Edinburgh Research Explorer Long-term CH3Br and CH3Cl flux measurements in temperate salt marshes

. Fluxes of CH 3 Br and CH 3 Cl and their relation-ship with potential drivers such as sunlight, temperature and soil moisture, were monitored at fortnightly to monthly intervals for more than two years at two contrasting temperate salt marsh sites in Scotland. Manipulation experiments were conducted to further investigate possible links between drivers and ﬂuxes. Fluxes followed both seasonal and diurnal trends with highest ﬂuxes during summer days and lowest (negative) ﬂuxes during winter nights. Mean ( ± 1 sd) an-nually and diurnally-weighted net emissions from the two sites were found to be 300 ± 44 ng m − 2 h − 1 for CH 3 Br and 662 ± 266 ng m − 2 h − 1 for CH 3 Cl. The ﬂuxes from this work are similar to ﬁndings from this and other research groups for salt marshes in cooler, higher latitude climates, but lower than values from salt marshes in the Mediterranean climate of southern California. Statistical analysis generally did not demonstrate a strong link between temperature or sunlight levels and methyl halide ﬂuxes, although it is likely that temperatures have a weak direct inﬂuence on emissions, and both certainly have indirect inﬂuence via the annual and daily cycles of the vegetation. CH 3 Cl ﬂux


Introduction
The methyl halides CH 3 Br and CH 3 Cl are the most abundant natural vectors of bromine and chlorine into the stratosphere and consequently play an important role in stratospheric ozone destruction. The current knowledge of their respective natural sources is incomplete leading to large uncertainties in their global budgets. Besides the issue of quantification, characterisation of possible sources is needed to assist modelling the impacts of future land-use and climate change on the stratosphere.
During the last decade a number of studies of CH 3 Br and CH 3 Cl fluxes at salt marshes have been conducted in Europe, Australia, the US and China with widely differing findings in regard to magnitude and drivers. Whilst extrapolation of flux measurements conducted in southern California (Rhew et al., 2000(Rhew et al., , 2002Manley et al., 2006) suggested that salt marshes could be responsible for 7-12% and 4-5% of global CH 3 Br and CH 3 Cl production, respectively, extrapolation of measurements in Tasmania by Cox et al. (2004), in Ireland by Dimmer et al. (2001) and in Scotland by Drewer et al. (2006) suggested that salt marshes were likely to account for less than 2% of global CH 3 Br and no more than 0.03% of global CH 3 Cl production (all estimates based on a global salt marsh area of 0.38 × 10 12 m 2 , Woodwell et al., 1973). Recent measurements by Rhew and Mazéas (2010) at northern California salt marshes show notably smaller methyl halide net Published by Copernicus Publications on behalf of the European Geosciences Union. emissions in this more temperate climate zone than those previously reported for southern California. Wang et al. (2006) identified a Chinese salt marsh as a net sink of CH 3 Cl. All groups except Wang et al. (2006) identified plants as a strong source of methyl halides whilst sediments were identified as weak emitters or sinks. The different groups found either sunlight, temperature, season or different plant growth stages as the most likely drivers of temporal methyl halide flux variations, with Rhew et al. (2000), Manley et al. (2006) and Rhew and Mazéas (2010) identifying plant species as the cause of spatial flux variation. Of the other studies, only Drewer et al. (2006) attempted to identify factors underlying the spatial variation and suggested sediment bromide content as a possible contributor. In general, fluxes of CH 3 Br and CH 3 Cl were found to be positively correlated with each other.
In summary, data remain sparse for both the size and drivers of methyl halide emissions from salt marshes. Furthermore many of the above studies comprised only a limited number of individual measurements often missing possible annual and/or diurnal flux variations. Thus gaps remain in both quantification and characterisation which must be filled in order to gain a clearer picture of the current and future impact of salt marshes on atmospheric chemistry. The aim of this study was to extend considerably a previous monitoring project for CH 3 Br fluxes at Heckie's Hole salt marsh on the east coast of Scotland (Drewer et al., 2006) by including a second site at Hollands Farm salt marsh on the west coast and measurement of CH 3 Cl. Measurements were conducted for more than 2 yr at the two salt marshes, comprising more than 450 individual daytime measurements in all seasons and more than 100 diurnal measurements.

Site descriptions
Heckie's Hole salt marsh (56 • 1 N 2 • 35 W), located in East Lothian at the mouth of the River Tyne flowing into the North Sea ( Fig. 1), is surrounded on three sides by (wooded) dry land and protected from direct wave action. The site is characterised by very gentle inclines (<0.04%) dissected by deep tidal creeks. Vegetation cover is dense and dominated by Puccinellia maritima in the lower parts and an Armeria maritima-Puccinellia maritima vegetation mixture in the upper marsh but without strong zonation. Other, minor, species found at this salt marsh are Aster tripolium, Glaux maritima, Salicornia europea, Spergularia media and Suaeda maritima as well as Plantago maritima and Triglochin maritima which often occur as dense, mono-specific stands. Sediments at Heckie's Hole were characterised as silty loam with organic matter content in the range 10-22%.
Hollands Farm salt marsh (54 • 58 N 3 • 29 W) is situated besides the estuary of the River Nith as it flows into the Solway Firth on the west coast of the British mainland (Fig. 1).
The marsh is open to the sea and subject to constant tidal wave action, and has accumulated appreciable amounts of sediment over the last few years. The landscape is characterised by a gently inclining lower marsh (∼0.13%), inland of which a small upper marsh area rises via a small step of ∼20 cm. The lower marsh contains at least three distinct topographical and vegetation zones starting from the sea with a flat, sparsely vegetated zone mainly comprising Puccinellia maritima, intermittent areas of hollows and hummocks covered by Aster tripolium, and a narrow zone nearest to the mainland which is covered by a mix of Aster tripolium, Juncus gerardi, Plantago maritima, Puccinellia maritima and Triglochin maritima and which is markedly wetter than the other parts of the lower marsh. The upper marsh is covered by a mixture of mainly graminoid species including Carex binerva, Festuca rubra, Juncus gerardi, Puccinellia maritima and Triglochin maritima. Minor species that occur sparsely at the site are Armeria maritima, Glaux maritima and Trifolium repens. Sediments at Hollands Farm were characterised as sandy loam with organic matter content in the range 1.8-4.4%.
The climate at both sites is classified as Cfb in the Köppen-Geiger climate classification, being temperate maritime, wet all year round with long cool summers (Peel et al., 2007).

Field enclosures
The main body of the work at the two salt marshes was the regular measurement of fluxes at the same sampling points throughout the whole year. At each sampling point a collar constructed of 10-20 cm high, 40 cm internal diameter opaque PVC pipe with a fitting flange of outer diameter 50 cm of the same material was permanently inserted into Biogeosciences, 7, 3657-3668, 2010 www.biogeosciences.net/7/3657/2010/ E.  Br and CH 3 Cl flux measurements in salt marshes 3659 the ground. The collars were inserted flush to the ground to reduce any impact of the collar rims on the enclosed space. At Heckie's Hole two pairs of collars were installed in each of the lower (collar pairs A and B) and upper salt marsh areas (collar pairs C and D) on the 26 May 2007. All collars were situated within the eulittoral (intertidal) zone, i.e. between the spring high tide line and the neap low tide line, with a range of vertical elevation across all collars of only 14 cm. All collars at this site were therefore subject to regular inundation.
At Hollands Farm in the lower marsh area a single collar was installed on a bare patch of sediment (collar E) and one collar pair each on the previously described flat area (collar pair G), on the hummocks (collar pair F) and in the narrow wetter zone nearest to the mainland (collar pair H) on the 30 and 31 May 2007. In the upper marsh one collar pair (I) was installed at the same time, and a second collar pair (J) on the 8 May 2008. At Hollands Farm the range in vertical elevation across all collars spanned 106 cm; all collars designated "lower" (collar E and collar pairs F-H) were situated in the eulittoral zone with regular inundation, whereas collars designated "upper" (collar pairs I + J) were in the supralittoral zone with only occasional inundation.
At both salt marshes the two collars of a collar pair were placed no more than 5 m apart, representing the typical vegetation of a particular part of a salt marsh but were not intended as replicates. In nearly all cases collars contained mixtures of species, as described in Table 1. Halfway between each collar pair a 50 cm depth dip well was installed to measure water table depth. Flux measurements were conducted at fortnightly to monthly intervals between 4 June 2007 to 20 July 2009 at Heckie's Hole and 6 June 2007 to 21 July 2009 at Hollands Farm.
To start a flux measurement a cylindrical, 40 cm diameter chamber, made of 0.5-1.0 mm thick translucent polyethylene terephthalate (typical transmission in the photosyntheticallyactive radiation (PAR) region ≈90%) and equipped with forced air circulation and a gas sampling port, was placed on top of the collar flange and clamped airtight with bulldog clips. The chambers were 21.5 cm high (except for the 6 June 2007 at Hollands Farm where 10.5 cm high chambers were also used). Typically, sampling for flux measurements took place between 10:00 a.m. and 03:00 p.m. (except for diurnal studies). At the end of the 10 min enclosure time a 550 mL air sample was extracted and stored in a 1 L Tedlar bag. Two samples of on-site ambient "background" air were collected at 1.5 m height at the beginning and the middle of each measurement day and stored similarly.
Concurrently with each enclosure, internal chamber temperature and PAR were recorded at 2 s intervals and a reading of ambient air temperature, soil temperature at 10 cm depth and water table depth in the dip well was taken. The mean internal chamber temperature during an enclosure was 5.8 • C above ambient air temperature. Soil samples (0-8 cm depth) for moisture determination were also taken on most measure-ment occasions from the immediate vicinity of the individual collars, stored in air-tight bags for transport to the laboratory and then refrigerated until analysis.

Diurnal flux measurements
To investigate the behaviour of methyl halide fluxes throughout the day diurnal measurement campaigns were conducted. Fluxes were measured for one collar of each collar pair (normally choosing the stronger emitting one) every 4.8 h 5 times at Hollands Farm and 6 times at Heckie's Hole. Individual background air samples (at 1.5 m height) were collected for each enclosure occasion. At Heckie's Hole the time of day of the first and last measurement cycle therefore coincided. Diurnal measurements were conducted three times at both locations, twice during summer and once during winter.

Shading experiment
On the 31 May 2009 a separate field experiment to study the influence of sunlight on methyl halide fluxes was carried out at Heckie's Hole. First, fluxes from each collar pair were measured in the usual way with the transparent 21.5 cm high chamber for 10 min. Then, for one collar of each pair, the measurement was repeated with the chamber and the PAR and temperature sensors beneath a 95 cm long × 62 cm wide × 47 cm high scaffold covered with a semi-transparent green netting. Finally, the scaffold was covered by a blanket and a third enclosure undertaken in total darkness. This was repeated for all four collar pairs at the salt marsh. The aim was to reduce the incident sunlight whilst maintaining the temperature inside the chamber as constant as possible, so as to distinguish between sunlight and air temperature as drivers for methyl halide fluxes. Due to problems with the analytical equipment at the time only CH 3 Br fluxes were quantified.

Vegetation removal experiment
On the 16 and 18 June 2009 at Hollands Farm and Heckie's Hole, respectively, above-ground vegetation from all salt marsh collars was removed using scissors to determine the mass and make-up of the vegetation. In the laboratory it was sorted by species and dead plant material removed. The sorted plant material was then oven dried at 70 • C as described below. Methyl halide fluxes were measured immediately before (in most cases) and after vegetation removal and then on two more occasions at fortnightly intervals. Between measurements the collars were covered with a black landscaping textile to slow vegetation regrowth and any new vegetation was removed before undertaking subsequent measurements. Due to problems with the analytical equipment at the time only CH 3 Br fluxes could be retrieved in the initial stages of this experiment. Table 1. Summary of methyl halide fluxes and vegetation composition of collars at Heckie's Hole and Hollands Farm salt marsh sites. Shown are the mean, minimum and maximum fluxes and number (n) of daytime flux measurements, as well as the Pearson product-moment coefficient (R) between CH 3 Br and CH 3 Cl fluxes, the latter printed in bold where statistically significant at P < 0.05. Vegetation composition is characterised by the total dry mass (m) of above-ground vegetation present in each collar at the end point together with the weight % of the two main species in each collar relative to the total vegetation composition. Species abbreviations are Arm(eria maritima), Ast(er tripolium), Junc(us gerardi), Car(ex binerva), Fest(uca rubra), Plan(tago maritima), Pucci(nellia maritima), Suae(da maritima) and Trig(lochin maritima). The last two columns give the total bromine and chlorine content of the above-ground vegetation calculated from species-specific halogen content and dry mass in each collar. All collars at Heckie's Hole were situated within the eulittoral (intertidal) zone. At Hollands Farm, all collars designated "lower" (collar E and collar pairs F-H) were situated in the eulittoral zone, whereas collars designated "upper" (collar pairs I + J) were in the supralittoral zone. (Further details in the main text.)

Testing of flux linearity
An enclosure time of 10 min was chosen to ensure a large enough accumulation/depletion of methyl halides even from weakly emitting/uptaking collars for subsequent GCanalysis. To ascertain if, and by how much, inferred fluxes differed over the enclosure duration on five occasions fluxes were measured from individual collars at increasing enclosure durations, ranging from 2.5 to 40 min. Derived fluxes at each enclosure time for a given experiment were normalised to their value at 2.5 min and pooled. Regression fitting indicated that CH 3 Br and CH 3 Cl fluxes extrapolated to the ideal "0" min enclosure time could be on average a factor of 1.5 times the fluxes derived after 10 min enclosure. However, where analyte concentration was small (for shortest enclosure times, and for small net fluxes), the larger inherent analytical uncertainty associated with the derived fluxes yielded a large statistical confidence interval on this extrapolation factor. Thus this factor is not applied in the reporting of the measured fluxes, but its effect is considered as potential uncertainty in the subsequent tentative scale-up.

GC analysis
Briefly, 100 mL aliquots of each sample were analysed using an HP5890 GC-ECD with a custom-built two-stage cryogenic pre-concentration unit (Hardacre et al., 2009 (15.80 ± 0.47 ppmV in N 2 , Air Liquide, Paris, France). Accuracy and precision uncertainty combined was quantified as ± 15% for both gases at mixing ratios around ambient. Net flux was derived from the difference in concentration between the enclosure sample and the contemporaneous background air sample and expressed per m 2 per h. The uncertainties in flux values quoted here combine the uncertainties in both instrumental determination of methyl halide concentration in a gas sample and in the enclosure parameters used to convert mixing ratios to fluxes. The main sources of uncertainty derive from interpolation of the calibration curve. Blank and storage experiments were conducted (in light and dark, and over a range of temperatures), and storage under ambient conditions for a week or so was observed not to contribute uncertainty greater than the general magnitude of the other uncertainties present and quantified. Filled sample bags were at slight excess pressure to ambient, so ingress of outside air was unlikely and those that subsequently leaked showed signs of deflation and were discarded from analysis. Since both the background air and enclosure samples were stored in identical conditions such uncertainties are further mitigated by the experimental procedure of quantification by difference.
Discrimination of a significant net flux depends on the ability to determine significant difference in analyte mixing ratio between an enclosure sample and background air. The LOD for determination of a net flux was thus set at twice the uncertainty in the associated background value. The LOD values vary between individual flux measurements because the uncertainty in the background measurement varies with the calibration curve used for that pair of sample and background determinations. The interquartile ranges of individual sample LOD values were 34-71 and 340-750 ng m −2 h −1 for CH 3 Br and CH 3 Cl fluxes, respectively, with lower sample values having lower associated LOD values (because of better linear calibrations over smaller concentration ranges). Where a flux measurement was below its estimated LOD the flux was set to zero for all futher data processing. Before January 2008 the ECD had not yet been converted to oxygen doping and therefore only CH 3 Br measurements are available prior to then.

Determination of soil and plant dry mass
Soil and plant samples were oven dried at 105 • C and 70 • C, respectively, to constant mass. For soils, gravimetric water content was determined by weighing before and after oven drying.

Bromine and chlorine content of plant material
Plant materials were analysed for both total chlorine and total bromine content using a custom-built XRF instrument. The analysis of the samples was carried out at the University of Heidelberg by Andriy K. Cheburkin.
Vegetation from each species was randomly sampled from both salt marshes (except Salicornia europea which was collected only at Heckie's Hole) and the sub-samples mixed into a single sample per species. The fresh material was washed with de-ionised water and then dried in an oven at 70 • C to constant mass. Once dried, the samples were stored in airtight bags and shipped to Germany for analysis.
Combined total chlorine and bromine analysis was carried out on the TITAN-XRF (Cheburkin and Shotyk, 2005), an energy-dispersive XRF spectrometer specifically built for analysis of peat and plant samples. Calibration was carried out with certified plant materials with known halogen content and analytical uncertainty was estimated to be less than 10%.

Methyl halide flux quantification summary
The mean, minimum and maximum daytime methyl halide fluxes measured for the collars at both salt marsh sites together with data on collar vegetation are displayed in Table 1. Both inter-collar and intra-collar CH 3 Br fluxes at Heckie's hole (Table 1a) were on average larger and less variable than fluxes recorded at Hollands Farm (Table 1b). Moreover there was a marked zonation of mean CH 3 Br fluxes at Hollands Farm, with fluxes at the lower marsh collars (pairs F-H) being up to 4 times greater than at the upper marsh collars (pairs I and J). CH 3 Br fluxes at Hollands Farm also tended to be more extreme with flux maxima being much larger than those at Heckie's Hole. No CH 3 Br net uptake was recorded at either salt marsh site during the 26 months of measurements.
Mean CH 3 Cl fluxes at both salt marsh sites were larger than CH 3 Br fluxes and were negative (net uptake) on some occasions. In contrast to CH 3 Br, mean unweighted CH 3 Cl fluxes recorded at Hollands Farm were up to a third larger than fluxes at Heckie's Hole, but again were spatially more variable. Individual CH 3 Cl flux magnitudes from collars at Heckie's Hole were positively correlated with the total dry mass of Puccinellia maritima present (R = 0.94, P = 0.0004). For this reason mean CH 3 Cl fluxes were less homogeneous than CH 3 Br fluxes and mean fluxes from the lower marsh collars dominated by Puccinellia maritima (A2, B1 and B2) were up to 3 times larger than fluxes from the upper marsh area which has a larger proportion of Armeria maritima. For the same reason CH 3 Cl fluxes from collar A1 were considerably smaller than for the rest of the lower marsh as it was dominated by Plantago maritima. In contrast, at Hollands Farm, mean flux magnitudes were negatively cor-   the fact that the vegetation mixture and abundance of plant species differs between the two salt marsh sites. No such correlation between vegetation and flux magnitudes was found for CH 3 Br. For 14 out of the 18 collars with vegetation across both salt marshes there was a statistically significant (P < 0.05) positive correlation between CH 3 Br and CH 3 Cl fluxes, in line with previous findings by Cox et al. (2004), Rhew et al. (2000), Manley et al. (2006) and Rhew and Mazéas (2010). Mean CH 3 Cl/CH 3 Br mass flux ratio was 2.2.

Annual and diurnal flux variations
In Fig. 2, annual and diurnal fluxes from an example collar, D2 at Heckie's Hole, are presented together with air temperature and PAR values at the time of enclosure. Also shown in Fig. 2a is the growing season as a grey shade from the beginning of April to the end of October (days 91-304). The growing season was taken as the period between the appearance of the first green shoots in spring to the dying off of above-ground vegetation in autumn.
As shown in Fig. 2a CH 3 Br fluxes follow an annual cycle with highest fluxes during the growing season and low-est or no fluxes during the non-growing season. An interesting observation was that fluxes from the different collars at Heckie's Hole followed similar patterns to each other in a given year, i.e. had some spatially coherent response, but patterns differed between the three calendar years and peaked at different times. This indicates that there is indeed a common factor driving annual emissions from salt marshes. However, at Hollands Farm, the patterns of methyl halide fluxes from individual collars in the different years varied more between the different collar pairs than between the two collars of a collar pair indicating that either ground conditions or vegetation make-up are also important in the response of methyl halide fluxes to climatic drivers. From studies with stands of mono-specific flowering plants it has been suggested that peaks in methyl halide emission may be associated with flowering events (Manley et al., 2006) but with the mixed-species enclosures in this work such a direct association was not discernible. Non-growing season daytime net emissions of CH 3 Br were on average 22% of growing season emissions at Heckie's Hole and 34% at Hollands Farm. The corresponding figures for CH 3 Cl are 13% and 16% at Heckie's Hole and Hollands Farm, respectively. Biogeosciences, 7, 3657-3668, 2010 www.biogeosciences.net/7/3657/2010/ Methyl halide fluxes during the growing season consistently showed a diurnal pattern with highest fluxes during daylight hours and lowest fluxes during night time (Fig. 2b). However, CH 3 Br nighttime fluxes typically did not cease but dropped on average to 70% of daytime fluxes at Heckie's Hole and 60% of daytime fluxes at Hollands Farm whilst CH 3 Cl fluxes often dropped to zero or were negative at some point during the night (average daytime/nighttime flux ratio 40% at Heckie's Hole and 60% at Hollands Farm). During the non-growing season CH 3 Br fluxes followed a much weaker diurnal cycle with a similar pattern whilst CH 3 Cl fluxes often appeared to be random. There is currently no suggested reason for this differing behaviour.

Statistical analysis of potential factors impacting flux magnitudes
Statistical analyses were performed in two different ways: 1. Net methyl halide fluxes were tested simultaneously for significant differences in flux magnitudes between the two measurement sites, between the growing and nongrowing season, and between the collars within a site using the General Linear Model (GLM) function of the Minitab ® software package. The raw net flux data from the two sites were tagged according to site, collar and season, where the factor "collar" was nested within the factor "site". As well as testing for significance of these factors, interaction between season and site was also tested. Two data sets, one for CH 3 Br and one for CH 3 Cl, were derived.
2. Normalised methyl halide fluxes within each of the two sites for the growing season and non-growing season were analysed for their association (through correlation) with the measured environmental factors (PAR, ambient air temperature, soil temperature, internal chamber temperature, water table depth and soil water content). For these analyses, the following modifications were made to the raw data: (a) net methyl halide fluxes from an individual collar were normalised against the highest flux of that collar for the whole of the measurement period; (b) numerical data for PAR, ambient air temperature, soil temperature and internal chamber temperature were left unchanged whereas data for soil water content and water table depth were adjusted for spatial variations within a salt marsh site by subtracting the lowest value measured at each collar from the raw data for that collar; (c) the data so derived from the individual collars were pooled for each salt marsh site and season to allow for analysis of common factors affecting flux variations over time.
For both these procedures the fluxes from collar E were excluded since it did not contain any vegetation. All testing was performed to a 95% significance level, adjusted for multiple testing using the Bonferroni correction (Abdi, 2007).
The GLM analyses of variations according to site, season and collar revealed no significant differences in flux magnitudes between the two salt marsh sites for either CH 3 Br or CH 3 Cl (Table 2), but significant difference in flux magnitudes between season, and between collars at a site, for both methyl halides. This indicates a greater variability in net fluxes within a salt marsh than between the two salt marsh sites. It also affirms the appropriateness of separating the year into these two different seasons. There was no significant interaction between season and site, i.e. the effect of seasonality was similar at both salt marshes.
The correlation analyses of temporal flux variations and their dependency on drivers such as temperature and light yielded the results shown in Table 3. Net CH 3 Br and CH 3 Cl fluxes from collars at Heckie's Hole during the growing season varied significantly (positively) with ambient air, internal chamber and soil temperature at 10 cm depth. At Hollands Farm there was a weaker (but still signficant) positive correlation between net flux variations of CH 3 Br and internal chamber temperature. However, there was no significant influence of PAR on net methyl halide fluxes from either salt marsh, during the growing season. This lack of correlation is surprising since it has been suggested that sunlight levels are important drivers for methyl halide production by plants (Dimmer et al., 2001;Manley et al., 2006;Wang et al., 2006;Drewer et al., 2006). However, separating individual effects of PAR and temperature is difficult because of their intrinsic tendency to correlate. Instead the findings from this study suggest that at the study sites the main process of methyl halide production from plants could be as secondary metabolites in plants. This S-adenosyl-L-methionine (SAM) methyl transferase dependent reaction (Attieh et al., 1995;Manley, 2002) is light independent since it is not directly coupled to the primary metabolic pathway of photosynthesis. Consequently a temperature dependency of the enzyme activity and therefore of the methyl halide production rate would be expected. During the non-growing season net CH 3 Br fluxes at the two sites varied significantly with both PAR and all three measures of temperature. Again this is surprising since the definition of non-growing season applied here is no aboveground growing plant material. It is possible that sunlight levels indirectly influenced CH 3 Br production by the soil microbial community by raising soil temperatures. For CH 3 Cl fluxes there was a significant correlation between net fluxes and PAR levels; however of the different measures of temperature only CH 3 Cl fluxes at Heckie's Hole showed a significant correlation to internal chamber temperatures.
There was no significant correlation between soil water content or water table depth and net methyl halide fluxes for either site and season. It therefore appears that these factors do not directly influence instantaneous methyl halide fluxes to a significant extent although soil moisture conditions will certainly influence the plant species prevalent in different zones of a salt marsh.
In summary, net methyl halide fluxes at Heckie's Hole were more strongly associated with the investigated drivers than at Hollands Farm. Temperature was found to be the best predictor of daytime methyl halide fluxes at both sites during the growing season when most of the methyl halides are produced. During the non-growing season light levels were equally good predictors. However, the influence of measures of temperature and sunlight was only assessed for their direct effect on methyl halide fluxes and did not take into account their indirect influences on vegetation growth or metabolism beyond the rather crude measure of growing and non-growing season. Despite some trends identified, the data do not show generally strong predictor drivers for flux variation across all sites and times.

Shading experiment
PAR levels during the shading experiment decreased on average from ∼1430 µmol m −2 s −1 without shade to ∼450 µmol m −2 s −1 under the netting to 0 µmol m −2 s −1 under the blanket, whilst internal chamber temperatures changed from 28.6 • C to 25.8 • C and then to 22.2 • C, respectively. CH 3 Br fluxes during the shading experiment did not change significantly between the different treatments (1-way ANOVA at P < 0.05) therefore strongly suggesting that PAR does not directly influence CH 3 Br emissions from vegetation within the ∼20 min time-scale of the treatment.

Vegetation removal experiment
CH 3 Br fluxes measured a few minutes after complete vegetation removal were for all collars significantly lower than the corresponding fluxes measured both directly before removal on the same day or most recently prior to that (t-tests at P < 0.05). However, as with the diurnal studies, CH 3 Br fluxes did not generally fall to zero but stayed at a low level. These emissions then decreased on the following two fortnightly measurement occasions. At Hollands Farm CH 3 Br fluxes after vegetation removal were compared in a twotailed t-test against all seasonal CH 3 Br measurements from collar E, which did not contain any vegetation, and no significant differences in flux magnitudes were found. This indicates that emissions/uptake of CH 3 Br from collar E is representative of fluxes from sediment for the whole of Hollands Farm.
As noted in Sect. 3.1.3, CH 3 Cl measurements were not available at the time of vegetation removal itself but on the two occasions two and four weeks after vegetation removal. After vegetation removal CH 3 Cl fluxes at both salt marshes dropped to values near or below the LOD. At Hollands Farm CH 3 Cl uptake at some collars was recorded. Onetailed t-tests showed that CH 3 Cl fluxes for both salt marshes were significantly lower after plant removal (P = 0.0001 for Heckie's Hole and P = 0.001 for Hollands Farm) and, similarly to CH 3 Br, CH 3 Cl fluxes at Hollands Farm after plant removal were found to be not significantly different from fluxes from collar E. Biogeosciences, 7, 3657-3668, 2010 www.biogeosciences.net/7/3657/2010/ In summary there is, as expected from earlier reports, strong evidence that both CH 3 Cl and CH 3 Br are primarily produced by vegetation rather than by micro-organisms in the soil. The findings from this experiment also indicate that below-ground plant biomass, such as roots and rhizomes, is either not involved or plays only a minor part in methyl halide production.

Total bromine and chlorine content of plant material
The total bromine and chlorine concentrations measured in the above-ground vegetation collected at Heckie's Hole and Hollands Farm (Table 4), together with the data on vegetation composition and mass for each collar, were used to estimate the average halogen concentration and total halogen content of the vegetation enclosed in each collar at both salt marsh sites. This was done by multiplying the dry mass of each species by the XRF-derived halogen concentration for the respective species and subsequent summation for all species in an individual collar. For Triglochin maritima the chlorine concentration was estimated from the bromine concentration and the orthogonal distance regression fit of all other available bromine and chlorine concentration pairs. Halogen concentrations for Festuca rubra were taken as the mean values from the other graminoid species Carex binerva, Juncus gerardi and Puccinellia maritima for which data were available. Comparing these calculated halogen content values with the mean methyl halide flux magnitudes from the collars did not reveal any statistically significant correlations. This indicates that variations in halogen content/concentration typically found in salt marsh vegetation do not explain spatial differences in methyl halide fluxes. Indeed, the amounts of bromine and chlorine released from plants via methyl halide emissions are only a small fraction of the total halogen pool available so conversion of halogens into methyl halides is likely to be the limiting factor. This makes it more likely that biological processes associated with plant species themselves are the determining factor in methyl halide flux magnitudes as argued by Manley (2002). As shown in Table 4 the Cl/Br mass ratio in the different plant species varied between 96 to 245 whereas the mean Cl/Br mass ratio in methyl halide emissions was only 1.8 (calculated from the observed CH 3 Cl/CH 3 Br mean net-emission mass ratio of 2.2). There is therefore a clear preference for CH 3 Br production over CH 3 Cl production or, for the corollary, suppression of CH 3 Cl production due to competitition by the more nucleophilic Br − ions.

Interpolation of study data and global upscaling
A weighted averaging of the CH 3 Cl and CH 3 Br fluxes, and extrapolation to global scale, was undertaken as follows. It is emphasised that this scale up utilises only the net fluxes measured in this work. 2. For each salt marsh, a mean methyl halide flux value, representing the entire salt marsh area, was derived for each day j by spatially-weighted averaging of fluxes from individual collars i at that salt marsh. Spatial weighting was based on the proportion of a vegetation zone relative to the total salt marsh area. The formula was where F j ,sa is the spatially-weighted mean methyl halide flux for the whole salt marsh on day j and F ij is the methyl halide flux for each individual collar derived from interpolation of the seasonal flux data. A i is either the relative proportion of a vegetation zone represented by a collar or, if a vegetation zone is represented by n collars, the n-th part of the relative proportion of a vegetation zone of the total salt marsh area (0 ≤ A i ≤ 1, 3. For both sites the number of hours of daylight and darkness for each Julian day were calculated using the formula cosχ = cos(LHA) × cos(LAT) × cos(DEC) where χ is the solar zenith angle at a given time and location, LAT is the latitude of the measurement site, DEC is the declination of the Earth to its orbital plane on that Julian day and LHA is a measure of the time www.biogeosciences.net/7/3657/2010/ Biogeosciences, 7, 3657-3668, 2010 within that day. By setting cos(χ) to 1 and solving for LHA the length of the day as the time between sunrise and sunset was derived.
4. For each site the average ratio of nighttime to daytime methyl halide fluxes (r n/d ) was calculated separately for the growing and non-growing season.
5. For each Julian day the daily flux (ng m −2 d −1 ) was calculated from the interpolated, spatially averaged hourly fluxes, the length of daytime and nighttime hours and the ratio of nighttime/daytime fluxes (Fig. 3). The total annual flux (ng m −2 yr −1 ) for each salt marsh was calculated by summing over all days as shown below.
Here F tot is the total annual methyl halide flux, and t i,daytime and t i,nighttime are the number of hours between sunrise and sunset or between sunset and sunrise, respectively, for each Julian day. for CH 3 Cl and scaling up by the total global salt marsh area estimates of 0.38 ×10 12 m 2 from Woodwell et al. (1973) and 1.65 ×10 12 m 2 from Costanza et al. (1997) yields global estimates of 1.0 and 4.3 Gg yr −1 for CH 3 Br and 2.2 and 9.6 Gg yr −1 for CH 3 Cl. These values would rise if taking into account the previously-mentioned possible non-linearity of measured fluxes with enclosure time, increasing the upper range values to 6.5 Gg yr −1 and 14.3 Gg yr −1 for CH 3 Br and CH 3 Cl, respectively. These ranges of values are equivalent to 0.5-3.2% and 0.05-0.33% of the global annual CH 3 Br and CH 3 Cl budgets, respectively, given in the most recent WMO report (Clerbaux et al., 2007), but again it is emphasised that these scale up estimates are based solely on the measurements presented here.
The mean flux results from this study are in line with the mean flux values reported by Dimmer et al. (2001) (320 ng m −2 h −1 CH 3 Br, 380 ng m −2 h −1 CH 3 Cl), Cox et al. (2004) (190 ng m −2 h −1 CH 3 Br, 300 ng m −2 h −1 CH 3 Cl) and Drewer et al. (2006) (350 ng m −2 h −1 CH 3 Br) for temperate salt marshes but are up to two orders of magnitude smaller than mean fluxes reported for Mediterranean climate salt marshes in southern California by Rhew et al. (2000) (4210 ng m −2 h −1 CH 3 Br, 51 100 ng m −2 h −1 CH 3 Cl) and Manley et al. (2006) (7300 ng m −2 h −1 CH 3 Br, 46 800 ng m −2 h −1 CH 3 Cl, vegetated areas only). On the other hand, the net fluxes reported by Rhew and Mazéas (2010) at northern Californian marshes are within the range observed at other higher-latitude temperate salt marshes. This work therefore confirms again the distinct flux magnitudes of methyl halides from salt marshes in these two climates. It now seems clear that a more nuanced scale up needs to take account of these differences and that salt marsh contribution to global annual emissions of CH 3 Cl and CH 3 Br will lie somewhere between those estimated here and those estimated by Rhew et al. (2000) and used in the WMO reports.
The mean CH 3 Cl/CH 3 Br net flux mass ratio of 2.2 observed in this study lies between the values of 1.2, 1.6, 10 and 6.4 for other salt marsh net emissions reported by Dimmer et al. (2001), Cox et al. (2004), Rhew et al. (2000) and Manley et al. (2006), respectively, with the lower values (2.2 and less) being associated with the more temperate climate salt marshes. Interestingly, the ratio 2.2 is considerably lower than the ∼60 mass ratio in fluxes from vegetation in a tropical rainforest (Blei et al., 2010), or the ∼22 mass ratio in global CH 3 Cl and CH 3 Br budgets. The difference is most likely due to a suppression of CH 3 Cl production by the more nucleophilic bromine/bromide when the concentration of bromine in the plant material is relatively high, as is the case for salt marsh vegetation. This further supports the finding that salt marshes are relatively more important for global CH 3 Br budgets than for global CH 3 Cl budgets, although again cooler temperate and Mediterranean salt marshes appear to behave somewhat differently.  Positive net fluxes of CH 3 Br and CH 3 Cl at temperate salt marshes are principally associated with the vegetation. Temporal flux variations are clearly linked to both seasonal and diurnal cycles but no strong direct link of instantaneous flux to either temperature or sunlight levels was found although temperature is indirectly important through seasonality. Spatial CH 3 Cl flux variations are associated with the abundances of different plant species, although this was not clearly discernible for CH 3 Br. Emissions of CH 3 Br are greater relative to CH 3 Cl from temperate salt marshes than from other terrestrial sources characterised. It appears that salt marshes in cooler temperate climates are unlikely to be a large global source of either methyl halide, amounting to at most a few percent of total global annual CH 3 Br production and an order of magnitude less for CH 3 Cl, compared with the contributions of larger emissions from salt marshes in warmer, Mediterranean-type climates.