Nicola Carslaw, Philip Jacobs, Mike Pilling
At Leeds, our modelling efforts are concentrated on the chemistry of OH, HO2 and peroxy radicals. The modelling studies take place in conjunction with measurements made by the atmospheric field studies group. Our aim is to use the measured data at campaigns to be able to reproduce OH, HO2 and peroxy radical concentrations measured by the FAGE and PERCA instruments. By using models to reproduce atmospheric measurements of radical species, we stand to learn much about the chemistry of the boundary layer.
The hydroxyl radical (OH) is widely recognised as the most important oxidant in the earth's troposphere, and is largely responsible for the removal of many of the pollutants that are released due to man's activities, as well as a significant number of biogenic species. The OH radical is produced in the troposphere primarily through the following reactions:
O3 + hv = O2 + O(1D) (< 320 nm) (1)
O(1D) + H2O = 2OH (2)
Most of the O(1D) is collisionally deactivated in the atmosphere, but about 10% reacts with water vapour to form the OH radical (reaction (2)). The two most important loss routes for the OH radical in the clean troposphere are the reactions with carbon monoxide (CO) and methane (CH4).
OH + CO (+O2) = HO2 + CO2 (3)
OH + CH4 (+O2) = CH3O2 + H2O (4)
The hydroperoxy (HO2) and methylperoxy (CH3O2) radicals are the first in the homologous series known as the peroxy radicals.
HO2 + NO = OH + NO2 (5)
CH3O2 + NO = CH3O + NO2 (6)
CH3O + O2 = HO2 + HCHO (7)
In more polluted atmospheres, the OH radical can undergo analogous reactions with a variety of non-methane hydrocarbons (NMHC). In the absence of NOx (the sum of NO2+NO), the peroxy radicals will undergo either self-reactions or cross-reactions with other peroxy radicals.
HO2 + HO2 = H2O2 (8)
CH3O2 + HO2 = CH3OOH + O2 (9)
Hydrogen peroxide (H2O2) and methyl peroxide (CH3OOH) are reservoirs for the peroxy radicals.
One of the major challenges of recent years has been to develop a technique where the concentration of OH can be measured accurately. A few suitable techniques have been developed in the last few years, and confidence in the results they yield are now being met with some confidence (Eisele et al., 1994; 1996). As the development of these techniques has been a relatively recent event, these instruments are only just being integrated into atmospheric photochemistry studies. In the few studies that have combined measurements and modelling of OH, modelled results nearly always exceed those measured (for previous comparisons refer to Eisele et al., 1994, 1996; Poppe et al., 1994, and the most recently reported study of OH chemistry, the Tropospheric OH Photochemistry Experiment, TOHPE in the Journal of Geophysical Research, Volume 102, D5). The previous studies have demonstrated two important points. First, it is essential that as many of the ancillary measurements are made as possible, and that these are made simultaneously to measurements of OH. Second, the model used to investigate the OH radical must contain a sufficient description of the atmospheric chemistry.
In order to test our understanding of atmospheric processes, it is essential to construct models to help interpret the chemistry. These models must then be validated by comparison with field data. Due to the short lifetimes of the OH (~1sec) and the HO2 radicals (~150 sec), they react quickly to changes in local ambient conditions (solar flux, concentration of NOx and NMHC), but are unaffected by transport. In this case, it is sufficient to use a zero-dimensional model with no spatial resolution to describe the fast chemical processes (Poppe et al., 1994). In order to make the model meaningful, frequent measurements of ancillary data in the same air mass are required, and this was the case for two recent campaigns at Mace Head.
The Mace Head Atmospheric Research Station is situated on the west coast of Eire (53o19'34"N, 9o54'14"W, 10 m above sea level). Two campaigns have been run as part of the OXICOA (Oxidative Capacity of the Oceanic Atmosphere) component of the NERC funded ACSOE (Atmospheric Chemistry Studies of the Oceanic Environment) project. The site is usually subject to clean westerly air that has blown over the Atlantic, experiencing no fresh sources of pollution. The air tends to be characterised by low concentrations of NOx and background levels of O3, CH4 and CO are 30 ppb, 1.8 ppm and 100 ppb respectively (Simmonds et al., 1996). During the campaigns, concentrations of OH and HO2 were determined by the Leeds group using the technique of Fluorescence Assay Gas Expansion (FAGE). The concentrations of about 30 NMHC were also monitored by the Leeds group. Other measurementsmade during the campaign included peroxy radicals, NO, NO2, O3, HCHO, peroxides, CO, CH4, NO3, j(O1D), j(NO2), halogen oxides, HONO, CO2, CFCs, HNO3, SO2, NH3, CH3I, NOy, PAN, aerosol data, meteorological data and black carbon.
To construct an appropriate model for the marine boundary layer at Mace Head, the concentrations of hydrocarbons observed on each day were investigated. The product of the concentration of each hydrocarbon measured during the campaign and its rate coefficient for the reaction of OH was calculated. As expected, the majority of OH loss calculated in this way was due to reaction with CO and CH4 (between 65-90%). About 95% (slightly less on more polluted days) of OH loss could be accounted for by including the following NMHC along with CO and CH4: isoprene, trans-2-butene, cis-2-butene, ethane, ethene, propene, DMS, and 1,3-butadiene (click NMHC to see chart).
A model was therefore constructed using the chemical schemes for these hydrocarbons. In addition, a comprehensive inorganic chemistry scheme is also included (Saunders et al., 1997). Although there are limitations to this method in that only the NMHC measured during the campaign are considered, this will be offset partly by the exclusion of the reactions of OH with NO2, NO, O3 and H2 in the calculation above.
With the exception of the DMS mechanism, the chemical mechanisms and rate coefficients have been taken from a Master Chemical Mechanism (MCM) devised by Jenkin et al. (1997a). The MCM is believed to be the most comprehensive chemical mechanism in the world treating the degradation of 120 volatile organic compounds (VOC) that are emitted into the atmosphere (Jenkin et al., 1997b). The MCM has been constructed by considering the chemical kinetics data concerning rate coefficients, absorption cross-sections and reaction products, including several recent evaluations and reviews. The MCM is an explicit mechanism, and as such, does not suffer from the limitations of a lumped scheme, or one containing surrogate species to represent the chemistry of many species. Rate coefficients that are unknown are determined by extrapolating data from previously measured systems using structural activity relationships (Atkinson, 1987; Kwok and Atkinson, 1995). The MCM has now been launched on the WWW and can be found here.
The mechanism includes the oxidation of the 120 VOC by OH, and O3, as well as the chemistry of the subsequent oxidation products. These steps continue until carbon dioxide and water are yielded as final products of the oxidation. The DMS mechanism has been taken from the work of Yin et al. (1990a, 1990b) and updated using more recent work where possible. The advantage of this technique is that the reaction scheme can be adapted easily. For instance, modelling a polluted situation would require more NMHC schemes to be taken from the MCM in order to describe the chemistry.
The model contains a large number of distinct, often generic photochemical reactions involving inorganic species, aldehydes, ketones and organic hydroperoxides. The photolysis coefficients for these reactions have been calculated after the method described by Jenkin et al., 1997b, and take the following form:
J = l x cos Z ^ m * exp. (-n * sec Z)
where l, m and n are coefficients that have been determined for each photolysis reaction by fitting the j-values calculated using the two stream scattering model of Hough (1988) to the functional form given above. These coefficients have been calculated by considering a combination of the solar actinic flux, the quantum yield and the absorption cross-section between 200 and 700 nm. The zenith angle is calculated at each time-step during the model run, from a knowledge of the latitude, longitude and time of year. In this way, the photolysis coefficients can be calculated for any location, time of year and time of day. These values have been calculated assuming clear skies. Therefore, to properly represent the conditions in the atmosphere, the effect of clouds must be taken into account. This has been achieved by considering the difference between the model predicted maximum of j(O1D) with that observed in the atmosphere. The attenuation in the atmosphere gives an indication of the cloudiness, and this factor can be applied to the photolysis parameters in the model that are not determined experimentally.
The dry deposition terms have been taken from Derwent, 1996. It is assumed that all of the PAN species in the model are lost at the same rate as that of PAN. In order to account for the increased deposition that occurs at nighttime in the much smaller boundary layer, the boundary layer height is varied in the model. The boundary layer height starts at 300 m at 06:00 hours and then builds up gradually to 1300 m by 14:00 hours. This is maintained until early evening when the height of 300m is restored which remains constant until the next morning (Derwent et al., 1996).
The rate coefficient of heterogeneous loss on the surface of aerosols is described by the following equation:
k=1/4. c . A. g
where g is the accommodation coefficient, A is the surface area of aerosol (cm2 cm-3) and c is the mean speed (ms-1) of the molecules. c is defined by the expression (8kT/PI.M)1/2, where k=1.381 x 10-23 JK-1 (Boltzmann's constant), T is the temperature and M is the relative molecular mass in kg. The aerosol surface area in this expression has been measured at Mace Head. The largest uncertainty within this equation lies in the values of the accommodation coefficients used, which vary greatly depending on the surface and the temperature (see for example, DeMore et al., 1994). In this model, heterogeneous loss of the following species are included N2O5, NO3, HO2, CH3O2, OH, CH3SO3H and HNO2. Once the species are lost on the aerosol surface, they are considered to take no further part in the chemistry.
The Leeds box model uses actual observations to predict the concentration of OH at Mace Head. The model is constrained by using observations of NO2, NO, O3, CO, CH4, NMHC, PAN, SO2, aerosol surface area, j(O1D), j(NO2), temperature and relative humidity. The data were available at various frequencies depending on the integration times of the various instruments. Available most frequently were measurements such as O3 and NOx each minute, but hydrocarbon measurements were made once an hour. For each day modelled, the minute data was averaged to 15-minute averages. For the hydrocarbon data, 15-minute averages were obtained by interpolating between points.
The model runs using FACSIMILE code (Curtis and Sweetenham, 1987). This is a Gear's method that integrates the differential equations numerically using stiffly stable techniques. At the beginning of each time-step, the solution is estimated using a polynomial fit, and this is improved upon by using a Newton-type iteration 'predictor-corrector' method. A 20 hour integration of the 1670 reactions takes about 10minutes on a high specification PC.
Preliminary results from both campaigns show a reasonable degree of agreement between the model and measurements, for OH, HO2 and peroxy radicals. A more detailed description of the model results from the first campaign has been submitted to J. Geophys. Res., November 1998 (Carslaw et al., 1998a)
For details on trajectory modelling studies of the Mace Head campaigns, check out the Centre for Atmospheric Sciences at the University of Cambridge.
Mechanism reduction
Introduction
Carslaw et al. (1998a) described a technique for constructing model mechanisms and presented results from the EASE96 (Eastern Atlantic Summer Experiment 96) campaign that took place on the west coast of Ireland in 1996. The initial results were encouraging and reasonable agreement was obtained with in situ measurements. Although the mechanism described by Carslaw et al. (1998a) is much smaller than the master chemical mechanism from which it is derived (1666 reactions compared with 7500), it is still too large to gain an understanding of which reactions are driving the chemistry. Mechanism reduction techniques have been applied to atmospheric field data for the first time. The full mechanism has been reduced according to two different criteria. First, the mechanism is reduced so that the model concentrations of OH and HO2 are within 5% of those predicted by the full mechanism. Second, the mechanism is reduced so that the model OH and HO2 concentrations are within 20% of those predicted by the full mechanism. We have selected a clean day (J216) and a semi-polluted day (J199) on which to carry out these tests. These two days have been described in detail by Carslaw et al. (1998a).
Mechanism reduction is ideal for the situation where a large general mechanism is to be applied to specific conditions, in this case to clean or semi-polluted conditions at Mace Head. Reduction of the mechanism involves the identification of a smaller subset of reactions, which can still accurately describe the chemistry to within a prescribed accuracy. By reducing a mechanism, it is possible to investigate species interactions and coupling between reactions in more detail. The reduced mechanism is also less computationally expensive. Full details of the mathematics behind mechanism reduction can be found in Zeng et al. (1997).
Basically, mechanism reduction consists of 5 steps:
1. Selection of important species (here, OH and HO2)
2. Identification and removal of redundant species
3. Identification and removal of redundant reactions of important and necessary species
4. Removal of fast reversible reactions
5. Application of quasi-steady state assumption (QSSA) to express concentration of some species algebraically
This procedure has to be carried out at several different time points throughout the profile to make sure that the species deemed important remain so throughout the day. If a species is found to be redundant at all time points then it can be eliminated from the mechanism along with its consuming reactions. At all stages of reduction, the reduced mechanism needs to be compared with the full mechanism to see if the difference in the important species concentration is within the prescribed accuracy. The software (kinalc) used to carry out the mechanism reduction is available on the World Wide Web kinalc
Results and Discussion
J216-clean day
The clean day reduction was carried out for the data obtained on J216 during the EASE96 campaign at Mace Head due to the high level of agreement between the model and measurements (Carslaw et al., 1998a). The full mechanism consisted of 1666 reactions and 505 species. The reduction was carried out at nine different time points over the day at 02:00, 06:00, 10:00, 11:00, 12:00, 13:00, 14:00, 18:00 and 22:00 hours. A species was included in the mechanism if it was deemed necessary at any of these time points. Two reductions were carried out for each scenario. The purpose of the first reduction was to retain a high degree of accuracy throughout the entire day for the OH and HO2 concentrations, to within 5% of the full mechanism. Hereafter this will be referred to as the reduced mechanism. The second reduction scenario aimed to sacrifice some accuracy in order to reduce the model further and get to the heart of the mechanism in terms of which reactions were driving the chemistry. The prescribed accuracy in this case was 20 % during daylight (09:00-17:00) hours. This caveat is added as the more extensive reduction eliminates much of the NO3 chemistry that drives nighttime chemistry and hence the errors during this period are very large. However, as the designated important species are OH and HO2, for the purpose of this exercise a focus on daytime chemistry is acceptable. The result of this reduction will be referred to as the mini mechanism. The reduced mecahnsims are shown in Table 2.
Table 2 - Summary of reduction conditions on J216 (clean day)
| Mechanism | Number of reactions | Number of species |
| Full | 1666 | 505 |
| Reduced | 894 | 249 |
| Mini | 25 | 17 |
J199-semi-polluted day
J199 was selected for reduction for two reasons. First, reasonable agreement was obtained between the full model and in situ measurements for both HO2 and the sum of peroxy radicals (Carslaw et al., 1998a). Second, the relatively high concentration of isoprene suggested that this day would show large deviations from clean air chemistry. Again, the full mechanism consisted of 1666 reactions and 505 species. The reduction was carried out at eight different time points over the day that covered the time period over which NOx data was available. The time points were at 10:00, 11:00, 12:00, 12:30, 13:00, 14:00, 15:00 and 16:00 hours. The reduction details are shown in Table 3.
Table 3- Summary of reduction conditions on J199
| Mechanism | Number of reactions | Number of species |
| Full | 1666 | 505 |
| Reduced | 986 | 279 |
| Mini | 64 | 39 |
Both sets of reduced mechanisms have been shown to be generally applicable to similar conditions during a campaign held at the same site the following year (see Carslaw et al., 1998b). Conclusions For both a clean and semi-polluted air mass during the EASE96 campaign we have provided reduced mechanism for predicting the concentrations of OH and HO2. This is possible using 64 reactions for a semi-polluted day, and 25 reactions for a clean day, and still maintaining an accuracy of 20% when compared to the full mechanism.
The mechansims described above are available here:
Reduced (clean, J216) and (semi-polluted, J199)
Mini (clean, J216) and (semi-polluted, J199)
Atkinson R., A structure-activity relationship for the estimation of
rate constants for the gas-phase reactions of OH radicals with organic
compounds, Int. J. Chem. Kinetics, 19, 799-828, 1987.
Carslaw N., D.J. Creasey, D.E. Heard, A.C. Lewis, J.B. McQuaid, M.J. Pilling,
P.S. Monks, B.J. Bandy and S.A. Penkett, Modelling OH, HO2 and
RO2 radicals in the marine boundary layer: 1. Model construction
and comparison with measurements submitted to J. Geophys. Res.,
1998a.
Carslaw N., P.J. Jacobs and M.J. Pilling, , Modelling OH, HO2
and RO2 radicals in the marine boundary layer: 2. Mechanism
reduction and uncertainty analysis, submitted to J. Geophys. Res.,
1998b.
Curtis A.R. and W.P. Sweetenham, FACSIMILE/CHEKMAT users manual. AERER12805.
London: HMSO, 1987.
DeMore, W.B., S.P. Sander, D.M. Golden, R.F. Hampson, M.J. Kurylo, C.J.
Howard, A.R. Ravishankara, C.E. Kolb, and M.J. Molina, Chemical Kinetics
and Photochemical Data for Use in Stratospheric Modelling: Evaluation Number
11, Jet Propulsion Laboratory, California, 1994. Derwent R.G., The influence
of human activities on the distribution of hydroxyl radicals in the troposphere,
Phil. Trans. R. Soc. Lond. A, 354, 501-531, 1996.
Derwent R.G., M.E. Jenkin, and S.M. Saunders, Photochemical ozone creation
potentials for a large number of reactive hydrocarbons under European conditions,
Atmos. Environ., 30, 189-200, 1996.
Eisele F.L., G.H. Mount, F.C. Fehsenfeld, J. Harder, E. Marovich, D.D.
Parrish, J. Roberts and M. Trainer, Intercomparison of tropospheric OH
and ancillary trace gas measurements at Fritz Peak Observatory, Colorado,
J. Geophys. Res., 99, 18605-18626, 1994.
Eisele F.L., D.J. Tanner, C.A. Cantrell and J.G. Calvert, Measurements
and steady state calculations of OH concentrations at Mauna Loa Observatory,
J. Geophys. Res., 101, 14665-14679, 1996.
Hough A.M., The calculation of photolysis rates for use in global tropospheric
modelling studies. AERE Report R-13259 (HMSO) London, 1988.
Jenkin M.E., S.M. Saunders, and M.J. Pilling, The tropospheric degredation
of volatile organic compounds: a protocol for mechanism development, Atmos.
Environ., 31, 81-104, 1997a.
Jenkin M.E., G.D. Hayman, R.G. Derwent, S.M. Saunders and M.J. Pilling,
tropospheric Chemistry Modelling: Improvements to current models and application
to policy issues, AEA/RAMP/20150/R001, 1997b.
Kwok E.S.C. and R. Atkinson, Estimation of hydroxyl radical reaction rate
constants for gas-phase organic compounds using a structure-reactivity
relationship: an update, Atmos. Environ., 29, 1685-1695, 1995.
Poppe D., J. Zimmermann, R. Bauer, T. Brauers, D. Brüning, J. Callies,
H.-P. Dorn, A. Hofzumahaus, F.-J. Johnen, A. Khedim, H. Koch, R. Koppmann,
H. London, K.-P. Müller, R. Neuroth, C. Plass-Dülmer, U. Platt,
F. Rohrer, E.-P. Röth, J. Rudolph, U. Schmidt, M. Wallasch, and D.H.
Ehhalt, Comparison of measured OH concentrations with model calculations,
J. Geophys. Res., 99, 16633-16642, 1994.
Saunders S.M., M. Jenkin, R.G. Derwent and M.J. Pilling, World wide web
site of a master chemical mechanism (MCM) for use in tropospheric chemical
models, Atmos. Environ., 31, 1249-1250, 1997.
Simmonds P.G., R.G. Derwent, A. McCulloch, S.O'Doherty and A. Gaudry, Long-term
trends in concentrations of halocarbons and radiatively active trace gases
in Atlantic and European air masses monitored at Mace Head, Ireland from
1987-1994, Atmos. Environ., 30, 4041-4063, 1996.
Turanyi T., T. Berces and S. Vajda, Reaction rate analysis of complex kinetic
systems, Int. J. Chem. Kin., 21, 83-99, 1989.
Turanyi T., Reduction of large reaction mechanisms, New. J. Chem.,
14, 795-803, 1990.
Yin F., D. Grosjean and J.H. Seinfeld, Photooxidation of Dimethyl Sulfide
and Dimethyl Disulfide: I Mechanism Development, J. Atmos. Chem.,
11, 309-364, 1990a.
Yin F., D. Grosjean and J.H. Seinfeld, Photooxidation of Dimethyl Sulfide
and Dimethyl Disulfide: 2 Mechanism Evaluation, J.Atmos. Chem.,
11, 365-399, 1990b.
Zeng G., M.J. Pilling and S.M. Saunders, Mechanism reduction for tropospheric
chemistry: butane oxidation, J. Chem. Soc, Faraday Trans., 93, 2937-2946,
1997.
