High Temporal and Spatial Variability of Atmospheric-Methane Oxidation in Alpine Glacier Forefield Soils

ABSTRACT

Glacier forefield soils can provide a substantial sink for atmospheric CH₄, facilitated by aerobic methane-oxidizing bacteria (MOB). However, MOB activity, abundance, and community structure may be affected by soil age, MOB location in different forefield landforms, and temporal fluctuations in soil physical parameters. We assessed the spatial and temporal variability of atmospheric-CH₄ oxidation in an Alpine glacier forefield during the snow-free season of 2013. We quantified CH₄ flux in soils of increasing age and in different landforms (sandhill, terrace, and floodplain forms) by using soil gas profile and static flux chamber methods. To determine MOB abundance and community structure, we employed pmoA gene-based quantitative PCR and targeted amplicon sequencing. Uptake of CH₄ increased in magnitude and decreased in variability with increasing soil age. Sandhill soils exhibited CH₄ uptake rates ranging from −3.7 to −0.03 mg CH₄ m⁻² day⁻¹. Floodplain and terrace soils exhibited lower uptake rates and even intermittent CH₄ emissions. Linear mixed-effects models indicated that soil age and landform were the dominating factors shaping CH₄ flux, followed by cumulative rainfall (weighted sum ≤4 days prior to sampling). Of 31 MOB operational taxonomic units retrieved, ∼30% were potentially novel, and ∼50% were affiliated with upland soil clusters gamma and alpha. The MOB community structures in floodplain and terrace soils were nearly identical but differed significantly from the highly variable sandhill soil communities. We concluded that soil age and landform modulate the soil CH₄ sink strength in glacier forefields and that recent rainfall affects its short-term variability. This should be taken into account when including this environment in future CH₄ inventories.

IMPORTANCE Oxidation of methane (CH₄) in well-drained, “upland” soils is an important mechanism for the removal of this potent greenhouse gas from the atmosphere. It is largely mediated by aerobic, methane-oxidizing bacteria (MOB). Whereas there is abundant information on atmospheric-CH₄ oxidation in mature upland soils, little is known about this important function in young, developing soils, such as those found in glacier forefields, where new sediments are continuously exposed to the atmosphere as a result of glacial retreat. In this field-based study, we investigated the spatial and temporal variability of atmospheric-CH₄ oxidation and associated MOB communities in Alpine glacier forefield soils, aiming at better understanding the factors that shape the sink for atmospheric CH₄ in this young soil ecosystem. This study contributes to the knowledge on the dynamics of atmospheric-CH₄ oxidation in developing upland soils and represents a further step toward the inclusion of Alpine glacier forefield soils in global CH₄ inventories.

INTRODUCTION

Aerobic oxidation of the potent greenhouse gas methane (CH₄) is one of many important ecosystem services provided by soils (1, 2). It is mediated by methane-oxidizing bacteria (MOB), which can utilize CH₄ as a sole source of carbon and energy (3). The activity of MOB in soils can substantially attenuate CH₄ emissions to the atmosphere from natural and anthropogenic sources, such as wetlands, rice paddies, and landfills (4–6). Moreover, in well-drained (upland) soils, so-called “high-affinity” MOB are capable of utilizing atmospheric CH₄ at ambient concentrations (≤1.8 μl liter⁻¹) (7). As a result, upland soils are usually a sink for atmospheric CH₄. In fact, they are CH₄'s only known terrestrial sink, contributing ∼5 to 15% of the total loss of CH₄ from the atmosphere (8–10).

As cultivation attempts with high-affinity MOB have been mostly unsuccessful, to date (Methylocystis sp. strain SC2 is the only potential high-affinity MOB isolate [11]), molecular and biochemical methods are employed for their identification in upland soils. Identification is typically based on the amplification of the pmoA gene, a widely used MOB biomarker that closely reflects phylogenies based on 16S rRNA gene sequences (e.g., see reference 12). The pmoA gene encodes a subunit of the particulate form of the enzyme methane monooxygenase (pMMO), which catalyzes the first step in the oxidation of CH₄ (3, 13). Based on pmoA, high-affinity MOB have mainly been assigned to the upland soil cluster (USC) alpha and gamma clades, and it was noted that they are distantly related to two groups of cultivable MOB, i.e., type II and type I, respectively (14–16). High-affinity MOB have been studied intensively in mature upland soils (e.g., tundra, forests, grasslands, savannahs, and deserts [17–21]), but only limited information is available on their performance in young, developing upland soils (e.g., recent volcanic deposits or proglacial sediments).

Whereas developing upland soils are still excluded from national and global CH₄ inventories (e.g., see references 8 and 22), recent studies indicate that they may provide a substantial sink for atmospheric CH₄. For example, Hawaiian volcanic deposits (andosols [FAO soil classification]) were found to be a sink for atmospheric CH₄ already 40 years after deposition. They exhibit CH₄ uptake rates (expressed by convention as negative soil-atmosphere CH₄ flux values) of −1.8 to −0.7 mg CH₄ m⁻² day⁻¹, similar to those of mature forest and grassland soils (22). Atmospheric-CH₄ oxidation was also detected in proglacial sediments from both Arctic (23) and Alpine (24, 25) environments. Specifically, young Alpine proglacial sediments, referred to here as glacier forefield soils (lithosols [FAO soil classification]), can provide a substantial sink for atmospheric CH₄ already at an early stage of soil development (<15 years; up to −0.7 mg CH₄ m⁻² day⁻¹) (26). As the extent of glacier forefields is expected to increase further with ongoing glacial retreat (26, 27), it is important to better understand the factors modulating community composition and activity of high-affinity MOB in this developing soil environment.

Glacier forefields are heterogeneous and dynamic environments which exhibit features that render their soils interesting but challenging systems for the study of microbial structures and functions. First, glacier forefields often feature a well-defined sequence of soil ages (soil chronosequence) (e.g., see references 28 to 30). This is a result of glacial retreat, when sediments are successively subjected to a radical shift from a subglacial to a proglacial environment, whereby soil formation is initiated. Thus, a soil chronosequence is a good model ecosystem for studies of microbial primary succession (e.g., see reference 31). Second, glacier forefield soils can be derived from diverse bedrock types. As bedrock type is a key factor determining nutrient availability, it may shape microbial communities in glacier forefield soils (31, 32). Third, the cooccurrence of different geomorphological processes (e.g., glacial erosion and deposition, debris movement, and physical and chemical weathering) can lead to the formation of numerous proglacial landforms (33, 34). Individual landforms often exhibit specific physicochemical properties, which may affect microbial community composition and activity. Finally, glacier forefield soils are also subject to dramatic short-term and seasonal variability in, e.g., physical parameters such as soil temperature and water content, which are known to affect the performance of microbial communities, including high-affinity MOB (e.g., see reference 9). Whereas such variability is usually low during the snow-covered season (35, 36), with the melting of snow the glacier forefield soils undergo sudden changes in soil water content, and they can be subject to large fluctuations in temperature and water regimens throughout the snow-free season (e.g., see references 37 to 39).

In a recent field study, we investigated atmospheric-CH₄ oxidation and MOB community composition along soil chronosequences (soil age, 6 to 120 years) in two Swiss glacier forefields situated on contrasting (siliceous versus calcareous) bedrock types (39). Soil age was found to be the main factor affecting atmospheric-CH₄ uptake, which increased with increasing soil age and ranged from −2.2 to −0.08 mg CH₄ m⁻² day⁻¹. In contrast, observed differences in MOB community composition were related mainly to bedrock type rather than to soil age, indicating that distinct, low-diversity MOB communities provided similar ecosystem services in the two forefields. However, as field sampling was restricted to two time points during the snow-free season and a single proglacial landform, the temporal variability of and effects of different landforms on atmospheric-CH₄ oxidation and MOB community composition in glacier forefield soils remained unknown.

Thus, the objectives for the present study were (i) to investigate temporal variability in atmospheric-CH₄ oxidation and MOB community composition as a function of soil age along an established glacier forefield soil chronosequence, (ii) to assess the effects of different landforms on the soil CH₄ sink within a single soil age class, and (iii) to provide an improved, high-resolution analysis of MOB diversity in glacier forefield soils. To this end, we conducted field campaigns to quantify soil-atmosphere CH₄ flux throughout the snow-free season of 2013 (Fig. 1), and targeted amplicon sequencing was used to assess MOB diversity and community composition in soil samples collected during this period. We employed linear mixed-effects (LME) models to test the dependence of the response variables soil-atmosphere CH₄ flux and MOB abundance on the model predictors soil age, landform, sampling time point, soil temperature, soil water content, and cumulative rainfall.

FIG 1.

Sampling locations at the Griessfirn Glacier forefield, mapped using Swiss square-projection coordinates (CH1903). Soil chronosequence locations are delimited according to soil age class (A, 0 to 20 years; B, 20 to 50 years; and C, 50 to 120 years). (Inset) Landform locations distinguish sandhill (S), terrace (T), and floodplain (F) landforms.

RESULTS

Soil physical parameters.

Soil water content was low at all sampling locations and time points, indicating that glacier forefield soils remained relatively dry throughout the snow-free season (Fig. 2a). Data from locations A to C indicate that topsoil was generally drier than bulk soil, with seasonal mean water contents ranging from 0.06 to 0.09 m³ m⁻³ for topsoil and from 0.10 to 0.12 m³ m⁻³ for bulk soil. Furthermore, topsoil water content increased with increasing soil age (from locations A to C) (Fig. 2a). The highest temporal variability in soil water content at individual locations was measured in soil age class A (topsoil, 0.04 to 0.09 m³ m⁻³; bulk soil, 0.06 to 0.19 m³ m⁻³) (data not shown). Among landform locations, the highest topsoil water content was observed at location F sites (Fig. 2a). Water content data for location S sites agreed reasonably well with topsoil data measured in close proximity, at location B sites, although different measurement techniques were employed.

FIG 2.

Heat maps showing median values for volumetric soil water content (a), soil temperature (b), and soil-atmosphere CH₄ flux (c) measured at soil chronosequence locations (A, B, and C) and landform locations (S, T, and F). For soil temperature and water content, “t” refers to topsoil measurements, and “b” refers to averaged (bulk soil) measurements across all depths. The column at right shows seasonal mean values ± 1 SD; white areas indicate that no measurements were performed on the corresponding dates.

After an initial increase in July following snow melting, the soil temperature decreased during the remainder of the snow-free season (from August to October) (Fig. 2b). During this period, the median topsoil temperature at locations A to C decreased ∼12°C, on average, whereas the bulk soil temperature decreased ∼8°C, on average. Moreover, bulk soil temperature moderately increased with increasing soil age (from locations A to C).

Rainfall events were quite evenly distributed throughout the sampling season, with a maximum of six dry days between any two consecutive events (see Fig. S2 in the supplemental material). Calculated cumulative rainfall rates ranged from 0.7 to 13.2 mm day⁻¹, with a mean value of 5.1 mm day⁻¹.

Soil-atmosphere CH₄ flux. (i) Chronosequence locations A to C.

Nearly all CH₄ concentrations measured in soil gas profiles were subatmospheric (example profiles are shown in Fig. S3), with a few measurements in the youngest soils falling at or slightly above the atmospheric value. As a result, soils at locations A to C exhibited mostly net CH₄ uptake during the sampling season (Fig. 2c). Median fluxes on individual sampling dates ranged from −2.34 to −0.03 mg CH₄ m⁻² day⁻¹, with no apparent temporal trend. However, CH₄ uptake increased substantially with increasing soil age. For example, the seasonal mean uptake rate was −0.30 mg CH₄ m⁻² day⁻¹ for location A sites and −1.31 mg CH₄ m⁻² day⁻¹ for location C sites (Fig. 2c). Conversely, the temporal variability of soil CH₄ concentrations and flux at individual locations was highest in the youngest soils (location A sites) and decreased with increasing soil age (Fig. S3). The highest spatial variability in CH₄ flux was also measured at location A sites (not shown).

(ii) Landform locations S, T, and F.

The three landforms exhibited substantial differences in soil-atmosphere CH₄ flux (Fig. 2c). Sites at location S exhibited substantial CH₄ uptake throughout the sampling season (median CH₄ flux of −0.54 to −0.11 mg CH₄ m⁻² day⁻¹). However, on comparing CH₄ uptake rates between adjacent locations S and B (Fig. 1), estimates for location B sites were ∼3-fold larger than the CH₄ uptake measured at location S sites. This may be due in part to the different methods employed to estimate/measure CH₄ flux. A slight apparent trend of decreasing CH₄ uptake with the ongoing sampling season at location S sites (Fig. 2c) was refuted by the LME model (see below). Uptake of CH₄ measured at locations T and F was substantially less than that at location S. In fact, positive CH₄ flux values for locations T and F even indicated intermittent net CH₄ emissions, particularly early in the sampling season (Fig. 2c). Similar spatial variabilities in CH₄ flux were observed for all landforms (not shown).

Abundance of MOB.

High-quality genomic DNA was isolated from all samples but S1-Jul25, which was consequently excluded from all further molecular analyses. pmoA copy numbers spanned 3 orders of magnitude, ranging from 8 × 10² to 5 × 10⁵ copies per g of dry soil (Fig. 3). At locations A to C, pmoA copy numbers increased substantially with soil age, and the highest copy numbers were detected in location C samples. The highest pmoA copy numbers among different landforms were obtained for location S samples (similar in magnitude to those for location B samples), whereas landform location F samples exhibited the lowest pmoA copy numbers, which were slightly lower than the copy numbers for location T samples. Notably, pmoA copy numbers for locations C and F differed significantly from those for other sampling locations (A, B, S, and T) (based on Mann-Whitney-Wilcoxon tests). Whereas location F samples showed the lowest variability in pmoA copy number during the sampling season, location C samples by far exhibited the highest spatial and temporal variability (Fig. 3).

FIG 3.

Box-and-whisker plot showing the total (spatial and temporal) variabilities in MOB abundance (pmoA gene copy number per gram of dry soil) for all samples collected for molecular analyses at soil chronosequence (A, B, and C) and landform (S, T, and F) locations (Table S1). The bottom and top of each box indicate the first and third quartiles, and the horizontal line inside the box shows the second quartile (median). Whiskers indicate maximum and minimum values.

Diversity of MOB communities.

We obtained an average of ∼130,000 pmoA sequences per sample, distributed among sampling locations as follows: 87,000 ± 43,000 (mean ± standard deviation [SD]) at location A sites, 115,000 ± 43,000 at location B sites, 184,000 ± 82,000 at location C sites, 131,000 ± 26,000 at location S sites, 154,000 ± 78,000 at location T sites, and 147,000 ± 66,000 at location F sites. High-throughput sequencing of pmoA genes allowed identification of 31 operational taxonomic units (OTUs). Analysis of the phylogenetic distance of the protein-derived pmoA partial sequences showed that most retrieved OTUs grouped with MOB-like sequences (Fig. 4). Half of the MOB-like OTUs grouped with either type Ic (mostly USC-gamma) or type IIb (mostly USC-alpha) sequences. Twenty percent of the MOB-like OTUs clustered with type Ib, 13% with type IIa, and 7% with type Ia. The remaining 10% of OTUs clustered with the pmoA/amoA-like group, designated for sequences clustering between the pmoA gene and the homologous amoA gene of ammonia-oxidizing bacteria. The applied taxonomic system for pmoA genes follows one reported previously (40). Thirty percent of the MOB-like OTUs were previously undetected pmoA sequences, which may represent novel species. Specifically, OTUs 07, 18, and 28 branched with USC-gamma but showed low nucleotide sequence identity with known pmoA sequences (79 to 86%) (Fig. 4). Similarly, OTUs 09 and 23 branched with type IIb and showed identities of 75% and 84%, respectively, with publicly available pmoA sequences. The OTUs grouping with pmoA/amoA-like sequences and OTU 31 (type IIa) also showed low identities with known pmoA sequences.

FIG 4.

Maximum likelihood tree with 100-bootstrap support showing the phylogenetic affiliation of the pmoA gene based on the derived amino acid sequence (partial sequence [150 amino acids]). Bootstrap numbers are shown for branches with bootstrap support of ≥50. Operational taxonomic units (OTUs) retrieved in this study are depicted in bold. GenBank accession numbers for representative sequences deposited in the database are given in parentheses. Clusters that do not include OTUs from this study are collapsed. The scale bar represents 0.2 change per amino acid position.

The presence and relative abundances of OTUs indicated the highest variability and alpha diversity at locations A to C and S (all part of the sandhill landform) (Fig. 5), whereas the lowest variability and alpha diversity were measured at locations T and F. Among the few exceptions were the ubiquitous and most abundant OTUs 01 (type Ic) and 02 (type IIb). Thirty-five percent of the retrieved OTUs were found only in the sandhill landform, with some OTUs being further specific to location C sites.

FIG 5.

Heat map showing the presence and relative abundances of operational taxonomic units (OTUs) retrieved through targeted pmoA gene amplicon sequencing. Individual samples are displayed on the x axis for soil chronosequence (A, B, and C) and landform (S, T, and F) locations, separated by a dashed line. OTUs shown on the y axis are grouped according to their phylogenetic affiliation (40). The average alpha diversity value ± 1 SD (Simpson index [D]; one-dimensional notation) is shown for all sampling locations. Higher values indicate higher diversity.

Analysis of the beta diversity of MOB communities in glacier forefield soils revealed similar results, i.e., significant differences in community composition related to landform (permutational multivariate analysis of variance [PERMANOVA]; P = 0.04) (Fig. 6). Based on pairwise tests among landform locations, MOB communities in location S samples differed significantly from communities in both location F (P = 4 × 10⁻³) and T (P = 6 × 10⁻³) samples, whereas no significant differences in community composition were detected between locations S and A to C. In fact, the high total variability in MOB community composition in location S samples put these communities closer to those at locations A to C than to those at locations T and F (Fig. 6). Total variability was comprised of both spatial and temporal variabilities; the latter was particularly noticeable at locations A and B (Fig. S4). Conversely, MOB community compositions at locations T and F were almost identical (with T and F data points plotting in the same region in space [Fig. 6]) and showed little variability.

FIG 6.

Principal coordinate analysis of MOB community beta diversity calculated with the weighted UniFrac metric. Average MOB community beta diversity values are shown for chronosequence (A, B, and C) and landform (S, T, and F) locations, with error bars (± 1 SD) representing total, i.e., spatial and temporal, variability. Together, the PCoA 1 and PCoA 2 axes explain 94.3% of the total variance. Average MOB community dissimilarities between and within sampling locations (A, B, C, S, T, and F) are displayed in the inset.

Linear mixed-effects model.

Several trends perceived in the soil physical data were confirmed to be significant by the LME model. For example, topsoil water content at locations A to C increased significantly with increasing soil age, whereas soil water content at locations S, T, and F was significantly influenced by the predictors sampling time point and landform (Table 1). On the other hand, soil temperature significantly decreased over the sampling season at all locations, whereas it significantly increased with increasing soil age at locations A to C.

TABLE 1.

P values of LME models fitted to sample time series of chronosequence (A to C) and landform (S, T, and F) locations^a

Predictor	Correlation, P value for response variableb
	Soil-atmosphere CH4 fluxc		MOB abundance		Soil water content			Soil temp
	Locations A to C	Locations S, T, and F	Locations A to C	Locations S, T, and F	Locations A to C		Locations S, T, and F	Locations A to C		Locations S, T, and F
					b	t	t	b	t	t
Soil age	−, 5.7 × 10−3	NA	+, 7.5 × 10−3	NA	NS	+, 5.3 × 10−3	NA	+, 2.5 × 10−5	+, 4.9 × 10−3	NA
Landformd	NA	2.8 × 10−5	NA	1.3 × 10−4	NA	NA	0.013	NA	NA	NS
Sampling time point	NS	NS	NS	NS	NS	NS	−, 1.8 × 10−3	−, 2.2 × 10−16	−, 2.2 × 10−12	−, 1.86 × 10−5
Cumulative rainfall	−, 0.015	+, 0.032	+, 0.039	NS	NS	NS	NS	NA	NA	NA

^aWe tested the dependence of the response variables soil-atmosphere CH₄ flux and MOB abundance on the predictors soil age, landform, sampling time point, cumulative rainfall, soil temperature, and volumetric soil water content (the data for the predictors soil temperature and soil water content are not shown, as individual P values were not significant). In addition, we also tested the dependence of bulk soil (b) and topsoil (t) temperature and water content on soil age, landform, sampling time point, and cumulative rainfall.

^b+, positive correlation; −, negative correlation; NS, not significant (P ≥ 0.05); NA, not applicable.

^cNote that CH₄ uptake is expressed by convention as a negative soil-atmosphere flux value. Thus, a negative correlation of soil-atmosphere flux with a predictor (e.g., soil age) indicates a positive correlation of CH₄ uptake with these predictors, and vice versa.

^dThe direction of correlation is not applicable.

Soil CH₄ uptake significantly increased with increasing soil age (locations A to C) (Table 1). In addition, CH₄ flux was significantly affected by landform (locations S, T, and F). The results of the LME model further revealed that other than soil age and landform, the only predictor significantly influencing CH₄ flux was cumulative rainfall, whereas soil temperature and soil water content were not significant (P ≥ 0.05) predictors (not shown). The perceived trend of decreasing CH₄ uptake with the ongoing sampling season for location S sites was refuted as not significant by the LME model.

The results of the LME model fits further indicated that the observed increase in MOB abundance with soil age at locations A to C was statistically significant and that MOB abundance was also significantly affected by landform (locations S, T, and F) (Table 1). As in the case of CH₄ flux, soil temperature and soil water content did not significantly influence MOB abundance (not shown).

DISCUSSION

Temporal variability of the soil CH₄ sink as a function of soil age.

Soil-atmosphere CH₄ flux exhibited substantial temporal variability at locations A to C during the snow-free season. This variability was clearly attenuated with increasing soil age (Fig. 2c; see Fig. S3 in the supplemental material). Thus, not only is soil age the main factor affecting the strength of the soil CH₄ sink (39), but increasing soil age also contributes to the stability of atmospheric-CH₄ uptake in these developing soils. In this context, soil age serves as a proxy for all edaphic factors that change with soil development. The importance of soil age in modulating the soil CH₄ sink was corroborated by our LME model, in which soil age was the main environmental factor (P = 5.7 × 10⁻³) (Table 1) explaining CH₄ flux.

Temporal variability in soil physical parameters (Fig. 2a and b) appeared to have little effect on CH₄ flux in glacier forefield soils. A weak dependence on soil temperature was previously explained by gas diffusion being the main, but only mildly temperature-dependent, factor limiting atmospheric-CH₄ uptake (9, 18). Low soil water content and thus a high CH₄ availability in these fast-draining glacier forefield soils (39) may partially explain the lack of an expected dependence of CH₄ flux on soil water content (e.g., see reference 9). In addition, small but potentially important variations in soil water content may have been masked by the measurement uncertainty (0.04 m³ m⁻³), which was often large compared to measured values, in particular for topsoils (0.038 to 0.12 m³ m⁻³) (Fig. 2a).

Conversely, atmospheric-CH₄ uptake at locations A to C showed a significant, positive dependence on cumulative rainfall (P = 0.015) (Table 1). At first glance, this result appears to be at odds with our findings for soil water content. However, it may indicate that atmospheric-CH₄ uptake in these soils was at least occasionally limited by water availability, i.e., when the latter dropped below the lower limit of favorable conditions (∼20% water saturation) (41, 42). At a low soil water content, water exists primarily in small pores and as films coating mineral particles (43, 44). Through continued evaporation, water films may disappear and soil water be present only in the form of small pendular rings between particles (45, 46). Under these conditions, microorganisms inhabiting particle surfaces will be faced with highly unfavorable conditions, which can lead to cessation of their primary metabolism. Cumulative rainfall can be seen as an integrative parameter accounting for recent rainfall events, which replenish water films (47) and thus contribute to maintaining or reestablishing microbial activity. Therefore, MOB survival and activity may strongly depend on cell distribution among microhabitats and, potentially, small changes in soil water content (e.g., see reference 48).

Total variability in MOB community composition at locations A to C (Fig. 6) included a substantial amount of temporal variability (Fig. S4), which is in agreement with previous findings on the variability of total microbial community composition in other glacier forefields (e.g., see reference 49). In particular, MOB communities sampled at locations B1 and B2 during our first campaign (June 18) markedly differed from communities sampled at later time points at these locations. This shift in MOB community composition may be a consequence of changes in environmental conditions that occurred during and immediately after snow melting (50). Although high-affinity MOB are thought to be slow-growing microorganisms (51), the observed temporal variability in community composition indicated that they are capable of promptly adapting to changes in environmental conditions. This may be facilitated, for instance, by the ubiquitous Methylocystis-like MOB (Fig. 4 and 5), which may switch to a dormant stage (cysts) under unfavorable conditions and rapidly initiate excystation once conditions become favorable, thus bypassing the growing phase (e.g., see reference 52).

Effect of landform on the soil CH₄ sink.

Our measurements indicated substantial differences between landforms in their contributions to the overall soil CH₄ sink of glacier forefields; seasonal mean CH₄ uptake at locations F and T was significantly lower than that at location S (Fig. 2). As our measurements represent net fluxes, i.e., the sum of simultaneous CH₄ transport, production, and consumption (53), differences in any of these processes may explain differences in CH₄ flux between the landforms. Although we were unable to distinguish between individual processes, occasionally observed CH₄ emissions at locations F and T support the presence of a CH₄ source of as yet unknown origin in these calcareous glacier forefield soils (54). We found no evidence of microbial CH₄ production on screening of the top 15 cm of soils at locations S, F, and T for presence of the mcrA gene, a biomarker for methanogenic Archaea (data not shown). However, we cannot fully exclude the possibility of microbial CH₄ production in deeper, water-logged soils at locations F and T. Alternatively, the CH₄ source may be related to slow dissolution of carbonate rock or particle aggregates in water-logged soils, thereby releasing entrapped CH₄ (54).

Low CH₄ uptake at locations F and T may also have resulted from low CH₄ availability due to gas diffusion limitation, particularly at location F (9), or from lower MOB abundances at locations F and T than at location S. The latter may be related to differences in soil texture between landforms. A particle size analysis of the <2-mm fraction of soils from locations S, F, and T showed that the clay and silt contents were significantly higher in the location S samples (E.-M. Rainer, E. Chiri, and M. H. Schroth, unpublished data). Bacterial biomass associated with mineral particles has been shown to be substantially higher in clay-silt minerals than in sand (55, 56), likely because microhabitats in fine mineral aggregates provide more favorable conditions for bacterial functioning (e.g., see reference 57). Fine mineral aggregates may have been washed out of soils at locations F and T as a result of flash floods during snow melting or summer rainstorms.

Diversity of glacier forefield MOB communities.

Our study employed a high-throughput sequencing technique (amplicon sequencing on a MiSeq platform [Illumina]) to investigate pmoA gene-based MOB diversity in young, developing glacier forefield soils. Using this technique, the number of identified OTUs was ∼6-fold larger than that in MOB diversity studies based on pmoA gene sequencing of clone libraries (25, 58).

Functional gene diversity can be low in extreme environments, such as glacier forefields (59). Nonetheless, we retrieved 31 OTUs (of which 30% were previously undetected pmoA sequences), which embrace all phylogenetic affiliations identified in previous studies. The USC-gamma clade was most prominent in all samples in terms of presence and relative abundance, as reported in a previous study of Alpine glacier forefield soils (25). These data confirm that high-affinity, USC-gamma MOB are widespread in Alpine soil environments (12, 39, 60). In addition, high-affinity MOB belonging to USC-alpha were prominently detected in the Griessfirn Alpine glacier forefield, likely as a result of using a high-throughput sequencing technique. Three other OTUs grouped with the Methylocystis-like cluster, including the ubiquitous and second most abundant OTU, OTU 02. This agrees with previous findings in which a well-represented Methylocystis-like OTU was exclusively retrieved from glacier forefield soils on calcareous bedrock (25). It may indicate that Methylocystis-like MOB possess an ecological advantage in calcareous glacier forefield soils, which may be linked to the existence of two pMMO isoenzymes, with different affinities for low (2 to 600 μl liter⁻¹) and high (>600 μl liter⁻¹) CH₄ concentrations, previously detected in cultures of a Methylocystis strain (11). Thus, Methylocystis-like MOB may profit from transient release of CH₄ (at elevated concentrations) entrapped in calcareous soil aggregates (54).

Conclusions.

This study contributes to our knowledge on the dynamics of atmospheric-CH₄ oxidation and MOB communities in developing soils and provides supporting evidence to include mountainous lithosols in global CH₄ inventories. Different landforms and their relative extents in a glacier forefield clearly modulate the overall CH₄ sink strength. We confirmed that the soil CH₄ sink and the MOB community driving the latter are subject to temporal and spatial variability during the snow-free season. Further work is needed to elucidate the underlying cause of this variability. Hence, to fully characterize atmospheric-CH₄ oxidation in this heterogeneous environment, an integrative approach will be required, combining, e.g., areal imagery and landscape classification maps with eddy covariance measurements of CH₄ flux. Finally, field studies of CH₄ oxidation during the snow-covered season, as well as investigations of interannual changes in CH₄ oxidation, are still needed to give more comprehensive knowledge on atmospheric-CH₄ oxidation dynamics in Alpine environments.

MATERIALS AND METHODS

Field site.

We performed sampling and measurements in the forefield of Griessfirn Glacier (Canton Uri, Switzerland), which has been described extensively elsewhere (39, 54). Sampling locations were positioned along two transects (Fig. 1). For the first transect, eight locations were selected along a well-defined soil chronosequence with increasing distance from the glacier terminus, on a band of lateral debris deposits (a landform referred to here as a sandhill). They represent a subset of sampling locations previously installed and categorized into three soil age classes (location A, 0 to 20 years; location B, 20 to 50 years; and location C, 50 to 120 years) (39). In the present study, each soil age class comprised 2 or 3 sampling locations.

The second transect was located in soil age class B and comprised five sampling locations each in floodplain (F), terrace (T), and sandhill (S) landforms (Fig. 1, inset), which were identified based on topographical features. Floodplains may be described as streamlined bed forms parallel to the glacial stream, which feature a shallow groundwater table. In contrast, terraces are elevated, ancient floodplains that present dryer conditions with a deeper groundwater table, but otherwise they exhibit a structure similar to that of floodplains. Finally, sandhills are depicted as an unoriented hummocky landform exhibiting a disorganized deposition pattern and the deepest groundwater table among the three landforms (34).

Sampling and measurement procedures.

At locations A to C, we employed the soil gas profile method to determine soil-atmosphere CH₄ flux (milligrams of CH₄ per square meter per day) (e.g., see reference 19). To this end, depth-resolved soil gas samples were collected using a polyuse multilevel sampling system (61). Details of the installation, sampling, and measurement procedures were described previously (54). Soil gas sampling was performed under dry weather conditions during nine sampling campaigns in 2013 (June to October) (see Table S1 in the supplemental material). Sampling started on June 18, when most of the glacier forefield was still snow covered, followed by more frequent sampling shortly after the snow melted (July 2 to 18). The sampling frequency was reduced toward the end of the snow-free season. Concurrent with soil gas sampling, we measured the depth-resolved volumetric soil water content (cubic meters per cubic meter of soil) (PR2/6 capacitance probe; Delta-T Devices Ltd., Cambridge, United Kingdom) at each location, and we recorded the depth-resolved soil temperature (iButton temperature loggers; Maxim Integrated, San Jose, CA) in 1-h intervals throughout the sampling season for one location per soil age group (Table S1). We report topsoil water content and temperature by using the uppermost measurement points (7.5 to 10 cm in depth) and bulk soil values by using the mean for all measured depths (7.5 to 97.5 cm).

For locations S, T, and F, we used the static flux chamber method to quantify CH₄ flux (e.g., see reference 62). Deployment of polyvinyl chloride chambers (31-cm diameter × 27-cm height) took place shortly after the snow melted (July 9 and 10). Chambers comprised a base collar inserted ∼15 cm into the ground. To allow soil consolidation around the collars, an idle phase of 8 days (which included several rainfall events) followed chamber installation. Flux chamber measurements were performed during six sampling campaigns (July 18 to September 19) (Table S1). For measurements, collars were fitted with gaskets and detachable lids, resulting in chamber headspace volumes ranging from 9.2 to 11.4 liters. Each lid featured an aluminum foil cap, to minimize temperature increases in the headspace from solar irradiation, and a port connected to a three-way valve, which allowed extraction of 20-ml gas samples from the headspace by use of a gastight syringe. Measurement periods lasted 90 min, with four gas samples collected at regular time intervals. Gas samples were immediately transferred to serum bottles capped with butyl rubber stoppers for subsequent analysis of the CH₄ concentration. Concurrent with flux chamber measurements, topsoil water content was quantified by time domain reflectometry (TDR100; Campbell Scientific, Loughborough, United Kingdom) at 2 or 3 spots per landform in the close vicinity of the flux chambers, using pairs of 30-cm-long brass rods permanently installed in the ground. Nearby, the topsoil temperature was measured at a depth of 11 cm at 1-h intervals, using iButton temperature loggers mounted on wooden rods.

A total of 56 soil samples were collected for molecular analyses on 6 days over the course of the sampling season (Table S1). Soil was collected from all eight sampling sites at locations A to C and from a total of nine sampling sites at locations S, T, and F. Details of the soil-sampling procedure were described previously (39). Individual samples and measurements are referred to by sampling location (A, B, C, S, F, or T) and number (1 to 5) and by sampling time point (date), e.g., A1-Jul02 or F5-Jul25.

Data for daily rainfall (millimeters per day) were obtained from the nearest automated weather station of the Federal Office of Meteorology and Climatology (MeteoSwiss). Cumulative rainfall (millimeters per day) for each sampling date was calculated as the weighted sum of daily rainfall within 4 days prior to sampling, using an exponential decay function with a decay constant of 1 day⁻¹ for weighting (i.e., the effect of preceding rainfall was halved every 16.6 h).

Determination of soil-atmosphere CH₄ flux.

Methane concentrations in all soil gas samples were quantified on a gas chromatograph equipped with a flame ionization detector (63). To determine soil-atmosphere CH₄ flux at locations A to C, analytical solutions to a steady-state diffusion-reaction model were fitted to individual soil CH₄ profiles in the R v3.2.1 software environment (64). Analytical solutions assumed CH₄ oxidation to be governed either by a single first-order reaction over the entire profile (65) (referred to here as the one-layer model) or by individual first-order reactions in two distinct (top and bottom) soil layers (39) (referred to here as the two-layer model). Applying Fick's first law of diffusion, we then computed the flux from the CH₄ concentration gradient at the soil-atmosphere boundary and best-fit parameters of that model, which yielded better agreement with measured soil CH₄ profiles. In cases where model convergence failed, CH₄ concentration gradients and fluxes were approximated by linear regression (39). For locations S, T, and F, flux values were obtained from the slope of CH₄ concentrations in the flux chambers plotted against measurement time by linear regression analysis (e.g., see reference 66).

pmoA gene marker-based molecular analyses of the MOB community. (i) Total DNA isolation and pmoA gene amplification.

Genomic DNA was isolated using a FastDNASpin kit for soil (MP Biomedicals, Solon, OH) according to the manufacturer's instructions, with minor modifications. Recovery and purity of the DNA extracts were tested prior to further molecular analyses. Applied protocols for DNA extraction and quality control were described previously (39). Amplification of the pmoA gene was the first step for all subsequent molecular techniques applied in this study. Primer sequences and amplification procedures are reported in Table S2.

(ii) qPCR.

To determine MOB abundance, copies of pmoA genes in DNA extracts were quantified by quantitative PCR (qPCR) on an ABI 7500 system (Applied Biosystems [now Thermo Fischer Scientific], Waltham, MA). Quantification of pmoA copy numbers followed a previously described procedure (67) and the assay specifications shown in Table S2. All samples were analyzed in triplicate. The efficiencies of the three assays performed were always ∼100%, with R² values above 0.99.

(iii) Targeted amplicon sequencing.

For each sample, we prepared, indexed, and paired-end sequenced amplicon libraries of the pmoA gene from genomic DNA extracts according to the targeted amplicon sequencing method (68). Primer sequences and amplification procedures are reported in Table S2. Amplicon library preparation, high-throughput amplicon sequencing, and sequence data processing are described in detail in the supplemental material.

Statistical analyses. (i) Linear mixed-effects models.

Using lme4 v1.1-10 (69) in R v3.2.1, LME models were fitted to sample time series, with random effects included for individual sampling locations. The drop1 function was used to find the optimal LME model for each response variable, based on the Akaike information criterion. We tested the dependence of the response variables CH₄ flux and MOB abundance on the model predictors soil age, landform, sampling time point, cumulative rainfall, soil temperature, and water content. Initial checks indicated a correlation between soil temperature and sampling time point, and thus an additional interaction term between these variables was included. Where residual diagnostics indicated nonlinearity and/or nonnormality, variables were log transformed (soil age, MOB abundance, cumulative rainfall, and sampling time point) or arcsine transformed (all water content data). Missing values were replaced with the mean for the respective soil age group or landform.

(ii) Diversity and structure of MOB communities.

Phylogenetic distances of the assigned operational taxonomic units (OTUs) were assessed through nucleotide sequence alignment and phylogenetic tree building on the derived-protein level, using the software Seaview v4.5.4 (70). To identify the best evolution model, we tested 96 amino acid substitution models by using the software ModelGenerator v0.851 (71) and selected the one showing the lowest value for the Akaike information criterion. A phylogenetic tree was built according to the maximum likelihood method, with 100-bootstrap support, using the phylogeny software PhyLM v3.1 (72).

To assess the alpha diversity of MOB communities, we computed Simpson indices (D) and visualized the results in one-dimensional notation, in which alpha diversity ranges from 0 (low diversity) to 1 (high diversity). We further employed Mann-Whitney-Wilcoxon tests to assess differences in alpha diversity between sample locations (A, B, C, S, T, and F). To identify factors explaining differences in community structure, we analyzed the beta diversity of MOB communities (defined as the variation in composition between a set of communities) coupled with standard multivariate statistics (73, 74). Alpha and beta diversity calculations, as well as read count normalization of the pmoA sequences, were performed with the package phyloseq v1.12.2 (75) from the open source software Bioconductor. We applied a read count threshold of ≥35,000 counts per sample. To account for differences in numbers of reads between samples, we used rarefied OTU counts to an even sampling depth. The resulting data set was used for all subsequent analyses. Phylogenetic beta diversity was calculated using the tree-based unique fraction (UniFrac) metric weighted by the abundances of individual OTUs (76). Principal coordinate analysis (PCoA) ordination of the distance metric was used to identify community grouping (77). To determine whether the observed between-group distances were statistically significant, we performed permutational multivariate analysis of variance of the distance metric by using the PERMANOVA+ package of the software PRIMER-E v7 (PRIMER-E Ltd., Plymouth, United Kingdom).

Nucleotide sequence accession number(s).

Sequence reads of pmoA genes obtained from targeted amplicon sequencing and representative pmoA sequences of identified OTUs were deposited at the European Nucleotide Archive (ENA) under study number PRJEB20489.