehensive chemistry model, especially within clouds.For each 1 min data point, observations within cloud were identified.The observed-to-modelled HO 2 ratio was found to significantly reduce when sampling clouds with an increased liquid water content but appeared to be independent of the duration of cloud sampling.The time resolution of many species required to calculate HO 2 in chemical models is too low to interpret the short-lived depletions in HO 2 observed during AMMA (e.g.minimum 15 s for isoprene observed by PTR-MS).Therefore these rapid reductions in HO 2 are unlikely to provide any more than a qualitative sense of the heterogeneous uptake of HO 2 by cloud droplets.
The impact of HO 2 loss due to uptake on liquid cloud was investigated.The first order loss to surfaces was calculated with the diffusion of cloud droplets included (Schwartz, 1984): where k 0 loss is the rate constant for the loss of HO 2 to cloud droplets (s 1 ), A is the cloud droplet surface area per unit volume (m 2 m 3 ), is the uptake coefficient to liquid water, r is the droplet radius (m) and D g is the diffusion coefficient (for cloud droplets ⇠1⇥10 6 m 2 s 1 , Ravishankara, 1997).
The uptake of HO 2 onto cloud is not tightly constrained and the uptake coefficient is thought to vary between 0.1 and 1 (Jacob, 2000;Morita et al., 2004).The mean molecular speed of the gas, c g (m s 1 ) is: where R is the universal gas constant, T is the temperature and M w is the molecular weight of HO 2 .
Unfortunately no information was available on the size or number concentration of cloud droplets encountered for the observation period examined here but the HO x observations were made over the continent of West Africa and typical continental values of cloud droplet size (10 µm) and concentration (1.5⇥10 8 droplets m 3 ) have been assumed in the calculation of k 0 loss (Wallace and Hobbs, 2002).The main production and loss mechanisms of HO 2 were calculated for all HO 2 data where the liquid water content was greater than 0.2 g m 3 and an assumed mid-range of 0.5.
Using these constraints, the mean rate of HO 2 loss to cloud droplets was calculated to be 4.2⇥10 6 molecule cm 3 s 1 , slightly greater than the primary production rate (4.1⇥10 6 molecule cm 3 s 1 ) but much larger than the HO 2 self reaction (0.89⇥10 6 molecule cm 3 s 1 ).Therefore it is reasonable to assume that the uptake of HO 2 onto cloud droplets is a major sink of HO 2 within a cloud.Without the inclusion of HO 2 uptake on clouds, the mean steady state calculated HO 2 mixing ratio, HO 2 ss , is 35 pptv, much higher than the observed mean HO 2 mixing ratio of 10.6 pptv.It is possible to calculate a cloud-influenced HO 2 mixing ratio, HO 2 cloud , assuming that the primary production of HO 2 (through OH) is equal to the loss of HO 2 from both self reaction and uptake on clouds: With the inclusion of HO 2 uptake onto cloud droplets in the calculation of HO 2 , the mean HO 2 cloud mixing ratio was calculated to be 8.84 pptv, in closer agreement with the mean observed HO 2 mixing ratio (10.6 pptv), than the previous steady state calculation of 35 pptv.
In order to fully understand the role of clouds on the oxidative capacity of the troposphere, a more comprehensive field study of HO 2 uptake by clouds is required.The cloud droplet radius, surface area and the concentration of droplets should be measured at the same time resolution as HO 2 observations.In parallel with field studies, laboratory studies are needed to reduce the large uncertainty in current estimates of the HO 2 cloud uptake coefficient ( ), which, at present, limits our understanding of HO 2 behaviour in clouds.

HO 2 at night
HO 2 was observed at night on a number of flights over the forested region but no HO 2 data was collected over the Sahel region or ocean at night.A maximum HO 2 mixing ratio of 9.2 pptv was observed in the hour after sunset and a median HO 2 mixing ratio of 2.2 pptv observed overall at night.Isoprene reached up to 1 ppbv at night in areas north of the most heavily forested region, with the highest mixing ratios coincident with the highest HO 2 mixing ratios.This night-time maximum in isoprene at latitudes north of the day-time maximum is consistent with transport of the airmass north with the monsoon winds (Murphy et al., 2010).
Figure 17 shows the time series of HO 2 after sunset, showing that HO 2 mixing ratios decrease as the night progresses, with the suggestion of an increase again just before dawn.
At night HO 2 can be generated from the ozonolysis of alkenes, (e.g.isoprene) and terpenes (e.g.Kroll, 2002).Isoprene emissions are strongly dependent on sunlight, unlike terpene emissions which are also a function of temperature (Guenther et al., 2006).When night-time temperatures remain high, terpene emission may continue into the night and the reaction of these alkenes and terpenes with ozone leads to the production of HO x .Mixing ratios of HO 2 in excess of that reported here have been observed at night previously at surface sites (e.g.1-4 pptv HO 2 was observed in midlatitude deciduous forest (Faloona et al., 2001), up to 10 pptv HO 2 was observed on Rishiri Island, Japan, (Kanaya et al., 2002) and over 10 pptv HO 2 observed in China, Hofzumahaus et al., 2009).However, no altitude profiles of night-time HO 2 have been reported previously.
Figure 18 shows the altitude profile of (a) HO 2 (pptv), and (b) O 3 (ppbv) observed at night.The greatest HO 2 mixing ratio observed (less than an hour after dusk) was within 500 m of the surface.O 3 mixing ratios were lowest (<20 pptv) at the surface and increased steadily with altitude through the boundary layer.Isoprene, emitted from the surface, and Methyl Vinyl Ketone (MVK), a degradation product of isoprene, were highest in the boundary layer and generally decreased with altitude.This suggests that alkene ozonolysis may play a role as a source of HO 2 at night in the forest boundary layer.Between 6 and 8 km, both isoprene and MVK were observed to increase slightly.As there is no source of isoprene at these altitudes, these species may have been convectively injected into the mid troposphere (Murphy et al., 2010), resulting in the production of HO 2 at higher altitude at night.

OH and HO 2 with isoprene
Figure 19 shows the latitudinal variation of OH and HO 2 and highlights how the maximum mixing ratios of both OH and HO 2 were found in the boundary layer of the southern end of the forest region.The West African Monsoon (WAM) brings the majority of rainfall to the continent and produces a strong latitudinal gradient in vegetation and biogenic emissions such as isoprene: Murphy et al. (2010) show that the highest mixing ratios of isoprene were observed in the boundary layer above the southern forest regions (9-11 N).
Modelling studies of the latitudinal gradient of West Africa predict minimal HO x concentrations over the forested regions due to isoprene scavenging of OH (Saunois et al., 2009).Isoprene reacts quickly with OH (k=1.0⇥10 10 cm 3 s 1 at 294 K; Karl et al., 2004) and, given the concentrations of isoprene in the boundary layer above forests, this reaction should result in significant depletion of OH.
However, the observations presented here suggest that neither OH nor HO 2 mixing ratios are significantly depleted.Lelieveld et al. (2008) also observed higher than expected mixing ratios of HO x over forests in Suriname and suggested that additional production of OH from the degradation products of isoprene (Dillon and Crowley, 2008;Peeters et al., 2009;Paulot et al., 2009;da Silva et al., 2010) could explain the high concentrations of HO x .Given the complex nature of the oxidation of VOCs, the interpretation of these HO x mixing ratios requires a more comprehensive modelling study than presented here.Stone et al. (2010)   Mechanism (Saunders et al., 2003), to explore the interpretation of the HO x observations presented here.

Biomass burning plume
As shown in Figs. 13 and 14, OH and HO 2 sampled in a biomass burning plume showed some of the largest disagreement for the steady state calculations.The two distinct biomass burning episodes identified in Fig. 14 are from 22 July and 8 August 2006.The data from 8 August (observed HO 2 < calculated HO 2 ) is examined in more detail as there are more observations of both HO 2 and other data available.There is no NO x data available for the five data points identified as biomass burning on 22 July (observed HO 2 > calculated HO 2 ).
During the 8 August episode, the biomass burning plume was sampled at two altitudes over the Gulf of Guinea.At an altitude of 4-5 km, the mean O 3 mixing ratio was 50 ppbv and the mean CO mixing ratio was 150 ppbv.At an altitude of 3.5 km, mixing ratios of O 3 and CO were some of the highest observed during the campaign (100 ppbv and 300 ppbv respectively).Coincident with these observations was a sharp increase in NO, acetonitrile (CH 3 CN) and Peroxyacetyl Nitrate (PAN).Acetonitrile has been identified as a tracer for airmasses affected by biomass burning (de Gouw et al., 2003).
Figure 20 shows the time series of O 3 , CO, OH and P (OH) and HO 2 and HO 2 ss .The biomass burning plume was sampled at two altitudes, with the highest concentrations of both O 3 and CO observed at an altitude of 3-4 km (indicated as BB between the blues dashed lines).Within the plume at  3.5 km, mixing ratios of 400 pptv CH 3 CN (up from 100 pptv outside the plume), 600 pptv PAN (up from 50 pptv outside the plume) and 200 pptv NO (up from 50 pptv outside the plume) were observed.Between 4.5 km and 3.5 km, NO mixing ratios were found to increase from 150 pptv to 200 pptv, which is consistent with an increase in the observed OH at the lower altitude.The calculation of HO 2 ss does not incorporate the loss of HO x through reactions such as RO 2 +HO 2 , OH+VOC, OH+NO 2 or reactions with aerosols, so it is not surprising that HO 2 ss is significantly higher than the observed HO 2 .Previous observations of VOCs within aged biomass burning plumes (de Gouw et al., 2003) found that the OH inferred from the degradation of species such as benzene and toluene was a factor of 4 lower than that calculated by a OH climatology model (Spivakovsky et al., 2000).While Forberich et al. (1996) measured reduced levels of OH in a biomass burning plume on one afternoon, a systematic observation of oxidation in biomass burning plumes would help in our understanding of these events.Little biomass burning takes place in West Africa during the monsoon wet season.Within the plume, NO mixing ratios reach a maximum of 200 pptv (NO 2 observations were not available at this time), while PAN mixing ratios increased to over 600 pptv, suggesting that the biomass burning plume was transported from the Southern Hemisphere (Thouret et al., 2009;Liousse et al., 2010).
In this plume, HO 2 mixing ratios are much lower than HO 2 ss , as expected as HO 2 ss does not include all possible HO x loss processes.The reaction of OH with CO forms HO 2 , thus conserving the HO x budget.However, the reaction of OH with any of the VOCs enhanced in this plume www.atmos-chem-phys.net/10/8783/2010/Atmos.Chem.Phys., 10, 8783-8801, 2010 does not immediately form HO x and results in localised HO x depletion.The mixing ratios of a number of VOCs were observed to be greatly increased within the plume compared to outside the plume.Within the plume the greatest increase was observed in acetylene (factor of 7.5), followed by methanol (4.2), acetonitrile (3), benzene (2.9), ethane (2.9) and acetone (2.4).The CO mixing ratio increased from a mean of 98 ppbv outside the plume to 245 ppbv within the plume.The rate of production of HO 2 from the OH reaction with CO can be compared to the rate of production of RO 2 through the reaction of OH with VOCs.
The reactivity of OH with various VOCs can be estimated as a fraction of the OH reactivity with CO: where k i is the rate constant for the reaction of the VOC i with OH and i P k i [VOC] i is the sum over all the products of the VOC concentration and the appropriate rate coefficient for the reaction of OH with the VOC in question.From this ratio, the impact of enhanced VOC concentrations on OH can be examined.The VOC/CO ratio was found to be 1.6 times higher within the plume (71) than outside (43), consistent with the observed HO x depletion within the plume.

Conclusions
The Leeds aircraft FAGE instrument successfully measured tropospheric OH and HO 2 over west Africa during the summer monsoon in 2006.The instrument was deployed aboard the BAe-146 research aircraft as part of the African Monsoon Multidisciplinary Analyses (AMMA) campaign.For calibrations, known concentrations of HO x were generated by the photolysis of water vapour by a mercury lamp, with the lamp flux determined by NO actinometry.The instrument sensitivity did not change appreciably over the pressure range observed in the fluorescence cell during the campaign (1.5-2.1 Torr).The mean instrumental limit of detection observed during the campaign was 7.1⇥10 5 molecule cm 3 for OH (60 s integration time) and 7.5⇥10 5 molecule cm 3 for HO 2 (60 s integration time).HO x observations were made during 13 flights, ranging from 50 m to over 9 km.The aircraft sampled air over the Sahel (18 N), forest (8 -12 N) and over the ocean (4 N).OH and HO 2 mixing ratios show a diurnal profile but were highly variable.Observations of OH and HO 2 were compared with steady state calculation of HO x and differences highlighted a number of case studies.HO 2 data recorded at 1 Hz showed evidence of rapid reductions in the observed HO 2 in and around cloud, which could not be explained by changing primary production rates within the cloud.The heterogeneous uptake of HO 2 onto the cloud surface was included in the steady state HO x calculation and improved the agreement with the observed short-lived reductions of HO 2 in cloud.However, a more comprehensive field study of HO 2 and the properties of cloud droplets is required to fully understand the role of clouds on the oxidative capacity of the troposphere.Mixing ratios of HO 2 of 9 pptv were observed at night and the first altitude profile of HO 2 at night is presented.Up to 7 pptv of HO 2 was observed above 6 km, consistent with the convective uplift of HO 2 precursors such as isoprene.High mixing ratios of OH and HO 2 were observed in areas of high isoprene, in sharp contrast to the depletion calculated by models.Within a biomass burning plume, HO x was found to be depleted.Calculations show that OH was 1.6 times more likely to react with VOCs than CO, resulting in the observed depletion of HO x .

O 3 +
h⌫( < 340 nm) !O( 1 D) ) + M !O( 3 P) + M (R3) The fraction of O( 1 D) reacting with water vapour to produce OH (F OH ) rather than being quenched to O( 3 P) (R3) depends on the water vapour concentration present and is 0.1-0.15 in Published by Copernicus Publications on behalf of the European Geosciences Union.R. Commane et al.: HO x observations over West Africa typical boundary layer conditions.

Fig. 1 .
Fig. 1.Schematic outline of the Leeds aircraft FAGE instrument.Individual components are detailed in the text.The inlet delivers air to the fluorescence cell with OH and HO 2 detection axes.The cell is maintained at low pressure (1.5-2.1 Torr) by a rotary pump backed by a supercharger (Pump Set).The fluorescence cell and pump set are not joined for clarity.The Control Rack contains the instrument computer with photon counting cards, the photomultiplier power supplies and gating electronics, the laser trigger module and a nitrogen cylinder.The Laser Rack holds the YAG pumped Ti:Sapphire laser and chiller unit, the reference cell and a B2 size NO cylinder for HO 2 measurements.The NO cylinder is housed in a containment box designed to prevent any NO leak in the aircraft cabin in case of failure of the cylinder or regulator.Fibre optic cables (not shown) deliver the laser light from the laser to the OH, HO 2 and reference cells.An inlet and fluorescence cell for the detection of RO 2 have been fitted but are not yet operational.

Fig. 2 .
Fig. 2. Fluorescence cell for the sequential detection of OH and HO 2 .OH is detected in the first axis, with fluorescence collected perpendicular to the laser beam and sample flow.NO is then injected to titrate HO 2 to OH and the signal due to HO 2 is detected in the HO 2 axis with fluorescence collection also perpendicular to the laser axis and sample flow.

Fig. 3 .
Fig. 3.The raw signal observed in the (a) OH (cts s 1 ), (b) HO 2 (cts s 1 ) and (c) reference cell (cts s 1 ) for a measurement cycle during a flight on 8 August 2006.The Q 1 (2) rotational line is scanned over ("finding line", points 1-25, blue).Online (points ⇠25-375, red) represents the online measurement period.Offline (points 375-450, black) represents the background signal to be removed in the analysis.A 2.2% (1 ) variation in reference signal is observed online.

Fig. 4 .
Fig. 4. Schematic representation of the calibration system.Zero air is humidified and passed through the calibration tube, where it is photolysed to produce a known concentration of OH and HO 2 .

Fig. 5 .
Fig.5.Multipoint calibration for OH and HO 2 at a constant water vapour (332 ppmv).Calibrations were conducted for a range of water vapour mixing ratios between 10 000 ppmv and 332 ppmv.The lamp current was varied to produce a range of OH and HO 2 concentrations from 1⇥10 7 to 1.7⇥10 8 molecule cm 3 .Error bars indicate the standard deviation of the data within each data point.

Fig. 7 .
Fig. 7. Typical instrument performance for OH (black) and HO 2 (red) changes with the time interval between laser trigger and fluorescence collection.(a) Background signal (cts s 1 mW 1 ) and (b) Limit of Detection (molecule cm 3 ).

Fig. 9 .
Fig. 9. Time Series of an afternoon flight from 10:40 to 15:00 UTC over the forest region in the north of Benin showing (a) altitude (km), (b) latitude ( N), (c) O 3 (ppbv), (d) OH (pptv, 60 s resolution) and limit of detection (pptv, red dashed line, 60 s resolution) and (e) HO 2 (pptv, 10 s resolution) and limit of detection (pptv, red dashed line, 10 s resolution).Error bars on OH and HO 2 indicate the instrumental uncertainty.

Fig. 11 .
Fig. 11.Diurnal variation of (a) OH (pptv) and (b) HO 2 (pptv).Shaded areas indicate night.Specific data discussed in the text is highlighted: green points indicate high isoprene, red points are data identified as biomass burning, blue points indicate cloud sampling, white points highlight data high altitude data collected on flight B228.

Fig. 12 .
Fig. 12. Median altitude profile of (a) OH (pptv, black), (b) HO 2 (pptv, black) and (c) HO 2 /OH ratio for data between 09:00 and 15:00 UTC.Median values in 500 m altitude bins are shown as large joined circles, with error bars indicating one standard deviation from the mean.The median HO 2 /OH ratio of 95 is shown as a grey dashed line.60 s data are shown as small black dots, with specific data discussed in the text highlighted: Green points indicate high isoprene, red points are data identified as biomass burning, blue points indicate cloud sampling, white points highlight data high altitude data collected on flight B228.
Fig. 16.(a) Time Series of HO 2 /HO 2 ss ratio (black) and liquid water (blue); (b) HO 2 /HO 2 ss ratio as a function of liquid water (g m 3 ).The median HO 2 /HO 2 ss ratio is 0.4 and is shown as a dashed line.

Fig. 17 .
Fig. 17.HO 2 time series as a function of time from sunset.Dawn and dusk are indicated by dashed lines.

Fig. 18 .
Fig. 18.Night-time Altitude Profiles of (a) HO 2 (pptv, black) and maximum HO 2 limit of detection (pptv, red) for the night-time dataset.(b) O 3 (ppbv).Median values for 500 m altitude bins are shown as white circles.

Fig. 19 .
Fig. 19.Latitudinal variation of (a) OH and (b) HO 2 mixing ratio.Observations in the boundary layer are shown for data collected above the Sahel (13.75 -18 N, white), Forest (7.3 -13.2 N, green), Ocean (4 -6 N, blue) and Urban areas(6 -7.3 N  and 13.2 -13.75 N, red).Observations in the free troposphere in all regions are shown in black.

Table 1 .
Limit Of Detection (LOD) for a 60 s integration calculated for a typical run on 17 August 2006.

Table 2 .
Overview of HOx observations during AMMA flights.Flights were conducted at various times of day and over a range of terrain from Sahel to the north to heavily forested regions to the south, as well as over the Gulf of Guinea to the far south presents a detailed chemical box modelling study, based on the Master Chemical