HO x observations over West Africa during AMMA : impact of isoprene and NO x

Aircraft OH and HO2 measurements made over West Africa during the AMMA field campaign in summer 2006 have been investigated using a box model constrained to observations of long-lived species and physical parameters. “Good” agreement was found for HO2 (modelled to observed gradient of 1.23 ± 0.11). However, the model significantly overpredicts OH concentrations. The reasons for this are not clear, but may reflect instrumental instabilities affecting the OH measurements. Within the model, HOx concentrations in West Africa are controlled by relatively simple photochemistry, with production dominated by ozone photolysis and reaction of O(1D) with water vapour, and loss processes dominated by HO2 + HO2 and HO2 +RO2. Isoprene chemistry was found to influence forested regions. In contrast to several recent field studies in very low NOx and high isoprene environments, we do not observe any dependence of model success for HO2 on isoprene and attribute this to efficient recycling of HOx through Correspondence to: M. J. Evans (m.j.evans@leeds.ac.uk) RO2 +NO reactions under the moderate NOx concentrations (5–300 ppt NO in the boundary layer, median 76 ppt) encountered during AMMA. This suggests that some of the problems with understanding the impact of isoprene on atmospheric composition may be limited to the extreme low range of NOx concentrations.


Introduction
Tropospheric oxidation is responsible for the removal of many species from the atmosphere which would otherwise build up to dangerous concentrations.This is predominantly achieved through reaction with the hydroxyl radical (OH).For most of the troposphere, OH is produced through the short wavelength solar photolysis of O 3 in the presence of water vapour: This interconversion of OH and HO 2 can also occur through reactions of OH with VOCs (shown for a general hydrocarbon (RH) in reactions R5-R8) through the production of organic peroxy radicals (RO 2 ) and their subsequent reaction with NO: Chain termination occurs through peroxy-peroxy reactions, the reaction between OH and NO 2 and heterogenous uptake of HO 2 (R10-R13) (Jaegle et al., 2000;Monks, 2004;Olson et al., 2004): As a result of the rapid interconversion between OH and HO 2 , the two species are often considered together as HO x ([HO x ] = [OH] + [HO 2 ]).Owing to the intense radiation and high humidity which enhances primary production, the tropics are thought to possess the highest concentrations of OH, and are thus of central importance for the removal of many long lived gases such as CH 4 (Bloss et al., 2005).
Observations of HO x in tropical regions have been made on a number of previous field campaigns.The Pacific Exploratory Mission (PEM) Tropics-B campaign (Tan et al., 2001) made aircraft measurements of OH and HO 2 over the tropical South Pacific Ocean.Using a photochemical box model, Olson et al. (2001) found OH to be overpredicted at low altitudes and underpredicted at higher altitudes, with a mean modelled to measured ratio of 0.86.Simulations of HO 2 gave a mean modelled to measured ratio of 1.03.
The Intercontinental Chemical Transport Experiment B (INTEX-B) aircraft campaign over the North Pacific and western coast of North America made measurements of OH and HO 2 (Mao et al., 2009;Singh et al., 2009).Model calculations overpredicted the OH concentrations (Mao et al., 2009), particularly at altitudes less than 2 km, but found the overprediction to correlate with HCHO concentrations, indicating missing reactive VOCs in the model (Mao et al., 2009).Modelled HO 2 concentrations were generally found to agree with observations (Mao et al., 2009).
While HO x observations in marine tropical environments have generally been explained by modelling studies, fewer observations have been made in tropical forest regions to validate model predictions.Until recently, OH concentrations in forested regions have been expected to remain low as a result of high emissions of biogenic compounds such as isoprene (C 5 H 8 ) which react rapidly with OH (Wang et al., 1998;Lawrence et al., 1999;Lelieveld et al., 2002).
The INTEX-A campaign (Ren et al., 2008) was conducted over continental North America and the western Atlantic Ocean, covering a range of environments and chemical regimes.Model calculations for INTEX-A (Ren et al., 2008) show good agreement for OH in the continental planetary boundary at low isoprene concentrations (< 0.1 ppb), but display significant deviations from the observations when isoprene concentrations are high (upto and greater than 2 ppb), leading to the suggestion that there may be additional unknown HO x sources in forested regions.
Measurements made during the GABRIEL project in Amazonia (Martinez et al., 2010) also found much higher OH concentrations over tropical forests than predicted by model simulations (Lelieveld et al., 2008).Observations of HO 2 were also found to be higher than model predictions during GABRIEL, although not to the same extent as for OH.At the highest isoprene concentration observed (6 ppb) the modelled HO 2 was a factor of approximately 10 too low, compared to a factor of approximately 20 for OH.Such results led to the proposal for a novel mechanism for OH recycling in isoprene oxidation, in which the OH initiated oxidation of isoprene results in reformation of OH through subsequent reactions of HO 2 with RO 2 species, leading to marked improvements in the model simulations (Lelieveld et al., 2008;Kubistin et al., 2008;Butler et al., 2008).It was proposed that recycling of OH in such reactions is masked in regions with high NO x due to competition for reaction with HO 2 between RO 2 and NO, and is only apparent in regions with a combination of low NO x and high VOC concentrations such as Amazonia, where the mean NO concentration in the forest boundary layer during GABRIEL was (10 ± 10) ppt (Lelieveld et al., 2008).Hofzumahaus et al. (2009) were also unable to explain OH measurements made in the Pearl River Delta, China, with OH concentrations in the afternoon 3 to 5 times higher than expected by the model.Measurements of HO 2 , however, were at most only approximately 1.3 times higher than expected, and similarly to OH only showed significant deviation from the model in the afternoon.While VOC concentrations in the Pearl River Delta were relatively high, with mean isoprene concentrations between 1 and 2 ppb, the NO x concentrations were significantly higher (mean noontime NO ∼ 200 ppt) compared to those observed during GABRIEL.Similar model discrepancies for OH have also been found for the OP3 project in the Borneo rainforest (Hewitt et al., 2010;Pugh et al., 2010).
Our current understanding of HO x chemistry in regions with relatively low NO x and high biogenic emissions thus appears weak.It has been proposed that additional recycling of HO x occurs under such regimes through reactions of RO 2 with HO 2 and other RO 2 radicals (Lelieveld et al., 2008).This provides an additional source of HO x that is masked in high NO x environments owing to the prevalence of HO 2 + NO and RO 2 + NO reactions (Lelieveld et al., 2008;Kubistin et al., 2008;Butler et al., 2008;Hofzumahaus et al., 2009).Several theoretical (Peeters et al., 2009;Da Silva et al., 2010) and laboratory studies (Hasson et al., 2004;Le Crane et al., 2006;Jenkin et al., 2007Jenkin et al., , 2008;;Dillon and Crowley, 2008;Paulot et al., 2009) have now proposed mechanisms for additional HO x sources, largely based on RO 2 radical chemistry, yet no definitive conclusions have been reached.
Measurements made during AMMA provide the first opportunity to investigate the processes controlling HO x over West Africa, and can be used to assess the significance of recent findings in HO x chemistry for this region.
In this paper we examine aircraft observations of OH and HO 2 made in tropical West Africa (Commane et al., 2010) using a constrained box model in order to investigate the processes controlling the oxidising capacity in this region.We describe observations made over West Africa in Sect.2, outline a HO x modelling framework based on constrained box modelling in Sect.3, and describe the comparison between the modelled and measured OH and HO 2 in Sects.4 and 5, respectively.We also investigate the processes controlling the modelled HO x , and compare the dominant processes influencing HO x over forest and Sahel regions in Sects.5 and 6.Model sensitivity is discussed in Sect.7 and conclusions are drawn in Sect.8.

AMMA
The African Monsoon Multidisciplinary Analyses (AMMA) field campaign was a large international programme to look at a range of environmental issues in the West African region between 2005 and 2007 (Redelsperger et al., 2006;Lebel et al., 2010;Reeves et al., 2010;Mari et al., 2010).During the period between 17 July to 17 August 2006 the UK FAAM BAe146 research aircraft was stationed in Niamey, Niger, and undertook 21 flights from that base.The geographical extent of the flights is shown in Fig. 1.The aircraft flew in a range of environments, from the deserts in the north, through the Sahel, to the rainforests and ocean in the south.Air was also measured from the cities of Niamey and Lagos ( Hopkins et al., 2009).The aircraft was equipped to measure a number of atmospheric constituents (Reeves et al., 2010).Notably, this mission represents the first HO x measurements made over West Africa (Commane et al., 2010).A range of papers have been written about the observations collected thus far (Redelsperger et al., 2006;Stewart et al., 2008;Capes et al., 2009;Hopkins et al., 2009;Lebel et al., 2010) and only brief descriptions of the measurement techniques are given here.
Ozone was measured using a TECO 49C UV absorption detector.Carbon monoxide measurements were made using an Aero Laser AL5002 Fast Carbon Monoxide Monitor (Gerbig et al., 1999).Non-methane hydrocarbons (NMHCs), volatile organic compounds (VOCs) and oxygenated volatile organic compounds (oVOCs) were measured by a combination of Proton Transfer Reaction Mass Spectrometry (PTRMS) (Murphy et al., 2010) and Gas Chromatography (GC) with a flame ionisation detector (Hopkins et al., 2006(Hopkins et al., , 2009)).For several NMHC/VOC species (acetone, isoprene and MVK/MACR) observations are available from both PTRMS and GC instruments, with measurements by PTRMS made approximately every 15 s, while those by GC were made approximately every 15 min throughout the flight.The two instruments were in good agreement with each other for the species in common (acetone, isoprene and MVK/MACR) (Hopkins et al., 2009;Murphy et al., 2010).
NO x measurements were made by a chemiluminescence TECO 42C analyser with a heated molybdenum filament to enable conversion of NO 2 to NO.Additional NO x measurements were also made using a NO xy instrument on ten flights towards the end of the campaign (Brough et al., 2003;Stewart et al., 2008).The NO xy instrument measures the chemiluminescence from the reaction between NO and O 3 , and converts NO 2 to NO by photolysis.The detection limits for the NO xy instrument are 3 ppt for NO and 15 ppt for NO 2 at an averaging time of 10 s, while those for the TECO NO x analyser are 2 ppb for 30 s measurements and 50 ppt for both species at an averaging time of 120 s.While it has been shown that the total NO x measured by the two techniques are well correlated above 1 ppb (Stewart et al., 2008), the TECO measurements are systematically higher than those obtained using the NO xy instrument, and the individual NO and NO 2 measurements comprising the total NO x signals do not correlate well between the two instruments.Measurements from the NO xy instrument have been used in this work.
Measurements of formaldehyde were made using fluorescence techniques (Cardenas et al., 2000).Peroxy acetyl nitrate (PAN) was measured using gas chromatography with a two channel chromatograph coupled to a pair of electron capture detectors (Whalley et al., 2004).Carbon based adsorbent permeation tubes, with analysis using gas chromatography and detection using time of flight mass spectrometry (GC-TOFMS), were used on a number of flights during the campaign to measure monoterpene concentrations (Saxton et al., 2007;Capes et al., 2009).However, the time resolution of monoterpene measurements was low, and measurements were not made on all flights.
Total aerosol surface area was calculated using Mie Theory and scattering coefficients determined by a three wavelength integrating nephelometer (Anderson and Ogren, 1998;McNaughton et al., 2009).
Photolysis rates of NO 2 j(NO 2 ) and O 3 j(O 1 D) were measured by fixed bandwidth radiometry by the University of Leicester with 2 × 2π-sr filter radiometers (Edwards and Monks, 2003;Bohn et al., 2008).For several flights the upwelling component to the j(O 1 D) signal has been estimated from the j(NO 2 ) signal owing to instrumental problems during the campaign (Brookes, 2009).As a result of the approximations made in estimation of the total j(O 1 D), including those made during calibration to account for the non-linearity of j(O 1 D) with total ozone column, temperature and solar zenith angle, there may be a systematic error in the j(O 1 D), requiring a potential reduction by 30% (Brookes, 2009).Sensitivity of the model to this potential systematic error is discussed in Sect.7.
The focus of this paper is on the OH and HO 2 observations.The AMMA campaign represents the first HO x field measurements obtained with the Leeds aircraft FAGE instrument.The instrument is described in detail by Commane et al. (2010) where an overview of the FAGE observations made during AMMA is also given.The measurements were made by Fluorescence Assay with Gas Expansion (FAGE) using low pressure Laser Induced Fluorescence (LIF) to detect the on-resonance fluorescence of OH at 308 nm (Heard and Pilling, 2003;Commane et al., 2010).A fluorescence cell with two sequential laser axes was used to monitor OH and HO 2 , with HO 2 converted to OH by its reaction with NO, enabling simultaneous detection of OH and HO 2 .Calibrations were performed before and after the campaign using water photolysis at 185 nm to produce a known amount of HO x .The lamp flux was determined by NO actinometry (Commane et al., 2010).Uncertainties in the measurements resulting from the calibration procedure are estimated at 23% (1σ ).
The median limit of detection (LOD) for OH during AMMA was 7.1×10 5 cm −3 (1σ ) for a 60 s averaging period, close to ambient OH concentrations in the troposphere, resulting in many of the OH observations reported being close to or below the limit of detection and subject to large uncertainties.The detection limit is considerably higher than normally experienced for operation of the ground-based Leeds FAGE instrument (Smith et al., 2006), partly due to a lower precision caused by instabilities of the instrument when operated on the aircraft.Some of this instability was caused by temperature fluctuations inside the aircraft which resulted in changes of the timing between the laser pulse and the fluorescence detection (Commane et al., 2010).The poorer precision had the greatest impact on OH measurements as the signal and background signals are similar to one another.The median LOD for HO 2 was 7.5 × 10 5 cm −3 (1σ ), two orders of magnitude less than ambient HO 2 concentrations, resulting in very few measurements being below the LOD, in contrast to OH. Hence the HO 2 measurements were much less sensitive to any instabilities in the timing of the fluorescence detection.

Model approach
In order to interpret the observations a box model is used.Model calculations were made using the Dynamically Simple Model of Atmospheric Chemical Complexity (DSMACC) (Emmerson and Evans, 2009).DSMACC is a flexible zero-dimensional box model using the Kinetic Pre-Processor (KPP) (Sandu et al., 2006).
Simulations presented here used an observationally constrained version, with the chemistry scheme generated by the Master Chemical Mechanism (MCM) (Saunders et al., 2003;Jenkin et al., 2003).The MCM contains near explicit degradation schemes for 135 volatile organic compounds in the troposphere, resulting in over 5600 species and 13 500 reactions and representing a detailed and comprehensive chemistry scheme for tropospheric composition modelling.Rate coefficients for inorganic reactions and three body reactions were updated to recent recommendations (Atkinson et al., 2004;Sander et al., 2006).For these simulations the chemistry of the degradation of ethane, propane, iso-butane, n-butane, ethene, propene, acetylene, isoprene, methanol, ethanol, acetone, methyl ethyl ketone (MEK), methyl vinyl ketone (MVK) and methacrolein (MACR) is used.For all species a physical first order removal reaction was also included.This could be considered as a continuous deposition processes (such as dry deposition or a continuous wet deposition).The default value of the first order rate coefficient for the loss process is 1 × 10 −5 s −1 , giving a lifetime of approximately 1 day.Model results were not found to be strongly dependent on the rate of the deposition process.Sensitivity to this is described in Sect.7.
Loss of OH and HO 2 onto aerosol surfaces was represented in the model by inclusion of a first-order process, using the uptake coefficients γ = 1.0 for OH and γ = 0.1 for HO 2 (Ravishankara et al., 1997;Thornton et al., 2008;Whalley et al., 2010).
The model follows the approach of Olson et al. (2004Olson et al. ( , 2006)).For each observation of OH or HO 2 the objective is to calculate an expected value from the other parameters measured on board the aircraft and the chemical scheme used.All aircraft parameters are merged onto a 60 s timebase for expediency.For each observed time point, those species which have not been observed are set initially to zero.Those species which have been observed are set initially to the observed value and kept constant.The model is integrated forward in time with diurnally varying photolysis rates until a diurnal steady state is reached.Diurnal steady state is reached when the daily mean fractional change in the concentration of each unobserved species and the daily mean fractional change in the sum of all species concentrations between a day and the previous day is less than 0.01.
where N S is the number of species being integrated forwards, N T is the number of integration time steps in a day, [X i ] n,j is the mixing ratio of species i, during the j th time step on day n.Once a diurnal steady state solution has been found the model outputs the concentration of all species so they can be compared to observations.Thus unobserved species are considered to be at their diurnal steady state concentration.The impact of this approximation on HO x is considered in Sects.5 and 6, where the processes controlling HO x in the model are examined.
The model uses the TUV radiation model (v4.2) to calculate diurnally varying photolysis rates for clear sky conditions.An ozone column of 260 Dobson units was assumed (TOMS satellite data, http://toms.gsfc.nasa.gov/ozone/ozone v8.html).Instead of performing the radiation calculation for each integration time point, the calculation was performed for solar zenith angles between 0 • and 90 • in 5 • steps.As the integration occurred and time incremented forwards the solar zenith angle was calculated and the photolysis rates found from spline fitting.Correction factors (due to cloud, albedo etc.) for j (O 1 D) and j (NO 2 ) are calculated from the ratio of the observed and simulated photolysis rates, with the correction factor for j (NO 2 ) being applied to all photolysis reactions other than O 3 → O( 1 D) + O 2 .These photolysis correction factors are calculated for each observation point and applied for the integration forwards to diurnal steady state.
The concentrations of oxides of nitrogen are treated differently to other species.Over an individual 24 h period in the model the concentration of total NO x (defined as NO + NO 2 + NO 3 + 2N 2 O 5 + HONO + HNO 4 ) is kept constant.The concentration of an individual NO x species varies due to changes in photolysis rates etc.At the end of each 24 h period the calculated concentration of one individual NO x species (typically either NO or NO 2 ) is compared to the equivalent measured concentration and the concentration of all NO x species is fractionally increased or decreased so that the measured and the modelled concentrations of the observed species match.As the model simulation runs forwards and a diurnal steady state is reached the fractional change in NO x becomes increasingly smaller.
To be modelled, time points needed to have observations of OH or HO 2 together with the physical state (latitude, longitude, pressure, temperature and water vapour) and observations of CO, O 3 , NO, j (O 1 D) and j (NO 2 ).Since the TECO NO x instrument is known to be the less sensitive instrument, with possible interferences from NO z in the NO 2 measurement (Stewart et al., 2008), the NO xy NO x measurements are considered to be the more reliable.Model input for NO x has therefore been restricted to the NO xy measurements.Concentrations of methane and hydrogen were kept constant at values of 1770 ppm (SCIAMACHY Satellite Data, http://www.iup.uni-bremen.de/sciamachy/NIRNADIR WFM DOAS/) and 550 ppm (Ehhalt and Roher, 2009;Novelli et al., 1999), respectively.Where ethane, propane, iso-butane, n-butane, ethene, propene, acetylene, methanol, ethanol, acetone, formaldehyde, acetaldehyde, peroxy acetyl nitrate (PAN), methyl ethyl ketone (MEK), isoprene, methyl vinyl ketone (MVK) and methacrolein (MACR) were available they were also included.PTRMS measurements have been used where available due to the greater time resolution of the technique compared to GC measurements.Monoterpenes have not been included in standard model runs since measurements were not made on all flights and time resolution of www.atmos-chem-phys.net/10/9415/2010/Atmos.Chem.Phys., 10, 9415-9429, 2010 the measurements is not as great as those for PTRMS or GC measurements.Model sensitivity to monoterpenes is discussed separately in Sect.7.
In order to maximise the number of points in the model, for data points where measurements of VOCs were not available, parameterised alkane/alkene concentrations as a linear function of the CO concentration were used.Table 1 shows the fitting parameters used in this procedure, with typical fits given in Fig. 2. A summary of inputs to the model is given in Table 2.
Data from flight B228 have not been included in the model due to large differences between the observed photolysis rates and those calculated by TUV for a number of data points on this flight.The mean ratio of observed j (O 1 D) to TUV calculations was (0.83 ± 0.28) for all flights excluding B228, while there were a significant number of data points during B228 with j (O 1 D) ratios exceeding 1.5.

Observations and general model performance for OH
Figure 3 shows the comparison between the modelled and measured OH.Although there are some points where there is good model agreement, in general the model provides a systematic overestimate of OH for much of the atmosphere, with a best-fit gradient of 2.5.While some high OH concentrations were observed (Commane et al., 2010), the highest observed concentrations have not been included in this study due to the lack of supporting measurements.This model overestimate is not consistent with previous studies (Ren et al., 2008;Lelieveld et al., 2008;Hofzumahaus et al., 2009).However, as discussed above, and described in detail by Commane et al. (2010), instrumental instabilities led to an increased instrumental limit of detection for OH, and lower confidence in the measurements.The 1:1 line is shown by the solid red line, with dashed red lines indicating 50% limits on the 1:1 line.The best fit to the data is given in blue and is described by y = (2.53 ± 0.43)x + [(1.00 ± 1.60) × 10 6 ], with r2 = 0.75.
In order to reduce the modelled OH to levels similar to the observed values it is necessary to increase the CO concentration by an order of magnitude, or reduce the concentrations of O 3 , H 2 O vapour or NO by an order of magnitude.Such changes to the input parameters are greater than could be reasonable based on the expected observational errors and detection limits for these species, and adversely affect the model agreement observed for HO 2 (Sect.4.2), suggesting that the discrepancy between modelled and observed OH may result from experimental difficulties associated with the measurement of OH.Given this result, it has not been possible to investigate the OH chemistry using the AMMA dataset and we concentrate on understanding the HO 2 observations, where the experimental signals are much larger and well above the detection limits.The 1:1 line is shown by the solid red line, with dashed red lines indicating 50% limits on the 1:1 line.The best fit to the data is given in blue and is described by y = (1.23 ± 0.11)x + [(3.85 ± 3.06) × 10 7 ], with r 2 = 0.75.The data point in red marks the single point referred to in Sect.4.2.1 for which the sensitivity to isoprene and NO x was investigated and used to represent the midday forest boundary layer referred to in Sect.4.3 and Fig. 7.The data point in blue marks the single point used to represent the midday Sahel boundary layer referred to in Sect.4.3 and Fig. 10.

Observations and general model performance for HO 2
Figure 4 shows the scatterplot of modelled to observed HO 2 concentrations.The best fit line between the model and the observations is given by [HO 2 ] mod = (1.23 ± 0.11)[HO 2 ] obs + (3.85 ± 3.06) × 10 7 with a correlation coefficient squared of 0.75.This is just within the calibration uncertainty of the measurements (approximately 23% at the 1σ level) (Commane et al., 2010).The model was thus capable of reproducing much of the trends and variability in the observed HO 2 , although there is in general an overestimation of HO 2 .
While model performance could be considered good it is not altogether surprising.Assuming simple photochemistry, most of the HO x production should be due to the reaction between O(1 D) (produced by O 3 photolysis) and H 2 O vapour, and most of the HO x loss should be due to the HO 2 self-reaction.Thus at steady state the HO 2 concentration, [HO 2 ] ss , can be determined by a consideration of the rates of reactions below, and is thus given by Eq. (2).= 0.74.While the apparent separation of the data into two regimes does correspond to data points with high isoprene and those with low isoprene, discrimination on the basis of geography is not possible since both regimes contain a mixture of boundary layer and free troposphere data points over both forest and Sahel.

[HO
where F OH represents the fraction of O( 1 D) reacting with H 2 O to form 2 OH rather than being quenched (Eq.3).
Using this simple relationship on the dataset gives the calculated steady state HO 2 shown in Fig. 5.The best fit line between the calculated steady state HO 2 and the observations is given by [HO 2 ] ss = (1.73 ± 0.16)[HO 2 ] obs − (0.62±4.33) × 10 7 with a correlation coefficient squared of 0.74, showing that much of the photochemistry determining HO 2 concentrations during AMMA was indeed dominated by relatively simple processes and that almost all the variability in HO 2 is driven by photolysis rates and concentrations of water and ozone.There are, however, significant HO 2 sinks missing from the steady state model.In the full DSMACC model run the additional sinks are due to HO 2 + RO 2 reactions.

Impact of isoprene
The model agreement for HO 2 during AMMA (Fig. 4) appears somewhat contradictory to recent studies discussed in the introduction which imply a significant lack of understanding of HO x chemistry in forested areas with high isoprene and low NO x levels (Ren et al., 2008;Lelieveld et al., 2008).Figure 6 shows the model success (Eq.4) as a function of isoprene concentration for this study.Although there is some indication of model underestimation at high isoprene (ratio of observed to modelled HO 2 of ∼1.5), the underestimation is small compared to that observed during GABRIEL where a ratio observed to modelled HO 2 of approximately 7 was observed in the boundary layer in the afternoon, with a ratio between 3 and 5 at isoprene concentrations comparable to those observed during AMMA (Kubistin et al., 2008).
In order to further assess the processes controlling HO 2 , and the effects of isoprene, we have selected a single time point from the dataset corresponding to a measurement taken at 13:53 GMT on day 229 in the forest boundary layer (latitude 10.13 • N, longitude 2.69 • E, pressure 951.5 hPa) and conducted sensitivity tests to the isoprene and NO concentrations.The data point has a relatively high concentration of isoprene (1184 ppt compared to the mean of 294 ppt and standard deviation 333 ppt over the forest boundary layer) and a relatively low NO concentration of 51 ppt (compared to the mean of 59 ppt and standard deviation 30 ppt over the forest boundary layer).Figure 4 shows the data point in red on the HO 2 model to observation scatterplot.Concentrations of measured species for this point are listed in Table 3.
Figure 7 shows the instantaneous RO ), HO x and HO 2 budgets calculated for this point.Production of radicals (Fig. 7a) is dominated by ozone photolysis, followed by photolysis of formaldehyde (HCHO) and methylglyoxal (MGLYOX).The model is constrained with observations of both O 3 and HCHO.Both formaldehyde and methylglyoxal are oxidation products of isoprene, as are several of the principal species photolysed to produce RO x .Radical loss is dominated by peroxide formation from HO 2 + RO 2 and HO 2 + HO 2 , with reaction of HO 2 and isoprenyl peroxy radicals (ISOPO 2 , of which there are four isomers in the MCM) constituting the major radical loss process.
Diagnosing the HO x budget (Fig. 7b) shows that while OH + C 5 H 8 and HO 2 + ISOPO 2 are dominant HO x sinks, there is significant production of HO x from decomposition of isoprenyl alkoxy radicals (ISOPO), which not only produce HO 2 directly but also produce aldehyde species (HCHO, HC4ACHO and HC4CCHO, capitalised species names are those used in the MCM) which photolyse to generate further HO 2 .There is also significant production of HO x from CH 3 O, produced mainly by reaction of NO with CH 3 O 2 , which is derived from oxidation of both methane and isoprene.
If there is to be a net sink of HO x resulting from OH + isoprene then isoprenyl peroxy radicals (ISOPO 2 ) must react predominantly with HO 2 to produce peroxides.However, at the NO x levels encountered over the forest boundary layer during AMMA ([NO] = (59 ± 30) ppt) the reactions of ISOPO 2 with NO compete efficiently (70%) with those of ISOPO 2 with HO 2 , providing a means of generating ISOPO radicals and ultimately HO 2 through subsequent chemistry, thereby helping to conserve the total HO x radical budget.A generalised chemical scheme for these processes is depicted in Fig. 8.We therefore conclude that at the NO x and isoprene concentrations experienced in this location during AMMA, isoprene has little overall impact on HO 2 as the NO x concentrations are high enough to minimise peroxide formation and to increase the production of HO 2 source species such as HCHO and other aldehydes.Thus isoprene cycling provides a "null" cycle for HO 2 at these isoprene and NO x concentrations.We now investigate the chemical range for which this "null" cycle is valid.

Sensitivity of HO 2 to changes in isoprene at different isoprene and NO concentrations
Figure 9 shows the results of a sensitivity study, using the single point described above, to investigate the combined effects of isoprene and NO x on the modelled HO 2 in which the isoprene and NO x inputs to the model have been varied while maintaining all other species at their observed levels www.atmos-chem-phys.net/10/9415/2010/Atmos.Chem.Phys., 10, 9415-9429, 2010 (Table 3).The plot shows the ratio of the modelled HO 2 for a base run in which isoprene and NO x were varied, compared to an analogous model run in which the isoprene concentrations were multiplied by a factor of ten, thus showing the effects of increasing isoprene concentration under different isoprene and NO x regimes.We split the chemical space into four regimes dependent on the NO x concentration, and discuss each in turn below.
Impact of isoprene at 50 ppt > NO > 10 ppt Figure 9 allows identification of a number of regimes affecting the modelled HO 2 .At low NO x levels (NO < 50 ppt), loss of HO 2 is dominated by HO 2 + HO 2 and HO 2 + RO 2 , particularly ISOPO 2 , and HO 2 is sensitive to changes in isoprene concentration.Increasing the isoprene concentration at such low NO x levels leads to decreased modelled HO 2 .
Impact of isoprene at 500 ppt > NO > 50 ppt As the NO x level increases, loss of HO 2 becomes dominated by HO 2 + NO and the modelled HO 2 becomes insensitive to isoprene.Although there is some peroxide formation at higher NO x concentrations, leading to loss of HO x , the HO x loss is offset by direct HO 2 production from RO, formed by RO 2 + NO, and RO-generated HCHO which photolyses to produce 2 HO 2 radicals.
Impact of isoprene at NO > 500 ppt However, as the NO x level is increased further (NO > 500 ppt) the modelled HO 2 becomes sensitive to the isoprene input once more as the relative importance of HO 2 production from alkoxy radicals such as CH 3 O and ISOPO increases due to increased production of these species from RO 2 + NO reactions, and the modelled HO 2 increases if isoprene is raised at high NO x levels.Production of HO 2 from alkoxy radicals also results in co-production of HCHO, which increases the modelled HO 2 further.
Impact of isoprene at NO < 10 ppt At low NO x and low isoprene levels (< 10 ppt), HO 2 production and loss is dominated by OH + CO and HO 2 + HO 2 and HO 2 + CH 3 O 2 , and the modelled HO 2 is relatively unaffected by isoprene.As the isoprene concentration increases, HO 2 loss becomes dominated by reaction with ISOPO 2 radicals, and the modelled HO 2 is reduced on increasing isoprene.At higher isoprene concentrations and low NO x , however, production of HCHO from C 5 H 8 + O 3 reactions be- concentrations relative to a model run where the isoprene concentration was a factor of ten lower (th y-axis refers to the lower isoprene concentrations).Model runs described here were unconstrained HCHO and isoprene degradation products.The 1:1 ratio for the modelled HO 2 concentrations is show by the black line.Black circles indicate NO and isoprene concentrations encountered during AMMA with the average conditions for AMMA shown by the red diamond.The blue diamond indicates th average NO and isoprene concentrations encountered during GABRIEL (Lelieveld et al., 2008) an the purple diamond the median isoprene and NO concentrations encountered during INTEX-A whe isoprene was above 2 ppb (http://www-air.larc.nasa.gov/missions/intexna/intexna.htm).41 Fig. 9. Contour plot showing the fractional changes in modelled HO 2 at different NO and isoprene concentrations relative to a model run where the isoprene concentration was a factor of ten lower (the y-axis refers to the lower isoprene concentrations).Model runs described here were unconstrained to HCHO and isoprene degradation products.The 1:1 ratio for the modelled HO 2 concentrations is shown by the black line.Black circles indicate NO and isoprene concentrations encountered during AMMA, with the average conditions for AMMA shown by the red diamond.The blue diamond indicates the average NO and isoprene concentrations encountered during GABRIEL (Lelieveld et al., 2008) and the purple diamond the median isoprene and NO concentrations encountered during INTEX-A when isoprene was above 2 ppb (http://www-air.larc.nasa.gov/missions/intexna/intexna.htm).comes important, and the modelled HO 2 is actually increased on increasing the isoprene input due to increased production from HCHO photolysis.At such high isoprene concentrations the modelled HO 2 is relatively insensitive to the NO x input.
Overall, within the MCM concentrations of HO 2 in the atmosphere are a complex function of the isoprene and NO concentrations, and efficient production of HO x species through isoprene and isoprene derived peroxy and alkoxy radicals can be achieved through formation of HO 2 and HCHO, with subsequent photolysis, from ISOPO radicals generated by reaction of ISOPO 2 with NO.

Influence of isoprene and NO x on HO 2 during AMMA
During AMMA the mean isoprene concentration observed in the forest boundary layer was (294 ± 333) ppt, with a maximum of 1232 ppt, while the mean NO concentration was (59 ± 30) ppt.These concentrations are marked on Fig. 9, and compare to (2000 ± 760) ppt isoprene and (20 ± 20) ppt NO for the GABRIEL campaign (Lelieveld et al., 2008), also marked on Fig. 9, although it should be noted that differences in ozone and CO concentrations between AMMA and GABRIEL and INTEX-A may contribute to the placement of the different campaigns in different chemical regimes.
Data recorded during AMMA corresponds to a regime in which HO 2 chemistry is relatively insensitive to variations in isoprene and is dominated by NO x due to efficient conversion of ISOPO 2 back to HO 2 .This contrasts to the GABRIEL situation where NO concentrations are significantly lower and isoprene concentrations are significantly higher.At the lower NO x and higher isoprene concentrations observed during GABRIEL the modelled HO x will be more sensitive to changes in isoprene than for AMMA.Many of the suggested explanations for the problems simulating HO x rely upon the fate of the isoprenyl peroxy radicals when they do not react with NO (Lelieveld et al., 2008;Hofzumahaus et al., 2009;Paulot et al., 2009;Peeters et al., 2009;Da Silva et al., 2010).Under the AMMA conditions, mis-representation of this chemistry should not lead to significant failures to simulate HO x and given our observations of HO 2 this seems to be the case.

Sources and sinks of HO 2 and HO x
Although forest-impacted boundary layer air formed an important aspect of AMMA the aircraft campaign investigated air with very different characteristics, from the forest to the Sahel and from the surface to the mid-troposphere.We now examine the chemistry controlling HO x concentrations in these different geographical regions.We split the atmosphere into regions above rainforest (7.3 • N to 13.2 • N) and Sahel (>13.75 • N) and examine the chemistry responsible for determining HO x and HO 2 concentrations in each region.
Figures 7 and 10 show the RO x , HO x and HO 2 budgets for typical data points representative of the forest and Sahel boundary layers, respectively.The data points are shown in Fig. 4. The data point over the forest is that described previously in Sect.5, and refers to a data point taken at 13:53 GMT on day 229 in the forest boundary layer (latitude 10.13 • N, longitude 2.69 • E, pressure 951.5 hPa).The data point over the Sahel was taken at 14:01 GMT on day 223 in the Sahel boundary layer (latitude 14.48 • N, longitude 4.35 • E, pressure 961.3 hPa).Concentrations of measured species for these two points are listed in Table 3.
The main source of HO x in both forest and Sahel boundary layers is ozone photolysis followed by the reaction of O( 1 D) + H 2 O.This accounts for over 42% of the total HO x production over the forest and 61% over the Sahel.In both regions this is followed by reaction of CH 3 O, produced mainly by CH 3 O 2 + NO, with O 2 .Over the Sahel, these two processes account for approximately 90% of the total HO x production, with the remaining 10% predominantly from photolysis of HCHO and H 2 O 2 .Over the forest region, O 3 photolysis and CH 3 O + O 2 account for only approximately 50% of the total HO x production, with contributions from numerous species, but with notable contributions from ISOPO radicals and photolysis of HCHO.Loss processes for HO x over the forest are dominated by C 5 H 8 + OH, producing ISOPO 2 radicals, with a large number of reactions involving VOCs and their oxidation products contributing to the total HO x loss.In contrast, loss processes over the Sahel are dominated by the HO 2 selfreaction, CH 4 + OH and OH + HO 2 with considerably fewer contributions from other reactions.Losses due to aerosol were minimal, representing less than 1.5% of the total loss for HO 2 and less for OH.
Figures 7c and 10c show the modelled HO 2 budgets for the typical forest and Sahel points.Over the Sahel, the most significant reactions for HO 2 production are OH + CO and CH 3 O + O 2 .Over the forest, however, total production of HO 2 from all four ISOPO radical isomers combined is of greater importance than either OH + CO or CH 3 O + O 2 alone, and HO 2 production from ISOPBO alone is of equal importance as OH + CO or CH 3 O + O 2 .In addition, production of HO 2 from HCHO photolysis is twice as important over the forest than over the Sahel.
Loss of HO 2 is dominated by NO + HO 2 and HO 2 + HO 2 in both regions, but while these reactions account for approximately 85% of the total HO 2 over the Sahel they account for only 62% over the forest.The high biogenic emissions over the forest lead to a greater dominance of HO 2 loss through reaction with RO 2 , particularly those derived from isoprene.
In general, reactions of HO 2 with RO 2 , resulting in peroxide formation, are of greater importance over the forest than the Sahel, indicating the difference in VOC emissions between the two regions.

Model sensitivity
All models make various assumptions and approximations.The impact of some of these on the conclusions described earlier (Sects.4 and 5) are analysed here.

Influence of photolysis rates
Instrumental problems with the j (O 1 D) downward facing radiometer led to the total j (O 1 D) being estimated from measurements of the upward pointing j (O 1 D) and the total j (NO 2 ) for several flights during the campaign, and may result in some uncertainty in the photolysis input to the model.Assumptions made in the calibration procedure to account for the non-linearity of the j (O 1 D) calibration factors for total ozone column, temperature and solar zenith angle, may have also introduced some systematic error to the j (O 1 D), requiring a potential reduction in the j (O 1 D) used in the base model run by 30% (Brookes, 2009).
A reduction in j (O 1 D) by 30% changes the gradient of modelled to observed HO 2 from (1.23 ± 0.11) to (1.05 ± 0.09).This may provide some explanation for the slight overestimation of HO 2 in the model and illustrates the need for stringent and well characterised calibration procedures for constraints in the model.

Influence of monoterpenes
Monoterpene emissions are a significant source of carbon to the atmosphere (Adams et al., 2001).During AMMA monoterpene concentrations were measured on several flights (Capes et al., 2009), but measurements were sporadic and with relatively low time resolution.Average concentrations of α-pinene, β-pinene, camphene and limonene were included in the model as the mean concentration for the flight for every time point on each flight on which they were measured, with a maximum total monoterpene concentration of 130 ppt.The monoterpenes were treated as α-pinene (i.e.[α-pinene] = [α-pinene] + [βpinene] + [camphene] + [limonene]) due to the availability of a full α-pinene degradation scheme in the MCM.However, little difference on the model results was observed with the gradient of the modelled to observed HO 2 changing from (1.23 ± 0.11) to (1.18 ± 0.10).Thus we conclude that not having monoterpene observations does not significantly impact the simulations performed here or the conclusions reached.

Impact of deposition timescale
Removal of all species is included in the model to represent physical loss of species through deposition.The default unimolecular rate coefficient used to describe this deposition process in the model is 1 × 10 −5 s −1 , corresponding to a lifetime approximately of 24 h.During the sensitivity analysis the rate coefficient was varied between 2.78 × 10 −4 and 2.31 × 10 −6 s −1 , corresponding to lifetimes between 1 h and 5 days.Modelled HO 2 was not significantly affected by the changes to the deposition rate, with the gradient of the modelled to observed HO 2 changing from (1.10 ± 0.10) for a lifetime of 1 h to (1.23 ± 0.11) for a lifetime of approximately 24 h to (1.25 ± 0.11) for a lifetime of 5 days.Thus we conclude that our model simulations are not significantly influenced by this choice as most of the HO x sources and sinks in the model are observed, or are short lived and thus unaffected by deposition.

Conclusions
Aircraft measurements of OH and HO 2 made during the AMMA campaign in 2006 have been compared to model calculations using the Dynamically Simple Model of Atmospheric Chemical Complexity (DSMACC).The model makes a systematic overestimation of OH, most likely reflecting instrumental issues with OH detection during the campaign, and this work has focussed on understanding the HO 2 concentrations.Modelled HO 2 concentrations were found to agree with observations within the calibration uncertainties of the HO 2 measurement and uncertainties associated with the photolysis rate input to the model.HO x budget analyses indicate that HO x concentrations in West Africa are controlled by relatively simple photochemistry, with HO x production generally dominated by ozone production followed by reaction of O( 1 D) with water, while HO x removal processes are dominated by HO 2 + HO 2 and HO 2 + RO 2 .Isoprene chemistry was found to influence the HO x budget over forested regions.
Despite recent evidence indicating misrepresentation of HO x chemistry in high isoprene and low NO x environments we observed no dependence of model success on isoprene concentration, and conclude that under the conditions encountered during the AMMA campaign NO x levels were sufficiently high to enable efficient recycling of HO x species through RO 2 + NO reactions.

Fig. 1 .Fig. 1 .
Fig. 1.Locations of the BAe146 aircraft durin available, showing (a) the wider geographical are figure Observed ethane concentrations as a function of CO, r 2 =0.737 and (b) observed iso-butane rations as a function of CO, r 2 =0.268.The best fit lines are given in blue.
Observed ethane concentrations as a function of CO, r 2 =0.737 and (b) observed iso-butane rations as a function of CO, r 2 =0.268.The best fit lines are given in blue. 34

Fig. 2 ..Fig. 3 .
Fig. 2. (a) Observed ethane concentrations as a function of CO, r 2 = 0.737 and (b) observed iso-butane concentrations as a function of CO, r 2 = 0.268.The best fit lines are given in blue.

Fig. 4 .Fig. 4 .
Fig. 4. Comparison between modelled and observed HO 2 concentrations.The 1:1 line is shown by the solid red line, with dashed red lines indicating 50% limits on the 1:1 line.The best fit to the data is given in blue and is described by y=(1.23±0.11)x+(3.85±3.06)×10 7, with r 2 =0.75.The data poin in red marks the single point referred to in Sect.4.2.1 for which the sensitivity to isoprene and NO x was investigated and used to represent the midday forest boundary layer referred to in Sect.4.3 and Fig. 7 The data point in blue marks the single point used to represent the midday Sahel boundary layer referred to in Sect.4.3 and Fig. 10.

Fig. 7 .Fig. 7 .Fig. 7 .Fig. 7 .
Fig. 7. Processes controlling the instantaneous production and loss a typical data point in the forest boundary layer (shown in red in F the MCM.

Fig. 8 .Fig. 8 .
Fig. 8. Simplified schematic showing processes controlling radical budgets during AMMA.Bold lines indicate the dominant reactions in isoprene oxidation chemistry during AMMA.

Fig. 9 .
Fig.9.Contour plot showing the fractional changes in modelled HO 2 at different NO and isopren concentrations relative to a model run where the isoprene concentration was a factor of ten lower (th y-axis refers to the lower isoprene concentrations).Model runs described here were unconstrained HCHO and isoprene degradation products.The 1:1 ratio for the modelled HO 2 concentrations is show by the black line.Black circles indicate NO and isoprene concentrations encountered during AMMA with the average conditions for AMMA shown by the red diamond.The blue diamond indicates th average NO and isoprene concentrations encountered during GABRIEL(Lelieveld et al., 2008) an the purple diamond the median isoprene and NO concentrations encountered during INTEX-A whe isoprene was above 2 ppb (http://www-air.larc.nasa.gov/missions/intexna/intexna.htm).

Fig. 10 .Fig. 10 .Fig. 10 .Fig. 10 .
Fig. 10.Processes controlling the instantaneous production and lo for a typical data point in the Sahel boundary layer (shown in blue in the MCM.

Table 1 .
Parameters used to fit hydrocarbon data as a function of the CO concentration.Hydrocarbon concentrations were used in ppt, while CO concentrations were in ppb.RMS refers to the root mean square difference between the observed and calculated values.

Table 2 .
Summary of data inputs to the model.Chemical names are those used in the MCM.

.
Comparison between observed HO 2 concentrations and a simple steady state model.The 1:1 shown by the solid red line, with dashed red lines indicating 50 % limits on the 1:1 line.The best the data is given in blue and is described by y = (1.73 ± 0.16)x + (0.62 ± 4.33) × 10 7 , with r 2. While the apparent separation of the data into two regimes does correspond to data points with soprene and those with low isoprene, discrimination on the basis of geography is not possible since egimes contain a mixture of boundary layer and free troposphere data points over both forest and .

Table 3 .
Model inputs for the single points over the forest and Sahel used to investigate the sensitivity of the modelled HO 2 to isoprene and NO x concentrations and to investigate the processes controlling the chemistry of the forest and Sahel.For investigation of the sensitivity of the modelled HO 2 to isoprene and NO x concentrations the data point over the forest was used, with isoprene and NO x concentrations as shown in Fig.9.Chemical names are those used in the MCM.