Article pubs.acs.org/est
Multiannual Observations of Acetone, Methanol, and Acetaldehyde in Remote Tropical Atlantic Air: Implications for Atmospheric OVOC Budgets and Oxidative Capacity K. A. Read,† L. J. Carpenter,*,† S. R. Arnold,‡ R. Beale,§ P. D. Nightingale,§ J. R. Hopkins,† A. C. Lewis,† J. D. Lee,† L. Mendes,∥ and S. J. Pickering‡ †
National Centre for Atmospheric Science, University of York, York, YO10 5DD, U.K. Institute for Climate & Atmospheric Science, School of Earth & Environment, University of Leeds, LS2 9JT, U.K. § Plymouth Marine Laboratory, Prospect Place, Plymouth, Devon, PL13DH, U.K. ∥ Instituto Nacional de Meteorologia e Geofísica (INMG), Mindelo, Sao Vicente, Cape Verde ‡
S Supporting Information *
ABSTRACT: Oxygenated volatile organic compounds (OVOCs) in the atmosphere are precursors to peroxy acetyl nitrate (PAN), affect the tropospheric ozone budget, and in the remote marine environment represent a significant sink of the hydroxyl radical (OH). The sparse observational database for these compounds, particularly in the tropics, contributes to a high uncertainty in their emissions and atmospheric significance. Here, we show measurements of acetone, methanol, and acetaldehyde in the tropical remote marine boundary layer made between October 2006 and September 2011 at the Cape Verde Atmospheric Observatory (CVAO) (16.85° N, 24.87° W). Mean mixing ratios of acetone, methanol, and acetaldehyde were 546 ± 295 pptv, 742 ± 419 pptv, and 428 ± 190 pptv, respectively, averaged from approximately hourly values over this five-year period. The CAM-Chem global chemical transport model reproduced annual average acetone concentrations well (21% overestimation) but underestimated levels by a factor of 2 in autumn and overestimated concentrations in winter. Annual average concentrations of acetaldehyde were underestimated by a factor of 10, rising to a factor of 40 in summer, and methanol was underestimated on average by a factor of 2, peaking to over a factor of 4 in spring. The model predicted summer minima in acetaldehyde and acetone, which were not apparent in the observations. CAM-Chem was adapted to include a two-way sea−air flux parametrization based on seawater measurements made in the Atlantic Ocean, and the resultant fluxes suggest that the tropical Atlantic region is a net sink for acetone but a net source for methanol and acetaldehyde. Inclusion of the ocean fluxes resulted in good model simulations of monthly averaged methanol levels although still with a 3-fold underestimation in acetaldehyde. Wintertime acetone levels were better simulated, but the observed autumn levels were more severely underestimated than in the standard model. We suggest that the latter may be caused by underestimated terrestrial biogenic African primary and/or secondary OVOC sources by the model. The model underestimation of acetaldehyde concentrations all year round implies a consistent significant missing source, potentially from secondary chemistry of higher alkanes produced biogenically from plants or from the ocean. We estimate that low model bias in OVOC abundances in the remote tropical marine atmosphere may result in up to 8% underestimation of the global methane lifetime due to missing model OH reactivity. Underestimation of acetaldehyde concentrations is responsible for the bulk (∼70%) of this missing reactivity.
1. INTRODUCTION
A wide variety of primary and secondary OVOC sources means that their apportionment is complex. Primary emissions include those from the terrestrial biosphere (e.g., emission from vegetation8,9 and plant matter decay10,11), anthropogenic sources (e.g., fossil fuel12 and solvent usage13), biomass burning (BB),14−16 and the oceans.17−21 Plant-derived emissions are
Oxygenated volatile organic compounds (OVOCs) are ubiquitous in the troposphere (see refs 1−3 and references therein). In the marine boundary layer, they have a comparable if not greater effect than nonmethane hydrocarbons (NMHC) on the oxidizing capacity of the lower atmosphere through reaction with the OH radical.3,4 Photochemical breakdown of OVOCs in the upper troposphere is an important source of hydrogen oxide radicals (HOx = OH + HO2) and also produces organic radicals that can form peroxy acetyl nitrate (PAN), thus affecting the tropospheric ozone budget.5−7 © 2012 American Chemical Society
Received: Revised: Accepted: Published: 11028
May 30, 2012 August 30, 2012 September 10, 2012 September 10, 2012 dx.doi.org/10.1021/es302082p | Environ. Sci. Technol. 2012, 46, 11028−11039
Environmental Science & Technology
Article
thought to be the largest single source of acetone.19,20 Terrestrial vegetation is also the major net source of methanol, mainly via plant growth although with a smaller source from plant decay.6,17,18,22,23 Biogenic emissions of OVOCs show a strong light and temperature dependence similar to that of isoprene, with emissions maximizing in summer, but with higher variability across different plant types.24,25 The bottomup uncertainty of terrestrial emissions could be more than 50%.21 Secondary production from the oxidation of NMHCs is also a major OVOC source, particularly of acetaldehyde,21 allowing substantial concentrations to be present several days downstream from primary VOC emissions.3 The complexity of secondary production of OVOCs, with precursors including a multitude of >C1 alkanes and >C2 alkenes as well as higher OVOC, creates a large uncertainty in model predictions.3,6,26,27 In the remote marine atmosphere, oceanic sources and sinks are expected to play an important role in OVOC budgets; however, both the magnitude and direction of OVOC air−sea fluxes are a matter of debate,23,28−34 primarily because there are very few OVOC measurements in the remote marine atmosphere and fewer still in seawater. The dominant atmospheric sink for acetone, methanol, and acetaldehyde is through reaction with OH. Acetaldehyde and acetone are also readily photolyzed, particularly in the upper troposphere where photolysis loss of acetone exceeds OH loss.35,36 Estimates of wet and dry deposition to land are highly variable.13,37−39 Methanol may also be removed in precipitation and by in-cloud chemistry via oxidation by OH22 or through a heterogeneous sink such as uptake by sulfuric acid,40 ice41 or mineral oxide surfaces,42 although recent studies suggest that the heterogeneous sink is negligibly small.17,18 Estimated atmospheric lifetimes are around 0.8 days for acetaldehyde,21 9 days for methanol,23 and 35 days for acetone.36 In this paper, we present the first multiannual measurements of OVOCs in the remote marine boundary layer. The atmospheric measurements comprise part of the ongoing monitoring at the Global Atmospheric Watch (GAW) Cape Verde Atmospheric Observatory (CVAO) (16.848° N, 24.871° W). We utilize the five-year OVOC atmospheric time-series in conjunction with the global atmospheric chemical-transport model CAM-Chem to evaluate relative sources and sinks of the OVOCs and how these vary intra-annually. Oceanic OVOC fluxes are introduced into the model using a parametrization based on seawater measurements made in the subtropical Atlantic and in the northwest African upwelling region.
steel for the last 3 m to the instrument. No losses of OVOCs to the surfaces of the inlet tubing are observed. Water was removed from the sample using a condensation trap held at −27 °C within a bath of ethylene glycol/water; this was experimentally determined to produce no measurable losses of OVOC during sampling. OVOCs were preconcentrated onto a multiadsorbent bed Peltier-cooled to −18 °C and then desorbed into a helium flow by rapid heating (16 °C s−1) to 350 °C. The preconcentration trap was packed with Carbopack B and Carboxen 1000 in series, a combination not thought to be affected by O3 interferences at levels below about 90 ppbv.44 Average ambient levels of O3 observed at CVAO are 31.3 ± 8.1 ppbv; these levels are likely to be reduced substantially during air flow through the stainless lines and the water trap. The samples are separated using a dual column gas chromatograph with two flame ionization detectors (dc-GC-FID) (Agilent 6890N, Agilent Technologies, Wilmington, DE, U.S.A.). The instrument used two parallel capillary columns that simultaneously resolved C2−C8 NMHC (Al2O3 PLOT column, 0.53 μm, 50 m) and OVOC including acetaldehyde, methanol, and acetone (2 lengths in series of a more polar LOWOX column of 0.53 μm i.d. and 10 m length). Calibrations for acetone, methanol, and acetaldehyde were performed using permeation sources (KIN-TEK Trace Source) and a gaseous OVOC standard (ppbv level, Apel-Riemer Environmental Inc., Denver, CO, U.S.A.). Further details are given in the Supporting Information. 2.2. Oceanic OVOC Measurements in the Atlantic Ocean. Two oceanic data sets were used to provide the model with seawater concentrations of acetone, methanol, and acetaldehyde. Seawater measurements made during an Atlantic Meridional Transect (AMT) cruise from the U.K. to Chile (49° N to 39° S) in October−December 200945 allowed comparison of different Atlantic regions that are subject to varying levels of productivity, microbial species composition, and light intensity. The second cruise, also in the Atlantic Ocean, was off the Mauritanian Coast, Northwest Africa, in a region of natural upwelling (April−May 2009). 45 Although the latitudes encountered during this cruise (19−22° N) were also sampled during AMT, the upwelling cruise represented a contrast to the predominantly oligotrophic water of the northern Atlantic gyre. All samples were from the surface ocean (5 m) and were made using conductivity temperature depth (CTD) casts. Samples were drawn from a Niskin bottle via Tygon tubing into brown glass sample bottles. The measurement of OVOCs in seawater was achieved via the coupling of a membrane inlet system to a proton transfer reaction/mass spectrometer (MI-PTR/MS).46 The membrane is an interface separating the seawater sample from a precleaned flow of nitrogen gas. OVOCs are able to permeate through the membrane into the gas, which flows directly into the drift tube of the PTR/MS. OVOCs are ionized by hydronium ions (H 3 O + ) and detected at masses 33 (methanol), 45 (acetaldehyde), and 59 (acetone). Calibration is achieved with water standards prepared via solvent addition to seawater with subsequent serial dilution. Although interference to these compounds is suspected to be low via MI-PTR/MS,46 we cannot rule out the possibility of more than one compound being attributed to one mass, particularly for acetone, which is isomeric with propanal with an ionized mass of 59. Therefore, the OVOC oceanic concentrations used in the CAM-Chem model should be viewed as upper limits. Limits of detection are estimated as 27 nM, 0.7 nM, and 0.3 nM for methanol,
2. EXPERIMENTAL SECTION 2.1. Atmospheric Measurements of OVOC at the CVAO. The CVAO is located on a northeast facing lava field at Calhau on the volcanic island of São Vicente (16° 51′ 49 N, 24° 52′ 02 W), 50 m from the coastline and 10 m above sea level, with the northeasterly prevailing trade winds blowing directly off the ocean (97% of the air arrives from this sector). There are no obvious major coastal features such as seaweed beds, and air masses arriving at the site have been transported for a minimum of 2 days across the open ocean. Characteristics of the air masses and trajectories arriving at CVAO are described in detail in Carpenter et al.43 Hourly atmospheric OVOC measurements at the CVAO are made from a height of 10 m. The air sample is pulled at 55 L min−1 through 15 m of a heated sampling manifold (1″ glass: until July 2010 this was made of stainless steel, with a residence time of 4.5 s), then at 100 mL min−1 through 1/8″ stainless 11029
dx.doi.org/10.1021/es302082p | Environ. Sci. Technol. 2012, 46, 11028−11039
Environmental Science & Technology
Article
OVOCs at CVAO, we simulated the two-way exchange of acetone, acetaldehyde, and methanol between the ocean and the atmospheric model surface layer using the two-layer film model approach.53 Ocean surface layer concentrations of the OVOCs were taken from observations and applied globally to the oceans in eight latitude bands, summarized in Table 2. These concentrations were assumed to be invariant with time of year. Air-side (ka)54 and water-side (kw)55 transport velocities are calculated using model surface layer wind speed scaled to 10 m. Schmidt numbers for each OVOC are calculated using the kinematic viscosity of water and temperature-dependent diffusivities of each gas in water. 56 Henry’s Law data for the three OVOCs57 are corrected for temperature using model sea surface temperature. Bidirectional fluxes for the OVOCs dependent on oceanic and atmospheric concentrations, seasurface temperature, and wind speed are calculated for each oceanic model grid square and implemented in the model along with the standard emission and deposition terms. Net global ocean fluxes of the OVOCs calculated by the model are shown in Table S1 of the Supporting Information.
acetaldehyde, and acetone respectively, and precision for all three species is 800 km from the African continent. However, acetone and methanol show similar values in marine-influenced air masses to the Wisthaler et al.59 (early March) study. Acetaldehyde levels at Cape Verde (and similarly measured in situ at Mace Head) are significantly higher than those reported by Wisthaler et al. 59 in the Indian Ocean and by Singh et al. 2 in the Pacific Ocean. One potential reason is high photoproduction of acetaldehyde in the biologically active oceans of the west African coast, as computed by Millet et al.21 based on the results of Kieber et al.60 Figure 2 shows the annually averaged measured (mean and 90th and 10th percentiles) and modeled monthly OVOC concentrations, as well as the average % time spent by 10-day back-trajectories arriving at Cape Verde in a given geographical
Table 1. Details of the Model Sensitivity Runs model run NOBIO NOANTH NOFIRE
description No biogenic emissions of acetone, acetaldehyde, methanol, ethane, propane, ethene, propene, isoprene, and monoterpenes. No anthropogenic emissions of methanol, acetaldehyde, acetone, >C2 alkanes, >C2 alkenes, ethanol, methyl ethyl ketone. No biomass burning emissions of methanol, acetaldehyde, acetone, >C2 alkanes, >C2 alkenes, ethanol, methyl ethyl ketone.
2.4. Air−Sea Exchange Parametrization. To assess the impact of possible sea−air fluxes on the abundance of the
Table 2. Summary of Latitudinal Distribution of Measured Atlantic Seawater OVOC Concentrations (See Section 2.2) Used to Drive the Model Ocean Flux Parameterization cmpd
−40° to −10° (nM in seawater)
no. samples
−10° to 10° (nM in seawater)
no. samples
10° to 30° (nM in seawater)
no. samples
30° to 50° (nM in seawater)
no. samples
acetone methanol acetaldehyde
7 121 5
12 12 12
5 137 5
5 5 5
14 237 5
13 13 13
9 128 6
8 8 8
11030
dx.doi.org/10.1021/es302082p | Environ. Sci. Technol. 2012, 46, 11028−11039
Environmental Science & Technology
Article
Figure 1. Daily CAM-Chem model output (black) and daily observed (color) time-series of (a) acetone, (b) methanol, (c) acetaldehyde, (d) carbon monoxide, (e) ethane, and (f) propane.
Table 3. Measurements of Acetone, Methanol, and Acetaldehyde in the Marine Atmosphere acetone mean ±1σ (range) pptv 530 ± 200 (237−486)a 500 ± 205 (163−1670) 466 ± 97 624 ± 63 (450−2400) 400 ± 150 (200−1000) 360 ± 51 546 ± 295 (232−909)b a
methanol mean ±1σ (range) pptv
acetaldehyde mean ±1σ (range) pptv
890 ± 400 805 ± 176 (469−1553) 575 ± 211 708 ± 69 (500−1600)
(106−240)a 435 ± 190 (120−2120) 204 ± 40 212 ± 29 (125−500) 570 ± 280 (200−1400)
742 ± 419 (305−1327)b
428 ± 190 (191−671)b
time period Oct.−Nov. 2002 Oct. 93−Apr. 95 July−Aug. 2002 winter−spring 2001 March 1999 Oct. 1988 May−July 2004 annual 2006−2011
location tropical Atlantic (ship) Mace Head (canisters) Mace Head Pacific Ocean (ship) Indian Ocean (ship) Caribbean Sea (ship, canisters) western tropical Pacific (ship) Cape Verde Atmospheric Observatory (CVAO)
ref 29 78 3 2 59 79 30 this work
Monthly mean values. b10th−90th percentile range.
sector43 and the cumulative leaf area index (LAI). Acetone and methanol have winter minima, with methanol showing a slight maxima in spring and autumn and acetone showing quite a pronounced autumn (September) peak. Acetaldehyde does not show a distinct seasonal cycle. None of these features are well captured by the model. In spring, transport of North American air over the North Atlantic to Cape Verde reaches its seasonal peak (see Figure 2a), so primary (mainly biogenic) sources from the eastern U.S. influence the site during this period. The September maximum (particularly pronounced in 2009) may
be caused by African biogenic emissions; this is discussed in more detail later. 3.2. Model-Measurement Comparisons. Episodes of very high OVOC mixing ratios are only partly captured by CAM-Chem (Figure 1). This is, in part, likely due to the large grid resolution of the model, which dilutes plumes associated with transport from continental sources (Europe or North America). However, Figure 2 shows that there are also substantial differences between the monthly averaged Cape Verde OVOC observations and the CAM-Chem modeled 11031
dx.doi.org/10.1021/es302082p | Environ. Sci. Technol. 2012, 46, 11028−11039
Environmental Science & Technology
Article
(annual averages); however, it underestimates the observed autumn peak (July−September) and overestimates the winter/ spring data (November−April). Observed concentrations of methanol are a factor of 2.2 (annual averages) larger than those from the model and 4.5 times higher in March. Production of methanol is driven mainly by OH-oxidation of methane. Increasing the model methane concentrations by 3% leads to a maximum increase in methanol concentrations of 15 pptv (in September), consistent with calculations1 that up to 500 pptv of methanol could come from methane oxidation. Jacob et al.18 suggested that methanol production from methane oxidation may be underestimated by up to a factor of 2, due to uncertainties in the rate constant for the methyl peroxy radical (CH3O2) self-reaction and the corresponding yield of methanol. Increasing the methanol source from methane brings the “standard” model into closer agreement with the measurements; however, as discussed later, there are alternative explanations for this discrepancy. VOCs other than methane, including acetaldehyde, are also sources of CH3O2 and thus potential sources of uncertainty in methanol predictions. The model substantially underestimates the acetaldehyde mixing ratios observed at the CVAO by a factor of 10 averaged over the year. The simulations show a strong seasonal variation with acetaldehyde mixing ratios dropping to
