Sources and sinks of methyl bromide and methyl ... - AGU Publications

0 downloads 0 Views 533KB Size Report
the cryotrap loop was immersed in hot water to inject the analytes onto a 60 m,. 1.4 mm thick film capillary column (J&W DB‐VRX, Agilent. Technologies, Santa ...
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 116, G03026, doi:10.1029/2011JG001704, 2011

Sources and sinks of methyl bromide and methyl chloride in the tallgrass prairie: Applying a stable isotope tracer technique over highly variable gross fluxes Robert C. Rhew1 Received 14 March 2011; revised 13 May 2011; accepted 3 June 2011; published 1 September 2011.

[1] Methyl bromide (CH3Br) and methyl chloride (CH3Cl) are important stratospheric

ozone depleting compounds, but their natural terrestrial sources and sinks have large uncertainties. Gross fluxes of these compounds were measured during the growing season using a stable isotope tracer technique at a temperate tallgrass prairie in northeastern Kansas, United States. Results show that the tallgrass prairie acts as both a source and sink for CH3Br and CH3Cl, with large emissions associated with Amorpha spp. shrubs and a smaller soil sink than expected. Net flux behavior is not significantly altered by the addition of stable isotope tracers. Four models employed to calculate gross fluxes are largely in agreement with each other except at sites with high CH3Cl emissions, where one model most robustly quantifies gross consumption. Gross production rates may increase at Amorpha sites following simulated rainfall, but gross consumption shows no clear response, suggesting that late season microbial uptake of methyl halides is not strongly influenced by sudden changes in soil moisture. Citation: Rhew, R. C. (2011), Sources and sinks of methyl bromide and methyl chloride in the tallgrass prairie: Applying a stable isotope tracer technique over highly variable gross fluxes, J. Geophys. Res., 116, G03026, doi:10.1029/2011JG001704.

1. Introduction [2] Methyl bromide (CH3Br) and methyl chloride (CH3Cl) are presently responsible for one half of the total bromine and one sixth of the total chlorine in the troposphere, respectively [Montzka and Reimann, 2011]. Their moderately long atmospheric lifetimes (∼0.8 yrs for CH3Br and ∼1.0 yr for CH3Cl) ensures that a significant fraction of these compounds are transported to the stratosphere [Penkett et al., 1980; Schauffler et al., 1998], where the halogen atoms catalyze ozone destruction [McElroy et al., 1986; Molina and Rowland, 1974; Stolarski and Cicerone, 1974; Wofsy et al., 1975]. Anthropogenic sources are believed to account for ∼20–40% of the global CH3Br source and ∼10–15% of the global CH3Cl source [Clerbaux and Cunnold, 2007; Montzka and Reimann, 2011]. In the 1990s, CH3Br was used widely as an agricultural and structural fumigant, but because of its ozone depleting potential, it was added to the list of controlled substances as part of the Copenhagen Amendments to the Montreal Protocol [United Nations Environment Programme, 1992]. The international phase out of methyl bromide began in 1999, and subsequently, the globally averaged surface mixing ratio of CH3Br declined from a peak of ∼9.2 ppt in 1996–1998 to ∼7.4 ppt in 2008 [Montzka et al., 2003; Yokouchi et al., 1 Department of Geography and Berkeley Atmospheric Sciences Center, University of California, Berkeley, California, USA.

Copyright 2011 by the American Geophysical Union. 0148‐0227/11/2011JG001704

2002b; Yvon‐Lewis et al., 2009]. Methyl chloride, on the other hand, is not a Montreal Protocol gas, and its globally averaged mixing ratios have fluctuated between 530 and 550 ppt between 2004 and 2008, with different global networks showing slightly different values and trends [Montzka and Reimann, 2011]. [3] Tropospheric lifetimes for these methyl halides are commonly derived from their global sinks. For CH3Br, the major sinks are oxidation by hydroxyl radical, followed by chemical and biological degradation in the oceans and biological degradation in soils. The sinks for CH3Cl are similar, with oxidation by hydroxyl and chlorine radicals as the dominant sink, followed by loss to polar ocean waters and degradation in soils. Of these sinks, the largest uncertainty involves the role of soils, with published soil sink estimates ranging from 32 to 154 Gg yr−1 for CH3Br [Serça et al., 1998; Shorter et al., 1995], and from 100 to 1600 Gg yr −1 for CH 3Cl [Clerbaux and Cunnold, 2007; Keppler et al., 2005]. Reducing the soil sink uncertainty would improve the estimates of atmospheric lifetimes calculated for these compounds. [4] Global source estimates of these compounds are even more poorly constrained. The major known sources of atmospheric CH3Br include emissions from fumigation (soils, quarantine and preshipment), surface oceans, biomass burning, leaded gasoline combustion, wetlands, Brassica crops, and fungus [Montzka and Reimann, 2011, and references therein]. Using best estimates, the CH3Br sinks outweigh the sources by about 35 Gg yr−1, or roughly 20–25% of the total annual flux [Yvon‐Lewis et al., 2009]. This large

G03026

1 of 15

G03026

RHEW: TALLGRASS PRAIRIE CH3Br AND CH3Cl FLUXES

“missing source” for CH3Br appears in both prephaseout (1996–1998) and current (2008) budgets and appears to be terrestrial and nonindustrial in origin [Lee‐Taylor et al., 1998; Reeves, 2003; Warwick et al., 2006]. Our understanding of the CH3Cl budget had an analogous imbalance, but recent studies point to tropical and subtropical forests as the probable dominant source for CH3Cl, based on ship and aircraft based measurements [Gebhardt et al., 2008; Yokouchi et al., 2000]; plant incubations [Blei et al., 2010b; Saito et al., 2008; Yokouchi et al., 2002a, 2007]; isotope mass balance approaches [Keppler et al., 2005; Saito and Yokouchi, 2008]; and atmospheric chemistry and transport models [Lee‐Taylor et al., 2001; Xiao et al., 2010; Yoshida et al., 2006]. Other major sources of CH3Cl include industrial processes, biomass burning, oceans, wetlands and wood‐rot fungi [Montzka and Reimann, 2011, and references therein]. [5] Measuring methyl halide fluxes in terrestrial ecosystems is complicated by the presence of simultaneous production (typically associated with plants and/or fungi) and consumption (typically associated with soils). Thus, specific sites within an ecosystem can act as either a net source or net sink, depending on season, soil condition, or vegetative cover [Cox et al., 2004; Dimmer et al., 2001; Rhew et al., 2001, 2007; Varner et al., 2003]. Subbiome classifications are also important; for example, the temperate woodland (beech, ash and sycamore) soils in Scotland appear to be a net source for CH3Br [Drewer et al., 2008] while the temperate woodland (oak‐savanna) soils in California appear to be a net sink for CH3Cl and CH3Br [Rhew et al., 2010]. To interpret the highly variable net fluxes found in many of these ecosystems, gross production and consumption rates need to be measured. Gross fluxes may be derived from net flux measurements by clipping vegetation and quantifying soil uptake separately, by conducting accompanying dynamic chamber measurements to solve for equilibrium uptake rates [White et al., 2005] or by modeling soil uptake separately from the net flux [Varner et al., 2003]. These methods, however, are indirect methods that require major disruption to the plant‐soil system, pairs of consecutive measurements, or soil uptake rates derived from laboratory incubations. [6] An alternate method to measure gross fluxes (in addition to net fluxes) is to add a small amount of 13C labeled isotopes of the gas of interest to an incubation chamber and monitor the headspace concentration changes of both the 12C and 13C isotopologues [von Fischer and Hedin, 2002]. This stable isotope tracer technique has been developed for measuring gross fluxes of CH3Br and CH3Cl in laboratory soil incubations [Rhew et al., 2003], modified from a technique to separate biological and chemical degradation rates of CH3Br in seawater [King and Saltzman, 1997]. It has subsequently been applied to field studies in annual grasslands [Rhew and Abel, 2007], shortgrass steppe [Teh et al., 2008], oak‐savanna woodland [Rhew et al., 2010] and Arctic tundra [Teh et al., 2009]. Concurrent with the present work, the technique was applied to coastal salt marshes [Rhew and Mazéas, 2010] and rice paddies [Khan et al., 2011], where high emissions of methyl halides posed additional challenges for the isotope tracer technique.

G03026

[7] Grasslands account for roughly 20% of the global land surface, making this the second largest terrestrial biome globally, after forests [Matthews, 1983]. Prior studies indicated that temperate grasslands account for nearly 25% of the soil sink for CH3Br [Serça et al., 1998; Shorter et al., 1995], but more recent field studies of California annual grasslands and Colorado shortgrass steppe suggest much lower gross uptake rates [Rhew and Abel, 2007; Teh et al., 2008]. In this study, a stable isotope tracer technique is used to measure both gross production and gross consumption fluxes in the tallgrass prairie of Kansas, a mesic grassland in the Flint Hills region of North America. Because CH3Br and CH3Cl fluxes had not yet been measured in this ecosystem, the initial goals were to determine if gross production rates varied greatly between tallgrass prairie plant species and to test the relationship between gross uptake rates and soil moisture through a simulated rainfall experiment. Stable isotope tracer (or isotope pool dilution) techniques are still relatively novel for measuring gross fluxes of trace gases, and this study offered the opportunity for testing and refining the methodology in two ways. First, the effect of added stable isotope tracers on the net fluxes is tested through consecutive experiments without and with the tracers. Second, four different statistical approaches to calculate gross fluxes are compared here to identify the most versatile and accurate method to apply when a wide range of net fluxes is observed.

2. Field Campaigns [8] Two field studies (June 2006 and August 2007) were conducted at Konza Prairie (39°05′N, 96°35′W), a Long‐ Term Ecological Research field site located near Manhattan, Kansas. Konza Prairie is a temperate, mesic grassland growing on a terraced topography of limestone benches and shale slopes [Oviatt, 1998]. The 3487 ha (35 km2) of Konza Prairie are part of the largest remaining contiguous 1.6 million ha (16,000 km2) tract of tallgrass prairie [Knapp and Seastedt, 1998]. Konza Prairie is divided into watershed units, each of which is undergoing experimental long‐ term fire and grazing treatments. Prescribed fire frequencies are 1, 2, 4, 10, or 20 years, with burns typically occurring in the spring, but with some watersheds burned in the summer, fall or winter instead. Herbivory treatments involve the presence or absence of large grazing ungulates (bison and/or cattle) [Knapp and Seastedt, 1998]. [9] Konza Prairie is dominated by native, perennial C4 grasses such as Andropogon gerardii (big bluestem), Schizachyrium scoparium (little bluestem), Sorghastrum nutans (Indian grass) and Panicum virgatum (switchgrass). Nearly 600 vascular plants representing 96 families and 336 genera have been identified on Konza Prairie, with the 3 most species‐rich families being Asteraceae, Poaceae, and Fabaceae [Freeman, 1998; Towne, 2002]. Konza Prairie receives an average of 835 mm precipitation per year, but has ranged from half to nearly twice the mean between 1890 and 1995. About 75% of annual precipitation falls during the growing season (April to September), with the rainfall distribution slightly bimodal, peaking in the early (May/June) and late (September) months [Hayden, 1998]. The mean annual temperature at Manhattan, KS (1891–2006) is 13°C,

2 of 15

G03026

RHEW: TALLGRASS PRAIRIE CH3Br AND CH3Cl FLUXES

with monthly means of −1.8°C in January up to 26.5°C in June [Briggs, 2011]. [10] The first field campaign (Konza‐1) was conducted between 15 and 18 June 2006, during the early growing season. Grasses were in a vegetative stage, 0.3 to 0.9 m tall, with some forbs and woody shrubs in the flowering stage. There were no rain events during this time period. Sixteen flux measurements were performed at eight different sites (sites A–H) in three different watersheds (Table 1). At each site, two consecutive flux measurements were conducted: (1) a net flux measurement (“unspiked”) to quantify background exchange rates and (2) an isotope tracer measurement (“spiked”) where 13C tracers were applied to the chamber headspace to quantify gross production and consumption fluxes. The objective of the consecutive experiments was to test whether a small addition of 13CH3Br and 13 CH3Cl to the chamber headspace would significantly affect the amount of 12CH3Br and 12CH3Cl produced or consumed. Although no effect on net fluxes was expected, the isotope tracer method had not been extensively tested in the field previously. Between measurements, the chamber was ventilated for 10 min to restore chamber air to background atmospheric concentrations. [11] The second field campaign (Konza‐2) was conducted between 31 July to 9 August 2007, and all experiments included stable isotope tracers. Aboveground biomass on the prairie was at or near its annual peak; grasses were in the reproductive stage and reached maximum heights greater than 1.8 m. Three chamber sites (sites I–K), each in a different watershed, were chosen to test the short‐term effects of simulated rainfall on gross fluxes of CH3Br and CH3Cl (Table 1). Two of the sites contained Amorpha species as one of the two dominant plant species: Amorpha canescens (leadplant) at site I and Amorpha fruticosa (Indigo bush) at site J. Site K contained Panicum virgatum (switchgrass) as the dominant plant species. [12] Two flux measurements were conducted on each site at 24 and 1 h prior to simulated rainfall. A 12 mm rainfall event was then simulated by sprinkling three liters of water on the site over a 6 min period. The magnitude of this water addition is within the range of precipitation experienced during a Kansas thunderstorm at Manhattan, Kansas (http:// wdl.agron.ksu.edu/). Five flux measurements were made at 1, 4, 9, 24, and 48 h afterwards. Periodic rain showers occurred throughout the field campaign, but natural rainfall was excluded from each site during each 3 day study period.

3. Methods 3.1. Field Sampling [13] Gas samples for flux measurements were collected from two component, all aluminum, vented static flux chambers [Livingston and Hutchinson, 1995]. Opaque chambers yield similar net methyl halide fluxes to climate‐ controlled transparent chambers over the time scale of the enclosure [Rhew and Mazéas, 2010; Rhew et al., 2001]. Depending on the height of enclosed vegetation, either a short (23 cm height, 0.264 m2 footprint, 61 L) or tall (48 cm height, 0.252 m2 footprint, 121 L) base was used. The chamber base was placed 2 to 5 cm into the ground at least 1 h before the first sampling to allow the soil system to reequilibrate after the disturbance. To initiate the enclosure,

G03026

a chamber lid (127 L) was placed into a water‐filled channel on the rim of the base. Two internal fans mixed air during the enclosure, and the stainless steel sampling tube was outfitted with a 7 mm filter to prevent particulate contamination of air samples. [14] For ‘spiked’ chambers, an aliquot of a gas mixture (40–50 mL of 0.46 ppm 13CH3Br, 4.6 ppm 13CH3Cl, and 4.6 ppm CFC‐113 in N2) was injected into the chamber headspace with a glass and Teflon syringe immediately following closure. This isotope ‘spike’ yielded initial chamber concentrations of 30–180 ppt 13CH3Br, 830–2030 ppt 13 CH3Cl and 950–1950 ppt CFC‐113, depending on chamber volume. The added CFC‐113 (C2Cl2F3) served as an inert tracer [Khalil and Rasmussen, 1989]. [15] Air samples were drawn from the chamber into previously evacuated 3 L silica‐lined or 1 L electro‐polished stainless steel canisters (Restek, Bellefonte, Pennsylvania, United States; LabCommerce, San Jose, California, United States). In Konza‐1, air samples were taken at 1, 15 and 30 min from unspiked chambers, and at 2, 12, 22 and 32 min from spiked chambers. In Konza‐2 (spiked chambers only), air samples were taken at 2, 17 and 32 min. During sample collection, the upwind‐facing ventilation line was opened to allow for pressure equilibration [Livingston and Hutchinson, 1995]. Ambient air samples were also collected between chamber experiments. [16] Air temperatures (inside and outside of the chamber) and barometric pressure were recorded every 5 min during each enclosure. Soil temperatures at 5 and 10 cm and volumetric water content (VWC) at 0–6 cm (ThetaProbe ML2x, Dynamax Inc., Houston, Texas, United States) were measured immediately after each spiked experiment. To determine aboveground biomass, both dead and live vegetation were harvested, separated into identifiable species, and oven dried to a constant weight at 55–65°C for over 24 h. Following harvest, soil cores were collected from each watershed using a soil corer (AMS, Inc., American Falls, Idaho, United States) at 5 or 10 cm depth. Subsamples of the cores were dried to constant weight at 105°C to determine water content. 3.2. Gas Analysis [17] Two subsamples were drawn from each canister and analyzed on a gas chromatograph‐mass spectrometer (Agilent 6890N/5973 GC/MS). The sample inlet system was designed to analyze low volume air samples, utilizing an evacuated 300 mL end volume to draw air from the sample flasks through a 1/8″ stainless steel loop packed with glass beads immersed in liquid nitrogen [Aydin et al., 2002]. Analytes of interest were condensed on the trap while the permanent gases collected in the end volume, where a pressure transducer (Baratron 626A, MKS Instruments, Andover, Massachusetts, United States) quantified the amount of sample gas that passed through the loop. The analytes were thermally desorbed at 100°C from the trap under helium flow and cryofocused on a narrow bore 1/16″ stainless steel loop. While under helium flow (1.2 mL min−1), the cryotrap loop was immersed in hot water to inject the analytes onto a 60 m, 1.4 mm thick film capillary column (J&W DB‐VRX, Agilent Technologies, Santa Clara, California, United States). The column was divided into a 15 m precolumn and a 45 m main column connected via a six‐port Valco valve (VICI Inc.,

3 of 15

4 of 15

8–14

15–21

J

Kb

K2A

NT

SpB

SpB

SpB

2C

2C

2C

2C

4F

Watersheda

31–35

27–32

21–31

27–29

22–26

22

18–21

30

27–28

26–29

25–30

24–28

22–26

25

24

21

21

24

23

22

Tsoil 5 cm (°C)

10–26

31–39

26–40

30

31

23

19

17

21

16

Soil VWC (%)

136

271

388

3–6 August 2007 16 6–9 August 2007 15

105

47

302

310

31 July to 3 August 2007 3

2

18 June 2006 2

14

17 June 2006 14

318

497

16 June 2006 14 14

731

Dead Total Dry Weight (g m−2)

15 June 2006 14

Months Since Last Burn

449

323

526

361

180

211

366

191

319

239

Live Total Dry Weight (g m−2)

Panicum virgatum Andropogon gerardii

Amorpha fruticosa Carex spp.

Andropogon gerardii Amorpha canescens

Schizachyrium scoparium Sorghastrum nutans Amorpha canescens Schizachyrium scoparium

Psoralea spp. Andropogon gerardii Schizachyrium scoparium Sorghastrum nutans

Lespedeza spp. Andropogon gerardii Schizachyrium scoparium Sorghastrum nutans

Andropogon gerardii Carex meadii

Dominant Live Biomass (Two Most Abundant)

174 89

218 29

279 157

100 81 255 58

199 108 120 35

131 88 90 74

177 20

Dry Weight (g m−2)

a Watershed codes indicate burning frequency or location as follows: 4F, burn frequency 4 years; 2C, burn frequency 2 years; SpB, burned annually in spring; NT, Nature Trail watershed open to public and crossed with hiking trails; K2A, ungrazed watershed on north branch of Kings Creek. b Site K biomass collected at an adjacent site.

1–7

11–12

F

I

9–10

E

15–16

7–8

D

H

5–6

C

13–14

3–4

B

G

Experiments

Site

Tchamber (°C)

Table 1. Experimental Field Conditions and Predominant Vegetation at Konza‐1 and Konza‐2 Field Outings

G03026 RHEW: TALLGRASS PRAIRIE CH3Br AND CH3Cl FLUXES G03026

G03026

G03026

RHEW: TALLGRASS PRAIRIE CH3Br AND CH3Cl FLUXES

Houston, Texas, United States). This valve was switched after 10 min to back flush slower eluting compounds from the precolumn prior to the subsequent run. The GC oven temperature was held constant at 30°C for 348 s, ramped up at a rate of 25°C min−1 to 180°C, and maintained at 180°C for 492 s. [18] To determine the concentrations of 12CH3X and 13 CH3X, the peak areas of the following mass to charge ratios (m/z) were quantified in selective ion monitoring mode: m/z = 50 and 52 for 12CH3Cl, m/z = 51 and 53 for 13 CH3Cl, m/z = 94 and 96 for 12CH3Br and m/z = 95 and 97 for 13CH3Br. Ion fragmentation of the heavier isotopologues that had identical m/z ratios as the lighter isotopologues caused some signal overlap in spiked chamber samples. To correct for this, the ion fragmentation ratios associated with 12 CH3X and 13CH3X were determined for each set of samples using a separate set of high concentration standard injections [Rhew and Abel, 2007; Rhew et al., 2003]. Prior studies found no significant difference in concentrations or fluxes derived from the pairs of halogen isotopologues [Rhew and Abel, 2007; Rhew et al., 2003]. Thus, concentrations of 12CH3Cl, 13CH3Cl, 12CH3Br and 13CH3Br were derived from the peaks which required the smallest corrections: m/z = 50, 53, 94, and 97, respectively. CFC‐ 113 (m/z = 153) was also quantified. Concentrations are reported here as dry air mole fractions: ppt = parts per trillion (pmol mol–1). [19] Two seven‐point calibration curves (one before and one after sample analyses) were made by trapping different volumes of a reference standard gas. The working standard (9.4 ppt CH3Br and 561 ppt CH3Cl) was compressed natural air collected at Trinidad Head, California and calibrated at the Scripps Institution of Oceanography on the SIO‐2005 scale. The same standard was run multiple times daily during sample analyses to monitor and correct for daily instrument drift. Higher concentration synthetic standards were also used to extend calibration curves for high concentration samples. [20] For this study, instrumental precisions (1 s) based on daily standards after applying drift corrections were 4.4% for CH3Cl, 7% for CH3Br, and 5% for CFC‐113. Using these precisions and typical chamber conditions, we estimate the minimum detectable flux for the smaller chamber used in the first outing to be 37 nmol m−2 d−1 for CH3Cl, 1 nmol m−2 d−1 for CH3Br and 6 nmol m−2 d−1 for CFC‐113. For the taller chamber used in the second outing, the minimum detectable flux is 55 nmol m−2 d−1 for CH3Cl, 2 nmol m−2 d−1 for CH3Br and 9 nmol m−2 d−1 for CFC‐113. Chamber control experiments over an aluminum sheet with Viton gaskets showed no significant reactivity or emission for these compounds. 3.3. Net Flux Calculations [21] Net fluxes were calculated using the following equation:  Net f lux nmol m2 d1 m  n  ð1440 min=dayÞ  ð1000 nmol=pmolÞ ¼ ðSurface AreaÞ

time; ‘n’ represents the number of moles of air in the chamber; and ‘Surface Area’ (m2) represents the chamber footprint surface area. Net flux errors were calculated by propagating the standard error on the slope with the uncertainties (volume, temperature and pressure) associated with ‘n’. [22] When net uptake was observed (i.e., negative slopes), a first order reaction was assumed [Goodwin et al., 2001; Hines et al., 1998; Shorter et al., 1995; Varner et al., 1999], and a least squares fit was applied to the natural logarithm of the concentration (unitless) versus sampling time (min). The resulting slope (min−1) represents ‘−knet’ as follows: d ½CH3 X  ¼ knet  ½CH3 X  dt

Net uptake rates (ppt min–1) were then normalized for CH3Br and CH3Cl by multiplying knet by the seasonally averaged background concentrations in Northern Hemisphere air between 1998 and 2001 (10.4 ppt for CH3Br and 535.7 ppt for CH3Cl) [Simmonds et al., 2004]. This normalized net uptake rate was substituted for ‘m’ in equation (1) to yield the normalized net uptake flux. The chosen background concentrations were utilized in order to compare results to similar studies [Khan et al., 2011; Rhew and Abel, 2007; Rhew and Mazéas, 2010; Rhew et al., 2007, 2010; Teh et al., 2009], but it should be noted that uptake rates can be adjusted for different time periods by multiplying the fluxes by the ratio of new to old tropospheric concentrations. Unless otherwise specified, fluxes are reported in nanomoles per square meter per day (nmol m−2 d−1), although enclosures were 30–32 min long. For clarity, consumption (net or gross) rates are reported as negative values, while production (net or gross) rates are reported as positive values. When fluxes are averaged, errors are reported as ± one standard deviation (1 s). 3.4. Gross Flux Calculations [23] Gross production and consumption rates of CH3Br and CH3Cl were calculated using the stable isotope tracer method [Rhew and Abel, 2007; Rhew et al., 2003], which entailed adding 13CH3X (X = Br or Cl) tracers to the chamber headspace and subsequently monitoring the 12 CH3X and 13CH3X concentrations. Production rates ‘P’ were assumed to be linear over the time scale of the experiment, yielding the following time‐dependent model of methyl halide concentrations, [CH3X]: d ½CH3 X  ¼ P  k  ½CH3 X  dt

This equation can be applied to both headspace concentrations as follows:



ð1Þ

where ‘m’ (ppt min−1) represents the slope of the linear least squares fit to the measured dry air mole fractions versus

ð2Þ

13

CH3X and

ð3Þ 12

CH3X

  d ½12 CH3 X  ¼ ðF12 Þ  P  ðk12 Þ 12 CH3 X dt

ð4Þ

  d ½13 CH3 X  ¼ ðF13 Þ  P  ðk13 þ kL Þ 13 CH3 X dt

ð5Þ

where ‘F12’ and ‘F13’ indicate the fraction of production that is in the form 12CH3X (0.989) and 13CH3X (0.011),

5 of 15

G03026

RHEW: TALLGRASS PRAIRIE CH3Br AND CH3Cl FLUXES

respectively. While known sources have a range of isotopic values [Bill et al., 2002; Thompson et al., 2002], the differences are negligible in the model, as the typical d13C value of −40 ± 20 ‰ translates to an F12 value of 0.9893 ± 0.0002. Even if the source was exceptionally depleted at −135 ‰ [Keppler et al., 2004], the resultant F12 value would be 0.9904, a difference of 0.001. The first order uptake rate constants, ‘k12’ and ‘k13’, are related by experimentally derived stable carbon isotopic fractionation factors (a = k12/k13), with a measured as 1.069 ± 0.009 for CH3Br and 1.046 ± 0.004 for CH3Cl [Miller et al., 2001, 2004]. The uncertainty in a yields only a minor increased uncertainty of calculated fluxes [Rhew et al., 2003]. The loss of the inert tracer F‐113, which represents physical advective and diffusive processes [Andersen et al., 1998], is approximated as a first order loss [Rhew and Abel, 2007] and is represented in the model with the rate constant kL. [24] The time‐dependent solutions to equations (4) and (5) are represented by equations (6) and (7), respectively, 12

13

CH3 X

CH3 X

 t

 t

¼

¼

ðF12  PÞ k  12   ðF12  PÞ 12   CH3 X 0 expfk12 ðt  t0 Þgð6Þ k12    ðF13  PÞ ðF13  PÞ 13   CH3 X 0 ðk13 þ kL Þ ðk13 þ kL Þ  expfðk13 þ kL Þðt  t0 Þg

ð7Þ

where [12CH3X]0 and [13CH3X]0 represent the concentrations at the first sampling time (t0) after enclosure, and [12CH3X]t and [13CH3X]t represent the concentrations at sampling time t. 3.5. Four Models [25] There are several methods to proceed from here to estimate k and P from the flux chamber data, and different field situations may require different approaches. This study compares four slightly different models to identify which one performs the best across the range of fluxes observed at the tallgrass prairie. Models are given shorthand names based on their method or original literature citation. Model 1 (“k then P”) is the simplest model, using the 13CH3X measurements to derive the uptake rate constant ‘k’, which is then applied to the 12CH3X measurements to solve for P. Model 2 (“RAS2003”), model 3 (“R2011”) and model 4 (“vFH2002”) all simultaneously solve for P and k12, along with the initial concentrations of 13CH3X and 12CH3X; they differ only in the error minimization function used to solve for the best fit of the model to the observed data. 3.5.1. Model 1, “k Then P” [26] In model 1, the production of 13CH3X is assumed to be negligible, simplifying equation (5) to   d½ CH3 X  ¼ ðk13 þ kL Þ 13 CH3 X dt 13

ð8Þ

This assumption is applicable in cases where gross consumption exceeds gross production (i.e., negative net fluxes) because the 13CH3X tracer production rate would be 25% variation (1 standard deviation divided by mean).

9 of 15

G03026

RHEW: TALLGRASS PRAIRIE CH3Br AND CH3Cl FLUXES

G03026

Figure 6. Konza‐2 (sites I–K) chamber air temperature, soil moisture, and gross production and consumption fluxes of CH3Br and CH3Cl versus time (in h) relative to simulated rainfall. Dashed line represents t = 0, the time of added water. Error bars on individual measurements are smaller than symbol size.

[39] On the basis of the three methods of evaluation above, all models performed well where net uptake of methyl halides was observed. However, at sites with very high net emissions of CH3Cl, model 2 performed poorly. In those situations, both models 3 and 4 performed well, with model 3 slightly better in assessing gross production and model 4 slightly better in assessing gross consumption. Given the higher agreement with observed data, model 4 was selected as the most appropriate for the wide range of net fluxes observed in the tallgrass prairie (Figures 1 and 2), and especially to characterize soil uptake at sites with large net emissions. Thus, the following results utilize the gross flux results of model 4. 4.3. Gross Fluxes and Soil Moisture [40] At Konza‐1 (Figure 3), the gross production rates varied widely, from 2 to 485 nmol m−2 d−1 for CH3Br and 0 to 9200 nmol m−2 d−1 for CH3Cl. As with the net fluxes, site H (Amorpha canescens) stood out as having the largest gross methyl halide production rates of that outing. Gross consumption rates, on the other hand, were constrained to a much more limited range, from −6 to −14 nmol m−2 d−1 for CH3Br and −330 to −620 nmol m−2 d−1 for CH3Cl. Thus the site‐to‐site (intersite) variability of net fluxes (Figure 1)

can largely be explained by variability in the gross production rates between these sites. At Konza‐1, the average gross consumption rates were −10 ± 3 for CH3Br and −430 ± 120 nmol m−2 d−1 for CH3Cl. Gross consumption rates increased (i.e., trended more negative) with volumetric soil moisture from 15 to 23% (sites B–F, R2 = 0.66 for CH3Br and 0.87 for CH3Cl) and decreased at the higher (30–31%) soil moisture sites of G and H, although these observations are based on limited data (n = 7 for each gas). [41] At Konza‐2 (Figure 6), gross production rates also varied widely between the three separate sites, with averages varying by a factor of 9 for CH3Br and by a factor of 100 for CH3Cl. The largest CH3Br production occurred at site I (Amorpha canescens), while the largest CH3Cl production occurred at site J (Amorpha fruticosa). Over the course of the Konza‐2 outing, the intrasite variability (including the pre‐ and postrainfall measurements) was much smaller than the intersite variability, although intrasite gross production fluxes did vary by up to a factor of 3 for CH3Br and by up to a factor of 5 for CH3Cl. Gross consumption rates of the methyl halides were much more constrained, similar to Konza‐1. Average gross consumption rates for CH3Br were: −9.4 ± 1.8, −13.6 ± 2.8, and −10.5 ± 1.7 nmol m−2 d−1 for sites I, J and K, respectively. Average gross consumption

10 of 15

G03026

RHEW: TALLGRASS PRAIRIE CH3Br AND CH3Cl FLUXES

rates for CH3Cl were −320 ± 60, −490 ± 90 and −410 ± 90 nmol m−2 d−1 for sites I, J and K, respectively. These limited ranges occurred despite spanning 24 h before simulated rainfall (n = 2 per site) to 48 h postrainfall (n = 5 per site). [42] In Konza‐2, the soil moisture (at 0–6 cm depth) increased 12 ± 5% VWC immediately following the simulated rainfall at the three sites (Figure 6). The soil moisture change (averaged postrainfall versus averaged prerainfall measurements) showed site‐specific differences: relative increases of 20% for site I, 10% for site J, and 103% for site K. The small changes at sites I and J are partly a result of high soil moisture measured 24 h prior to the experiment, presumably because of natural rainfall events that preceded the measurements. [43] As with the net fluxes, the response of gross methyl halide fluxes to the simulated rainfall did not show consistent trends across the sites. At site I, average gross production rates increased (post‐ to prerainfall) by 51% for CH3Br and 60% for CH3Cl, and average gross consumption rates increased as well (37% for CH3Br and 47% for CH3Cl). At site J, average gross production rates of CH3Br and CH3Cl increased (25% and 65%, respectively), but average gross consumption decreased (−25% and −15%, respectively). At site K, where soil moisture doubled postrainfall, average gross production rates increased 18% for CH3Br and decreased by 40% for CH3Cl, and gross consumption rates increased only 2% for CH3Br and decreased by 15% for CH3Cl.

5. Discussion 5.1. Versatility of the Stable Isotope Tracer Method [44] Both sources and sinks of CH3Br and CH3Cl exist in the tallgrass prairie, manifested by the range of positive and negative net fluxes. Because the net flux represents a competition between local sources and sinks, it is essential to separately determine the gross production and consumption fluxes to clarify the source of that variability. Although net CH3Br and CH3Cl fluxes are slightly greater in spiked versus unspiked chambers at Konza‐1, this enhancement is relatively minor. In other ecosystems where this method was tested, no statistically significant difference in net fluxes was observed between consecutively run spiked and unspiked chambers, including studies conducted in annual grasslands (n = 18 pairs), shortgrass steppe (n = 8 pairs) and oak‐savanna woodland (n = 3 pairs) [Rhew and Abel, 2007; Rhew et al., 2010; Teh et al., 2008]. The cumulative evidence indicates that the stable isotope tracer technique does not significantly alter the behavior of the system. 5.2. Model Choice is Important at Sites With Large Emissions [45] The results of the four‐model comparison demonstrate a sensitivity to curve fitting procedures and in particular, the error minimization technique employed. All four models performed well for sites where net consumption (or small to moderate net production) occurred, but model 1 was not designed for sites with large production rates and model 2 also performed poorly where high CH3Cl production was observed (sites I and J). At these sites, internal chamber

G03026

12

CH3Cl concentrations exceeded 5000 ppt (5 ppb), and gross CH3Cl production rates were over an order of magnitude greater than gross consumption rates. The high concentrations have varying degrees of leverage in the curve fitting procedure. For example, model 2 minimized the difference at high 12CH3Cl concentrations at the expense of fitting the smaller changes in 13CH3Cl concentrations, leading to underpredicted uptake rates. This model shortcoming was not evident for the case of CH3Br, but chamber CH3Br concentrations were less than 200 ppt in all but one experiment (Konza‐1, site H). [46] This does not invalidate the usage of model 2, which was applied previously in situations where gross production rates were within the same order of magnitude of gross consumption rates [Rhew and Abel, 2007; Rhew et al., 2003, 2010]. In those situations, model 2 is predicted to perform equally well as models 3 and 4. Under situations where large emissions of methyl halides are observed, however, model 3 or 4 is more appropriate. Recently published isotope tracer studies at field sites with large emissions of methyl halides include coastal salt marsh [Rhew and Mazéas, 2010] and rice paddies [Khan et al., 2011]. These studies, however, utilized model 4 in anticipation of the results presented here. 5.3. Amorpha and Panicum Species as Sources of Methyl Halides [47] While Andropogon gerardii (big bluestem) and Schizachyrium scoparium (little bluestem) were present in all chamber sites in the tallgrass prairie, the largest observed gross production rates of CH3Br and CH3Cl are associated with plant species of the genus Amorpha Interestingly, the Amorpha canescens (leadplant) at sites H and I showed the higher production of CH3Br while the Amorpha fruticosa (indigo bush) at site J showed the highest production of CH3Cl. At sites H, I and J, the gross CH3Br production rates were 485 ± 10, 277 ± 71, and 76 ± 15 nmol m−2 d−1, respectively, and gross CH3Cl production rates were 9200 ± 200, 7600 ± 1900, and 25,000 ± 7500 nmol m−2 d−1, respectively. [48] These gross production rates are greater than most emission rates measured in temperate terrestrial ecosystems. For example, gross production rates from the highest producing Atriplex canescens (four‐wing saltbush) sites on the shortgrass steppe were 5 nmol m−2 d−1 for CH3Br and 1250 nmol m−2 d−1 for CH3Cl [Teh et al., 2008]. In California grasslands, the maximum observed gross production rates were 23 nmol m−2 d−1 for CH3Br and 535 nmol m−2 d−1for CH3Cl (net fluxes at the sparsely vegetated alkaline dry playa, however, were up to 2650 nmol m−2 d−1 for CH3Br and 42100 nmol m−2 d−1 for CH3Cl). Seasonal maxima of net emissions observed in California rice fields ranged from 130 to 280 nmol m−2 d−1 for CH3Br and 1300– 2400 nmol m−2 d−1 for CH3Cl [Khan et al., 2011; Redeker et al., 2000]. [49] Surprisingly, the gross production rates of methyl halides from the Amorpha sites are comparable to temperate coastal salt marshes, which are perhaps the largest known natural terrestrial source of methyl halides outside of the tropics and which are regularly replenished with halides from seawater. For example, the temporally and spatially weighted mean emission rates from two Scotland salt marshes were 75 ± 11 (maximum of 860) nmol m−2 d−1 for

11 of 15

G03026

RHEW: TALLGRASS PRAIRIE CH3Br AND CH3Cl FLUXES

CH3Br and 315 ± 126 (maximum of 20,100) nmol m−2 d−1 for CH3Cl [Blei et al., 2010a]. However, southern California salt marshes had much larger emissions, with mean net emission rates of 1000–1800 nmol m−2 d−1 for CH3Br and 22000–24000 nmol m−2 d−1 for CH3Cl [Manley et al., 2006; Rhew et al., 2000]. Salt marsh sites containing Frankenia grandifolia and Batis maritima had emission rates up to an order of magnitude larger. [50] The smaller gross production rates of CH3Br (33 ± 14 nmol m−2 d−1) and CH3Cl (312 ± 200 nmol m−2 d−1) associated with Panicum virgatum (switchgrass, site K) should also be considered, as this genus has been proposed as a major feedstock for the production of biofuels. If our measurements were representative of switchgrass during a 180 day growing season, the conversion of 3.1–21.3 million hectares of prairie to switchgrass [McLaughlin and Kszos, 2005] would correspond to a gross emission rate of 0.02– 0.12 Gg CH3Br and 0.09–0.60 Gg CH3Cl, minus the gross emission rate of the grassland it displaced. These measurements suggest that the hypothetical conversion of grassland to switchgrass would not cause a significant increase (