Temperate forest methane sink diminished by tree emissions

Introduction

Upland (free-drained) soils are estimated to consume 20–45 Tg methane (CH₄) per year (Topp & Pattey, 1997; Dutaur & Verchot, 2007; Kirschke et al., 2013; Schlesinger & Bernhardt, 2013), a sink comparable to the rate of CH₄ accumulation in the atmosphere and, therefore, capable of influencing the radiative forcing caused by this potent greenhouse gas. Global CH₄ budgets, Earth system models, and carbon accounting policies have generally assumed that the role of upland forests can be determined by measuring the rate of CH₄ fluxes at the soil surface. This assumption is problematic in forests where soils but not whole trees can be enclosed in gas flux chambers, the most common technique for quantifying upland CH₄ fluxes. A variety of evidence now makes it clear that all biological surfaces in upland forests have the potential to exchange CH₄. These include reports of novel sources of CH₄ emissions in nominally upland ecosystems (Keppler et al., 2006; Martinson et al., 2010; Lenhart et al., 2012), eddy flux evidence of hot spots or hot moments of forest CH₄ emissions (do Carmo et al., 2006; Shoemaker et al., 2014) and elevated CH₄ concentrations in tree stems (Bushong, 1907; Zeikus & Ward, 1974; Covey et al., 2012).

Despite significant advances in identifying novel sources of CH₄ in upland forests, the consequences for upland forest CH₄ budgets have been highly speculative due to a lack of in situ observations of CH₄ emissions across surfaces other than soils. Abiotic emissions driven by UV radiation (Keppler et al., 2008; Megonigal & Guenther, 2008), fungal emissions from wood surfaces (Lenhart et al., 2015) and microbial emissions from wood cores (Zeikus & Ward, 1974; Wang et al., 2016) have been measured in laboratory settings. Emissions from living tree stems were modeled from CH₄ concentration gradients and estimates of gas diffusion constants (Covey et al., 2012). Until recently it was difficult to judge whether these potential upland CH₄ sources are quantitatively important because there were no direct, in situ emissions data other than from tropical forest tank bromeliads (Martinson et al., 2010). This changed recently with direct CH₄ flux measurements from the living stems and shoots of three temperate forest tree species – Populus davidiana, Carya cathayensis and Larix gmelinii – in China (Wang et al., 2016); the stems and shoots of Pinus sylvestris in a European boreal forest (Machacova et al., 2016); the stems of several hardwood species in a temperate hardwood forest in eastern North America (Warner et al., 2017); and stems of Fagus sylvatica in a temperate European forest (Maier et al., 2017).

Representation of CH₄ emissions from upland ecosystems has been acknowledged in global models and budgets (Saunois et al., 2016), but remains limited by a lack of in situ flux measurements from non-soil surfaces. The global contributions of CH₄ from abiotic production, fungi and epiphytes are difficult to estimate, but are expected to be too small to adequately explain the potential source–sink imbalance of 8–46 Tg yr⁻¹ (Kirschke et al., 2013). Coarse woody debris is a source of CH₄ emissions from termites and microorganisms that may be significant (Carmichael et al., 2014; Covey et al., 2016; Warner et al., 2017). Living tree stems are potentially a large CH₄ source in upland forests (Covey et al., 2012; Machacova et al., 2016; Wang et al., 2016; Warner et al., 2017), and are known to be a significant CH₄ source in floodplain forests (Terazawa et al., 2007, 2015) and wetland forests (Pulliam, 1992; Gauci et al., 2010; Pangala et al., 2013). In floodplain and wetland forests, the source of tree-emitted CH₄ has been assumed to be biological production in saturated soils with subsequent transport through aerenchyma tissue or transpiration. But in upland forests there is evidence that the CH₄ emitted from living tree stems is produced biologically in wood (Bushong, 1907; Zeikus & Ward, 1974; Covey et al., 2012; Wang et al., 2016). Distinguishing between sources (e.g. wood, soil) is critical for global CH₄ models.

Our objectives were to quantify CH₄ emissions from tree stems in an upland temperate forest and to evaluate whether soil moisture, tree species and tree activity regulate the size and timing of emissions. We hypothesized that upland trees emit CH₄ from stems; soils are a source of stem-emitted CH₄; and stem emissions offset a meaningful fraction of net CH₄ consumption by soils.

Materials and Methods

Flux measurements

Seventeen trees were fitted with opaque acrylic chambers (Ryan, 1990; Fig. S2). Each tree was paired with a soil gas flux chamber within 1 m of the base. Stem chambers were permanently fixed at 30–60 cm above the soil and secured with elastic shock cord. An airtight seal between the chamber and the stem was created with closed-cell neoprene foam, sealed with non-VOC caulk normally used for making dental molds (Examix™; GC America, Alsip, IL, USA). Soil rings were constructed from 30.5 cm-diameter schedule 80 PVC pipe buried 5 cm into the soil surface. Chambers were mounted a minimum of 1 wk before taking flux measurements, and remained in place for the duration of the study.

Gas concentrations were measured using a portable Off-Axis Integrated Cavity Output Spectroscopy (OA-ICOS) instrument (Los Gatos Research, Los Gatos, CA, USA). The instrument measures CH₄ in a range of 0.01 to 100 ppm with a precision of 0.002 ppm at 0.5 Hz, and CO₂ in a range of 200 to 20 000 ppm with a precision of 0.3 ppm. The closed system drew headspace gas from the chamber, measured CH₄ concentration non-destructively, then returned the gas to the chamber. Each flux was measured over 5–10 min and generated ≥ 150 observations (Fig. S3).

Stem and soil CH₄ and CO₂ flux rates were measured monthly from May to September 2014 for a total of five campaigns. In the first campaign (22–23 May 2014), fluxes were measured once on seven individual trees. In the second campaign (20 June 2014) a set of 10 completely different trees were measured, such that there was no overlap between the individual trees measured in the May and June campaigns. Thereafter, all 17 trees (Table S1) were measured once during each of the final three campaign (i.e. 8–10 July, 20–29 August and 17–22 September 2014). Each tree measurement was paired with a single measurement of soil CH₄ and CO₂ flux taken from the adjacent soil chamber. The full dataset (Table S2) consists of four measurements (one on each of four different campaigns) for every tree-soil pair. There were 68 tree observations and 68 soil observations for a total of 136 flux measurements.

Both CO₂ and CH₄ were quantified from ≥ 150 concentration observations. We removed the first 20% of the observations to eliminate artifacts caused by closing the lid, estimated the slope using linear regression, and calculated gas flux by the following equation:

(where F is the flux in μl m⁻² h⁻¹, P is atmospheric pressure, T is temperature, R is the universal gas constant, A is the collar surface area and V is the volume of the air enclosed by the chamber). Air temperature was measured by the OA-ICOS unit on gas circulating between the unit and the chamber. Atmospheric pressure was based on a nearby weather station (< 1 km).

We constructed an automated system for high-frequency tree CH₄ flux measurements to gain insights on the source and mechanism of CH₄ emissions from upland trees. The system was installed on two new trees that had not been measured previously. The elevation of the trees was similar to those at the highest elevation in the transect study, and separated by c. 300 m. During a 3-d period (28–31 July 2014) CH₄ and CO₂ fluxes were measured at 45 min intervals from the trunk of a single Liriodendron tulipifera at three heights above the soil surface (75, 165 and 245 cm) and a single F. grandifolia at one height (75 cm). Automated measurements were made using the same chamber design (Fig. S2), modified with a lid that was opened and closed by pneumatic cylinders and solenoids controlled by an Arduino Mega microcontroller. The manifold sampled from one chamber at a time in a closed loop for 8 min, flushing with ambient air for 3 min between measurements.

Results and Discussion

Using the criterion that fluxes with P-values < 0.05 were significant, 65 of the 68 tree CH₄ flux observations were highly significant (P < 0.001), and three were statistically indistinguishable from zero (P > 0.05). Each of the seven upland tree species in the study was capable of emitting CH₄, namely F. grandifolia, Liriodendron tulipifera, C. tomentosa, Q. velutina, Q. michauxii, A. rubrum and Liquidambar styraciflua (Table S2). In six cases, the fluxes indicated low rates of CH₄ consumption (< −0.07 μmol m⁻² stem h⁻¹) by tree stems. Using the more conservative criterion that fluxes with r² < 0.80 are indistinguishable from zero, 46 of the 68 observations were greater than zero, all of which were positive fluxes to the atmosphere. All of the soil flux regression models had P-values < 0.001 and r² > 0.80 (Table S2). Soil CH₄ fluxes generally showed net CH₄ consumption from the atmosphere as expected in an upland forest, with one exception in which the flux was positive (Fig. 1). Thus, different terrestrial-atmosphere interfaces (soils vs tree stems) in an upland forest tended to simultaneously act as either a CH₄ sink (soils) or a CH₄ source (stems), each counteracting the influence of the other on net ecosystem CH₄ emissions.

Soil CH₄ flux averaged over the growing season was −4.52 ± 0.64 μmol m⁻² soil h⁻¹ (mean ± 95% CI), which is similar to the global average for temperate forest soils of −4.07 μmol m⁻² soil h⁻¹ (Dutaur & Verchot, 2007). Stem flux averaged 1.59 ± 0.88 μmol m⁻² stem h⁻¹ (P ≤ 0.0001). These averages and errors were the same whether calculated using the more relaxed or conservative criteria for assigning fluxes a value of zero, reflecting the fact that the highest rates of flux also had high r² values (Fig. S4). For this reason, we restrict discussion from here forward to the more conservative estimates. Tree emissions were positive and significantly greater than zero in June (9.53 ± 2.87), July (0.33 ± 0.16), August (0.19 ± 0.13) and September (0.19 ± 0.13) (P ≤ 0.002), but not in May (0.09 ± 0.10) (P = 0.13; all units μmol m⁻² stem h⁻¹). Soil fluxes were significantly different from zero (P ≤ 0.02) for all months.

Of the 46 stems with the highest CH₄ emissions (r² > 0.80), 45 were paired with a soil that was a net CH₄ sink. These data and those from a second North American temperate forest (Warner et al., 2017), two temperate upland forests in Asia (Wang et al., 2016), a boreal forest in Europe (Machacova et al., 2016), a temperate forest in Europe (Maier et al., 2017) and observations of super-ambient CH₄ concentrations inside temperate forest tree stems (Covey et al., 2012), collectively suggest that the size of the CH₄ sink ascribed to upland forests has been overestimated.

This study, and that of Warner et al. (2017), were similar in many respects. They were conducted at roughly the same time (2014 growing season), using similar methods (static chambers, Los Gatos OA-ICOS gas analyzer), similar sampling effort (16–17 trees measured 1–2 times per month), and on four of the same tree species. Climate is similar between the sites because they are at the same elevation (near sea level) and separated by 100 km, and both forests are on loamy soils classified as Typic Hapludults. Average tree stem CH₄ emissions over the growing season in the Warner study (0.40 ± 0.18 μmol CH₄ m⁻² stem h⁻¹) were comparable to the present study when the June data are excluded (0.23 ± 0.23 μmol CH₄ m⁻² stem h⁻¹), suggesting that tree CH₄ emissions may be similar when stratified by species, climate and soil characteristics under most conditions. However, stem CH₄ emissions in the present study rose an order of magnitude in the June sample (9.53 ± 2.87 μmol m⁻² stem h⁻¹), a ‘hot moment’ event of uncertain duration that was not observed by Warner et al. (2017). Because both studies sampled emissions over short time periods (minutes) at long intervals (2–4 wk), we cannot state whether the difference between studies is real or a limitation of the sampling designs. Regardless, the transient increase in stem CH₄ emissions observed in the present study is an illustration of the potential for high temporal variation in this process.

A small but growing collection of upland tree CH₄ emission studies indicates substantial spatial and temporal variability that must be resolved in order to account for tree emissions in upland forest CH₄ budgets. A 3-month study of the conifer Pinus sylvestris reported median rates of stem CH₄ emissions on a wet site that were 1–2 orders of magnitude lower than those in the present study (0.01–0.001 μmol m⁻² stem h⁻¹) (Machacova et al., 2016). At the high extreme is a Populus davidiana forest where annual average rates ranged 5.3–6.4 μmol m⁻² stem h⁻¹ (Wang et al., 2016), suggesting sustained rates comparable to the high June rate in the present study. Average rates from the present study (1.59 μmol m⁻² stem h⁻¹), a nearby temperate forest (Warner et al., 2017; 0.40 μmol CH₄ m⁻² stem h⁻¹) and a high emitting site in a European temperate forest (1.87 μmol CH₄ m⁻² stem h⁻¹ (Maier et al., 2017) fall between these extremes, but toward the upper end of the range. An important step toward improved experimental designs and scaling exercises is to understand the sources of this variability.

There were no statistical relationships between tree CH₄ emissions and stem diameter, soil moisture, air temperature or soil temperature that could be used for scaling fluxes (P > 0.05), consistent with the results of Warner et al. (2017) in a similar forest. By contrast, stem emissions were positively correlated to temperature in a Populus davidiana forest (Wang et al., 2016), perhaps because relatively high summer emissions and low winter emissions produced a wider range of rates. Previous studies reported that much of the variation in stem CH₄ emissions is related to tree species (Covey et al., 2012; Wang et al., 2016; Warner et al., 2017). In the present study there were no significant differences across all species (P = 0.07); however, for the species with the largest sample sizes – F. grandifolia (n = 28), Liquidambar styraciflua (n = 12) and Liriodendron tulipifera (n = 12) – there were significant differences (P = 0.04), with F. grandifolia supporting lower rates than the other species (Fig. 2). Such differences may be linked to characteristics such as stem morphology or disease resistance (Warner et al., 2017).

In the present study, there was variation across individuals and time that was not clearly related to species. Of the 17 trees in the transect study, five individuals contributed 84% of the cumulative CH₄ emitted from all stems (May–September fluxes scaled to stem diameter; Table S2). The five trees belonged to four species – F. grandifolia, Liriodendron tulipifera, Q. velutina and Q. michauxii – each of which contributed 12–23% of the total. Other than C. tomentosa of which there was just one stem, every species contributed > 10% of cumulative stem CH₄ emissions. About half of the individuals (8 of 17) were consistent emitters with measurable fluxes (i.e. P < 0.05 and r² > 0.80) during every sample event, while the remainder emitted CH₄ intermittently on three dates (two individuals), two dates (three individuals), or one date (two individuals). Two individuals never emitted significant amounts of CH₄. A similar result was reported in a study of 10 F. sylvatica stems across two Central European forests where nearly all stem CH₄ was emitted by a single tree (Maier et al., 2017). It is clear that robust estimates of stem CH₄ emissions at the stand scale will require a combination of large sample sizes and high-frequency measurements, and improved techniques such as eddy covariance flux capable of quantifying small fluxes over large areas.

Despite empirical evidence of several potential CH₄ sources in living upland trees, there is no evidence in this study or previous studies that can definitively assign the CH₄ emitted from trees to a single source. Potential sources include microbial production inside the tree stem (Zeikus & Ward, 1974; Covey et al., 2012; Wang et al., 2016), on stem bark (Lenhart et al., 2015) and in subsurface soils (Megonigal & Guenther, 2008; Machacova et al., 2016; Maier et al., 2017), and abiotic, UV-driven production by leaves and other tree surfaces (Keppler et al., 2008). UV-driven CH₄ production was not a source in the present study because emissions were measured in opaque chambers. Fungi on tree stems are unlikely to be an important source because reported rates are far less than those observed here (Lenhart et al., 2015). In the present study, evidence against a soil CH₄ source (thus consistent with a wood source) is the fact that stem emissions were not statistically related to soil moisture, soil temperature or water table depth (P > 0.05). Also, a rise in the water table in late August (Fig. S1) did not increase August emissions (Fig. 1). The trees in this study were not surveyed for rot or wet wood (e.g. Wang et al., 2016), but surveys of Atlantic coast temperate forest tree stems show that super-ambient internal CH₄ concentrations are commonplace and suggest an internal source (Covey et al., 2012).

Soils are a potential source of tree CH₄ consistent with some observations in this study (Megonigal & Guenther, 2008). Methane transport via transpiration is consistent with declining CH₄ emissions with increasing height on the same tree (Fig. 3). This pattern is expected for a soil CH₄ source (Terazawa et al., 2007; Gauci et al., 2010; Pangala et al., 2013, 2015), while an internal (wood) source is expected to peak at a height well above the base (Covey et al., 2012). We observed the same pattern in stem CO₂ emissions (Fig. 3) and found that CO₂ and CH₄ were correlated (r = 0.35, P < 0.01); similar observations have been observed for stem CO₂ emissions and interpreted as soil CO₂ entrained in the tree transpiration stream (Teskey et al., 2008). Furthermore, a soil source of CH₄ is expected to respond to changes in soil water content. Although we did not detect a statistical relationship between soil moisture content and stem CH₄ emissions, there were two instances when high emission rates followed a precipitation event. The first was the June sample in the transect study when all 10 individuals emitted CH₄ at rates one order of magnitude higher than in other months. The June sample occurred during a period of high volumetric soil moisture content (> 30%) and warm soil temperatures compared to mid-May when water content was also high but temperature was lower (Fig. 1). The combination of warm temperatures and high soil water content may have simultaneously increased CH₄ production and decreased methanotrophy due to O₂ limitation (Maier et al., 2017), as evidenced by the fact that soil CH₄ uptake rates were also lowest in June (Fig. 1). The second instance was in the 3-d record of automated flux measurements when rates in two individuals declined with time (Figs 4, S5); this occurred during a rain-free period that followed a damp period (2 mm rain over 5 d) and may reflect the influence of declining soil water content.

We observed striking differences in diurnal patterns of stem CH₄ emissions for the two individuals fitted with automated flux chambers (Fig. 4). The relatively subtle diurnal variation in the F. grandifolia tree compared to the Liriodendron tulipifera tree (Fig. S5) may reflect different CH₄ sources, different axial and radial stem diffusion rates, or different sinks (e.g. CH₄ oxidation), all of which influence stem CO₂ emissions (Teskey et al., 2008). Sorz & Hietz (2006) found that O₂ diffusion rates tend to increase in the order: conifers > ring-porous species > diffuse-porous species, but both Liriodendron tulipifera and F. grandifolia are diffuse-porous. With just one stem of each species, the differences may simply reflect characteristics of these particular individuals. However, mean CH₄ emissions were also higher for Liriodendron tulipifera in the well replicated transect study (Fig. 2).

High frequency flux measurements and time series analysis have the potential to be powerful analytical tools to disentangle tree CH₄ sources and transport pathways. The Liriodendron tulipifera fitted with an automated chamber showed a diurnal cycle with peak emissions in late afternoon (CH₄ = 16:20 h, CO₂ = 18:13 h; Table S3). The timing of this peak falls after peak sap flux density (12:00 h) and near the transpiration-driven minima in tree diameter (16:30 h), as measured 1 yr later on the same Liriodendron tulipifera tree (Herrmann et al., 2016). Long-term records of near-continuous CH₄ emissions combined with knowledge of the kinetics of gas and heat transfer in trees will enable inference of CH₄ sources.

The consequences of our observations for upland forest CH₄ budgets are difficult to judge because of the limited sample size of soils, trees, tree species and tree surface types (i.e. trunks only). Nonetheless, it is useful to estimate the balance of soil CH₄ consumption and tree CH₄ emissions because in situ tree and soil CH₄ fluxes have been measured simultaneously and scaled to the stand level in three other upland forests (Machacova et al., 2016; Wang et al., 2016; Warner et al., 2017). A simple scaling exercise suggests that upland tree emissions offset 6% of the soil sink over the May–September growing season, much of which occurred in a single June event of uncertain duration when this forest may have become a transient net source of CH₄ at a rate of 2.14 μmol m⁻² soil h⁻¹. A more conservative approach is to eliminate the June sample from the calculation, in which case tree emissions offset 1% of the soil sink. The range of estimates from our study (1–6%) brackets the only other estimate (3.5%) for a North American temperate forest (Warner et al., 2017). Both estimates are far lower than an estimate from a temperate deciduous forest in China where tree emissions offset 63% of soil CH₄ sink (Wang et al., 2016), but higher than a dry Pinus sylvestris site where the offset was 0.8% (Machacova et al., 2016). This wide range of estimates suggests that upland tree CH₄ emissions are highly variable and illustrates the need for new sampling and scaling strategies for quantifying local, regional and global upland forest CH₄ dynamics. The sampling challenge may be greater depending on whether small stems and leaves emit significant amounts of CH₄ (Machacova et al., 2016), consume CH₄ (Sundqvist et al., 2012), or have neither effect (Wang et al., 2016), all of which have been observed. It should be noted that while the degree to which stem emissions offset the soil sink is highly uncertain, in all cases upland forests remained net sinks for atmospheric CH₄.

Author contributions

J.P.M. conceived of the research project. J.P.M. and S.P. designed the transect study. S.P. designed the automated chamber study. S.P. performed the research. S.P. and J.P.M. analyzed data, interpreted data and wrote the manuscript.