Main

Tropospheric ozone is an important greenhouse gas in addition to its influence on air quality and public health, on the photochemical processing of atmospheric chemicals, and on food security and ecosystem viability. It is produced through the catalytic oxidation of carbon compounds in the presence of nitrogen oxides (NO x = NO + NO2), and has an additional smaller source from stratospheric influx of ozone into the free troposphere8. Ozone is lost to the surface through deposition and can be destroyed throughout the atmosphere by photochemical processes, predominantly by photolysis and the subsequent reaction of electronically excited oxygen atoms with water vapour. Whether a particular air mass is producing or destroying ozone depends broadly on the short-wave radiation environment and the water vapour and NO x concentrations. Thus, ozone is formed predominantly in continental regions where there are sources of NO x and is typically lost in marine regions where sources are small9. Because of its high water vapour content, high solar radiation levels and large geographical extent, the tropical marine boundary layer is the most important global region for loss of ozone10. However, surface atmospheric observations in this region are extremely sparse.

The Cape Verde archipelago lies within the tropical Eastern North Atlantic Ocean. The archipelago is volcanic in origin and the island shores shelve steeply to the deep abyssal plain beyond the African continental shelf. An Ocean–Atmosphere Observatory has been newly established on the northeastern side of the island of São Vicente within the Cape Verde archipelago. The atmospheric site (16.85° N, 24.87° W, Supplementary Fig. 1) receives the prevailing northeasterly trade winds directly off the ocean for around 95% of the time. In contrast to many other atmospheric monitoring stations in the Northern Hemisphere, there are no seaweed beds or other local coastal sources. It can therefore be assumed to be representative of the surrounding open-ocean marine boundary layer. The ocean surrounding Cape Verde is in general biologically productive because of both Saharan dust input11 and proximity to the northwest African coastal upwelling system, which lies a few hundred kilometres to the northeast.

A year of observations from the Observatory has now been obtained (October 2006 to October 2007). These include measurements of O3, H2O, CO, NO, NO2, CH4, volatile organic compounds (VOCs), and oxygenated VOCs, dimethyl sulphide, the halogen oxide radicals BrO, IO and OIO, the ozone photolysis rate coefficient J(O1D), broadband ultraviolet radiation, wind speed and direction (see Supplementary Information). During the summer of 2007, a research aircraft measured composition over the site to assess the representativeness of the overlying boundary layer and to determine any diurnal variability in boundary layer depth (see Supplementary Information).

The annual cycle in ozone (see Supplementary Fig. 2) displays a maximum in spring and a minimum in late summer, consistent with other coastal sites12,13. The daily cycle shows a loss during the day and a recovery at night (due to entrainment from the free troposphere) with an annually averaged loss of 3.3 ± 2.6 p.p.b.v. (parts per 109 by volume) per day between 09:00 and 17:00 h ut (local plus 1 h). The aircraft observations in June 2007 confirm the vertical extent of the ozone loss throughout the boundary layer (Fig. 1). This agreement between surface and aircraft-determined ozone concentration in the boundary layer is typical and was observed on 12 further flights not shown here.

Figure 1: Example data from aircraft observations 1–20 km upwind of São Vicente.
figure 1

a, Vertical temperature profiles obtained at 09:00, 13:00 and 18:30 ut on 27 May 2007 showing the constant inversion height. b, Mean ozone measured at four vertical levels from the aircraft (symbols, 10-min averages) and at the Cape Verde Observatory (solid line, 20-min running mean) on 27 May 2007. The 1σ standard deviations of the mean of the aircraft ozone data from 200 m are shown as an example of the precision of the data (see Supplementary Information). c, Temperature and CO on 27 May 2007.

Full size image

The monthly averaged daytime (09:00–17:00 ut) concentrations of ozone and NO observed at Cape Verde between October 2006 and October 2007 were compared with simulations using the global tropospheric chemistry transport model GEOS-CHEM14. The measured NO mixing ratios were extremely low throughout the year (09:00–17:00 ut average 3.0 ± 1.0 p.p.t.v. (1σ) where p.p.t.v. is parts per 1012 by volume) and showed broad agreement with the GEOS-CHEM simulated levels (09:00–17:00 ut average 3.2 ± 1.3 p.p.t.v.; Fig. 2). However, the ozone observations show a significantly greater monthly averaged daytime depletion of ozone (the difference between concentrations at 09:00 and 17:00 ut, referred to hereafter as ΔO3) and a more exaggerated seasonal cycle in ΔO3 than the model simulations. The ozone budget in GEOS-CHEM includes NO x -catalysed ozone production and odd-hydrogen photochemical ozone loss, in addition to advection, convection and deposition terms. During the period of maximum solar activity, GEOS-CHEM computes a maximum net ozone loss of ∼3 p.p.b.v. d-1 (up to 8% of the mean GEOS-CHEM ozone concentration) compared with the observed losses of ∼5 p.p.b.v. d-1 (up to 13% of the mean measured ozone concentration) (Fig. 2). Similar results are obtained from box model simulations (see Supplementary Information) of the O x –HO x –NO x system, constrained using the observations.

Figure 2: Measurements and modelling results.
figure 2

a, Monthly averages of observed and modelled (GEOS-CHEM) NO mixing ratios; b, monthly averages of observed ΔO3 over 8 h (09:00–17:00 ut) compared with predictions from GEOS-CHEM and from the box model with and without halogen chemistry.

Full size image

Because of slowly changing sea surface temperatures and continuous strong winds, the marine boundary layer around Cape Verde and the Observatory is subject to minimal diurnal variations in dynamics such as boundary layer depth, which can complicate interpretations and modelling of changes in O3 concentrations. The surface temperature varies by no more than ±0.5 °C throughout the day and drops by only 1 °C at night. Aircraft measurements made on four days, each day with flights typically at 09:00, 13:00 and 18:30 ut, confirm negligible changes in the inversion depth over any given day (Fig. 1). Other key parameters for accurately simulating the daily ozone variation in low NO x regions are the entrainment rate—the largest ozone budget term in this region—and the photolytic destruction rate. For the box model simulations, the observed average monthly nocturnal increase in ozone at Cape Verde was assumed to comprise the difference between the entrainment and deposition terms over the month. This parameter demonstrated a seasonal variation with a minimum in November (0.18 p.p.b.v. h-1) and a maximum in April (0.48 p.p.b.v. h-1). Hourly GEOS-CHEM values were used for the ozone photolysis rate coefficient J(O1D). These are calculated using Fast-J code15 which uses the cloud fields from the Goddard Earth Observing System of the NASA Global Modelling and Assimilation Office (GMAO), and ozone columns derived from satellite observations. The calculated monthly averaged peak values of J(O1D) agreed with data obtained in January–February and May–June 2007 to within 13%.

That both modelling approaches significantly underestimate the observed daily ozone loss is clearly evidence for an important missing loss process. A widespread effect of halogens on tropospheric oxidants in the marine boundary layer has been suggested by a number of theoretical4,5,7 and observational16,17 studies; however, these so far remain unconfirmed predominantly because of a lack of observations of atmospheric species representative of the pristine marine boundary layer18,19. Ozone is destroyed directly via catalytic cycles with the rate determined by the self-reactions of XO (where XO = IO, BrO) and the reaction of XO with HO2 (reactions (1)–(5)). A reduction in ozone production also occurs through a decrease of the HO2/OH ratio as a result of reactions (4) and (5), and the suppression of NO x caused by the hydrolysis of halogen nitrates on aerosol surfaces20.

Although halogen oxides also cause a decrease in the NO/NO2 ratio through reaction (6), the subsequent formation of NO2 does not necessarily lead to an increase in ozone concentrations because a halogen atom is also formed which will destroy ozone through reaction (1). In addition to its effect on ozone and OH, it has been proposed that BrO causes marked changes in dimethyl sulphide levels and oxidation pathways, reducing its cooling effect on climate5,21.

Halogen oxide measurements by differential optical absorption spectroscopy (Supplementary Information) at the Cape Verde Observatory from November 2006 until June 2007 reveal the presence in daytime of IO and BrO radicals with mean daytime maxima of 1.4 ± 0.8 (1σ) p.p.t.v. and 2.5 ± 1.1 (1σ) p.p.t.v., respectively (Fig. 3). Both radicals show diurnal cycles that seem to be dependent on solar radiation, with mixing ratios below the detection limits (BrO: 0.5–1 p.p.t.v., IO: 0.3–0.5 p.p.t.v.) at night. On average, slightly higher average values are apparent in spring; but the variability of the observations does not allow for a robust conclusion on seasonal halogen oxide trends. Models for the acid-catalysed activation of bromine from sea salt aerosol predict a marine boundary layer BrO concentration of about 1–4 p.p.t.v. with a diurnal cycle similar to a top-hat distribution4,22. Iodine oxide levels of about 1 p.p.t.v. with a similar diurnal profile were predicted from photolysis of ‘typical’ quantities of organoiodine compounds in the marine boundary layer4, although there are large uncertainties in such predictions, especially because of the lack of data on the iodine atom precursors. The concentrations and diurnal behaviour of IO and BrO are thus consistent with prior understanding.

Figure 3: Halogen oxide observations.
figure 3

Averaged diurnal profiles for BrO (a) and IO (c). Errors (1σ) are indicated as grey lines. b, Seasonal variation in BrO; d, seasonal variation in IO. The points show average concentrations seen from 09:00 to 17:00 ut.

Full size image

When observed hourly averaged halogen concentrations are included in the box model, the simulated annual average daily ozone loss (3.2 ± 1.1 (1σ) p.p.b.v. d-1 between 09:00 and 17:00 ut) is comparable to that observed, and the monthly average cycle in ozone loss is better reproduced (Fig. 2). The ozone loss is greatest between March and May and is thought to be a consequence of increased photolysis rates and therefore increased photochemical ozone loss in addition to seasonally changing halogen oxide concentrations. The halogen-mediated O3 destruction contributes an average of 1.8 ± 0.4 p.p.b.v. d-1 to the total ozone loss (Fig. 4). Contributions of 0.35, 1.24 and 2.07 p.p.b.v. d-1 are attributed to BrO, IO and the sum of IO and BrO, respectively, during the maximum in March. The effect of IO and BrO together is greater than the sum of their individual contributions because of the IO + BrO reaction (reaction (3)). If halogens are excluded in the box model calculations, then the model underestimates the O3 loss by 47% and overestimates the O3 concentration by 12% (annual averages). Exclusion of halogens also leads to an underestimation of modelled OH concentrations by 5–12% in this region when compared with the full model concentrations, with corresponding overestimates of the lifetimes of important trace gases such as methane. A compelling argument for exclusion of localized sources of halogens as an explanation for the observations reported here is that if the halogen sources were restricted to the kilometre around the island, the ozone loss rate required to explain the measurements would be more than 100 p.p.b.v. h-1, requiring a halogen loading that is orders of magnitude higher than any tropospheric observations.

Figure 4: Monthly averaged contributions to the daily ozone budgets between 09:00 and 17:00 ut.
figure 4

The contribution from halogen oxides, calculated using the measured concentrations from Nov 2006 to Jun 2007, takes into account both the direct and indirect effects on ozone (see text).

Full size image

The combination of long-term localized chemical data with geographically widespread boundary layer O3 loss seen from aircraft leads us to the conclusion that halogens play an important role in this region and that their chemical influence extends at least over several thousand kilometres. Current understanding of the sources of bromine indicates that there is no a priori reason that this region should not be representative of bromine levels in the oceanic marine boundary layer in general6. The oceans around Cape Verde are biologically active, however, and it is possible that they represent an area of increased iodine release. The inclusion of halogen sources and chemistry into global climate models is now essential for understanding ozone as a climate gas, and for calculating tropospheric oxidizing capacity and the methane lifetime.

Methods Summary

Ozone was measured every minute from a height of 5 m using an ultraviolet absorption instrument (Model 49i Thermo Scientific). The instrument gives an absolute measurement of ozone with a precision of 0.07 p.p.b.v. for hourly averaged data. Mixing ratios for IO and BrO were retrieved from long-path differential optical absorption spectroscopy23,24 using a newtonian telescope acting as a transmitter and receiver, and an array of retro-reflectors placed 6.1 km across a bay from the observatory. Spectra were collected every 30 s and then further averaged over 20–30 min to improve the signal-to-noise ratio. Mixing ratios for CO from 10 m were determined using an Aerolaser 5001 fast response VUV analyser, with a detection limit of <0.5 p.p.b.v. on 1-min average. Measurements of NO x were made from 5 m using a single channel, chemiluminescence NO detector with a photolytic NO2 converter. The instrument25 (Air Quality Design) alternates between measuring NO and NO2 in a 10-min duty cycle, with hourly averaged data giving detection limits of 1.5 p.p.t.v. and 4 p.p.t.v. for NO and NO2, respectively. Hourly VOC measurements (C2–C8 non-methane hydrocarbons, dimethyl sulphide and C1–C3 oxygenated VOC) were made from 10 m using a dual-channel gas chromatograph with flame ionization detection26. Detection limits for VOC ranged between 5 and 10 p.p.t.v. Weekly CH4 measurements were obtained using flask samples with subsequent gas chromatographic analysis. Temperature, relative humidity and wind measurements were collected from 10 m at 1 Hz and then averaged over one minute. Atmospheric pressure and broadband ultraviolet radiation were recorded at 4 m. The rate coefficient J(O1D) was measured at 4 m with a 2π filter radiometer (Meteorology Consult) with <5% precision and ∼20% accuracy for solar zenith angles below 60°, and corrections were applied for the vertical overhead O3 column and changes in sensitivity due to changing solar zenith angle.