Study area

Our study area encompassed 180 km² of lowland tundra near Utqiaġvik, Alaska (71.2906° N, 156.7886° W; Fig 2). Mean temperature is ‒12°C during long winters and 4°C during short summers, with annual precipitation of 11 cm/y. The area has low relief (0‒10 m above sea level), and a thin active (seasonally frozen) layer (30‒40 cm) above continuous permafrost [27]. The landscape is dominated by a mosaic of thaw-lake basins. These basins are formed by creation and merging of flooded ice-wedge polygons and troughs to form shallow lakes, which later drain and revegetate before the cycle (lasting 2000‒5500 y) begins again [28, 29]. These landforms and interstitial tundra host an array of shallow wetlands vegetated mainly by water sedge (Carex aquatilus) or pendant grass (Arctophila fulva) [6]. Carex is emergent, but Arctophila can be emergent or submersed depending on water depth. Carex meadows also dominate uplands around the wetlands [27, 29]. In this area, tundra wetlands occupy a proportion of the landscape that is probably comparable to that in lowland tundra throughout the extensive Alaskan Arctic Coastal Plain (Miller et al. [Unpublished]).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Study area in lowland tundra near Utqiaġvik, Alaska, USA.

Map developed using 2002 Quickbird imagery from the National Snow and Ice Data Center [30].

https://doi.org/10.1371/journal.pone.0286368.g002

In the spring, frozen ground prevents penetration of snowmelt, which flows mostly into larger stream channels which flood seasonally. In summer and fall, the low topographic relief limits lateral run-off of rain as well as lateral transport of water through the active layer of soil, resulting in ponding at the surface. Thus, water budgets of the wetlands are driven mainly by direct summer precipitation and evapotranspiration [31]. Wetland sediments are generally flocculent and highly organic (~82% organic C). Microbial primary producers (mostly green algae and cyanobacteria in roughly similar proportions) are mostly attached to sediment and detritus particles, with relatively little production in the shallow water column [32]. Benthic algae generally do not form algal mats owing to bioturbation, and the main control on benthic algal biomass appears to be light limitation by mixing of algae into the sediments [32].

Sample collection and processing

We sampled six wetland types (based mostly on the classification of Bergman et al. [1]) characterized by depth and dominant macrophyte species: Shallow Arctophila (n = 14 wetlands), Deep Arctophila (n = 10), Shallow Carex (n = 11), Deep Carex (n = 7), Streams (n = 6), and Deep Open Lakes (n = 4). Wetlands dominated by Arctophila or Carex were termed “deep” if depth exceeded 40 cm, although depth was typically <110 cm. Deep Open Lakes were defined as having maximum depth >150 cm and surface area >1 ha, typically with no macrophytes present. Deep Open Lakes were generally >2 m deep, with some probably up to 8 m deep; they typically did not freeze in their entirety to the bottom in winter and some contained fish (mainly ninespine stickleback, Pungitius pungitius [33]). However, all of our sampling of Deep Open Lakes was in littoral areas <1 m deep that were accessible by wading. Streams were shallow, slow-moving, beaded streams between basins or between basins and coastal sloughs, usually with substantial overbank flooding during the spring. Equivalence of these different wetland types with the classification of Cowardin et al. [34] was documented by Derksen et al. [35].

We collected macrophyte and invertebrate samples in June and July of 2013, 2017, and 2018. In all three years, we used a pole-mounted, triangular sweep-net with 500-μm mesh (Wildco, Yulee, Florida, USA). In 2013, samples were collected for stable isotope analyses, but were not suitable for estimating invertebrate densities. One net sample was collected by scraping through the upper sediments, and the second sample by sweeping the net through vegetation (Carex or Arctophila); the two samples were combined into a single aggregate sample per wetland. In 2017 and 2018, we obtained quantitative net samples within stands of macrophytes at two different locations per wetland. At each location, we swept the net horizontally for 2.5 m through the water column 10 cm below the water surface to capture invertebrates found within the vegetation. Also in 2017 and 2018, at four different locations per wetland, we pressed an acrylic corer (inside diameter 5.2 cm, length 50 cm) through the sediments until it encountered the subsurface ice layer. If water depth exceeded 50 cm, a valve attached to the corer was closed to create a vacuum which assisted extraction of the core. The top 10 cm of the resulting core was filtered through a 500-μm sieve to retain invertebrates. If the corer encountered a frozen layer <10 cm deep in the sediment, biomasses per unit area were corrected based on volume sampled. All plant and invertebrate samples were frozen for later processing.

In the laboratory, invertebrate samples were thawed, sorted, and counted. Insects identified from our samples included larval forms of Diptera (Chironomidae and Tipulidae), Plecoptera, Trichoptera, and Coleoptera (mainly Dytiscidae). Other invertebrate taxa were Acari, Crustacea (mostly Copepoda with some Notostraca and Daphnidae), Oligochaeta, and Gastropoda of the family Physidae. Identification to lower taxonomic levels for some taxa was hindered by sample degradation during freezing. Macrophyte tissue was identified as Carex or Arctophila.

We did not separate larvae of Chironomidae into predatory (subfamily Tanypodinae) vs. non-predatory taxa. Butler et al. [36] reported that Tanypodinae averaged 3.8% (range 1.1 to 6.4%) of numbers of all chironomid larvae in three wetlands in our study area. In 12 wetlands in our study area, Lougheed et al. [11] found Tanypodinae to comprise 6.5% of total chironomid numbers in 1971‒1973 and 2.2% in 2007‒2009. Moreover, various Tanypodinae are facultative predators, often consuming partly or mainly detritus [37–39]. Given the small fraction of Tanypodinae and their often mixed feeding modes, we analyzed all Chironomidae as a single taxon.

Invertebrate masses and stable isotope analyses

For analyses of μg C per individual (S3 Table), samples for each taxon were pooled across wetland types to obtain large enough sample masses. For biomasses of each taxon within each wetland type and sample type (cores vs. sweeps), means were calculated across wetlands of each type.

For isotopic analyses, obtaining large enough masses required pooling of samples within taxon and wetland type. Specimens from each invertebrate taxon were counted, aggregated within wetland, oven-dried for 24 h at 55°C, ground with a mortar and pestle, and fumed with 10% HCl for 20 min to remove inorganic carbonates that might confound δ¹³C values of OM. These samples were then placed into folded, precombusted glass-fiber filters (Whatman GF/F, Cytivia, Marlborough, MA, USA). All individuals of a given taxon in samples from a given wetland were aggregated in a single filter. Lipids have lower δ¹³C values than proteins or carbohydrates, so that the whole-body δ¹³C values of the same organism can vary appreciably depending on its current state of lipid storage. To eliminate this variation, Post et al. [40] recommended lipid extraction when lipid content varies among consumers or between consumers and prey endmembers, and when δ¹³C values between endmembers differs by <10‒12‰. In our study, lipid content ranged from 3 to 23% among consumer taxa, and differences in δ¹³C values among endmembers were in most cases <10‰ (Table 1). Accordingly, we extracted lipids from all invertebrate samples. The filters with samples were immersed in 99% petroleum ether for 72 h to remove lipids, with the ether replaced every 8 h [41].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Means and SE of stable isotope values (δ¹³C, δ¹⁵N) of endmembers for organic matter sources used in mixing models, based on measured and reported values (S1 Table).

https://doi.org/10.1371/journal.pone.0286368.t001

After lipid extraction, invertebrate samples were oven-dried again for 48 h at 55°C and then ground with a mortar and pestle. Vegetation samples were dried and homogenized with a household coffee grinder. Samples from each filter (with varying total masses) were stored in sealed glass vials until being subsampled for analyses of δ¹³C and δ¹⁵N at the University of Wyoming Stable Isotope Facility, via a Costech 4010 elemental analyzer (Costech Scientific, Valencia, CA, USA) coupled with a Thermo Delta Plus XP isotope-ratio mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA). Stable isotope values are expressed in units of parts per thousand (‰), as the ratio of heavy to light isotope relative to a standard. This quantity is calculated by the equation δX = [(R_sample/R_standard)– 1] × 1000, where X is ¹³C or ¹⁵N and R is the ratio of ¹³C/¹²C or ¹⁵N/¹⁴N. Values are reported with respect to Vienna PeeDee Belemnite for δ¹³C and atmospheric N₂ for δ¹⁵N. Measurements of internal laboratory standards (peptone) yielded a precision of 0.1% for both δ¹³C and δ¹⁵N values.

Endmembers

For endmembers (organic matter sources) in mixing models, we used stable isotope values from our own samples for macrophyte tissue, and means among values reported in the literature for periphytic (benthic and epiphytic) microalgae, peat, and cyanobacteria (Table 1 and S1 Table). Obtaining a pure and consistent isotopic signal for periphytic algae from field samples can be notoriously difficult. Especially in flocculent, highly organic sediments, the periphytic layer is typically a complex community that includes fungi, bacteria, protists, organic tissue fragments, exopolymer secreted by both algae and bacteria, and DOM sorbed to and diffusing into the periphytic matrix from the water column [48, 49]. Indeed, algal cells are typically a minor component of periphytic carbon, averaging only 8% over a range of substrate types [50]. Thus, measurements of the bulk isotopic composition of periphytic samples often cannot be used reliably to represent that of primary producers alone, being too enriched in ¹³C. In our study area, broadscale sampling of periphytic communities in wetlands with flocculent, highly organic sediments, and active bioturbation that generally prevents formation of algal mats [32], was unlikely to yield accurate isotopic values for periphytic algae alone. Such measurements would not encompass the lower δ¹³C signatures for algivores and thus would not yield a suitable endmember. Owing to these technical challenges, similar studies have often used literature values of algae scraped from hard surfaces as the best proxy for periphyton in soft sediments [e.g. 51]. The available value for such periphyton samples in our region was ‒37‰ δ¹³C for streams in areas of ice-rich permafrost in northwest Alaska [42] (Table 1 and S1 Table).

Moreover, the guts of deposit-feeders typically contain both recognizable cells (e.g. diatoms, cyanobacteria) and unidentified particles termed “amorphous detritus” which are thought to include flocculated exopolymer secreted by algae and bacteria. Such mixtures in gut contents indicate ingestion of multiple components of the periphytic community [15, 49]. Literature values indicated that algae had by far the lowest δ¹³C value of the potential organic matter sources for wetlands in our study (Table 1 and S1 Table). Therefore, as has been done in similar cases with variable fractionation by microbes and uncertain selectivity by deposit-feeders, we considered the most reliable pure endmember for benthic microalgae to be the most extreme (most depleted in ¹³C) value among the primary consumers [cf. 52, 53]. In mixing models we used a δ¹³C value of ‒38‰, which encompassed almost all of the most depleted values for invertebrates (Fig 3) and was consistent with the literature value (‒37‰).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Stable isotope scatterplots for each wetland type.

Endmembers for organic matter sources with trophic discrimination factors applied (Table 2) are also plotted (yellow circles). (A) Shallow Arctophila, (B) Deep Arctophila, (C) Shallow Carex, (D) Deep Carex, (E) Streams, and (F) Deep Open Lakes.

https://doi.org/10.1371/journal.pone.0286368.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Trophic discrimination factors (TDFs, Δ¹³C, Δ¹⁵N) for macroinvertebrate consumers relative to organic matter sources (endmembers), and trophic route involved in the fractionation, in tundra wetlands.

Literature survey for the TDFs is in S2 Table. Values in parentheses are for endmembers (Table 1) plus the corresponding TDFs; resulting values are plotted in Fig 3.

https://doi.org/10.1371/journal.pone.0286368.t002

For peat in tundra environments, the isotopic signature can vary with depth through the active layer and the upper boundary of permafrost [42, 54]. However, within the active layer, permeability to lateral porewater movement decreases sharply with increasing depth [55, 56], and in our study area there is mostly laminar horizontal flow with little vertical mixing among depth layers [31]. Given that the active layer above impermeable permafrost deepens slowly as it thaws over several months [31], pore water moving laterally will collect DOM mostly from a relatively shallow layer of peat. In fact, most DOC entering shallow tundra wetlands is derived from this near-surface layer of more recently deposited organic matter [57].

This upper layer of peat, termed the acrotelm, differs from deeper layers in that it has much higher hydraulic conductivity, is periodically aerated, is rich in peat-forming bacteria and other microorganisms, and includes both living and dead plant material [58]. In peat formed mainly from Carex as around wetlands in our study area, the peat consists mostly of living and dead Carex roots that penetrate downward through the acrotelm [59, 60]. Isotopic measurements at five sites across an Icelandic landscape of peat derived mainly from Carex [45], as well as from six sites across our study region, yielded standard errors of only 0.21‰ in δ¹³C and 0.38‰ in δ¹⁵N (Table 1 and S1 Table). Given these very low variances among different areas for the same habitat, we considered the means and SE from literature values for peat to be suitable for our mixing models.

Cyanobacteria presented similar challenges as periphytic microalgae. Consequently, we used published values from a nearby site where samples could be collected from cyanobacterial mats, and from lakes in the Alaskan Arctic and a temperate region where deeper water allowed filtering adequate samples from the water column.

Note that for all endmembers derived from literature surveys, associated variances (S1 Table) are accounted for in mixing model software (MixSIAR) used to compare individual samples to different OM matter sources. When considering the range of possible values from the literature, there was clear separation of the endmember values (Fig 3) and good convergence of Monte Carlo simulations in mixing models (see section on Mixing models).

Because methane-oxidizing bacteria are strongly depleted in ¹³C, they can have important effects on the δ¹³C values of consumers [61]. However, because local measurements suggest that the role of methanotrophs in these systems is minimal (see details in the Discussion), we did not include an endmember for methane-oxidizing bacteria.

Trophic discrimination factors

Mixing models require trophic discrimination factors (TDFs), or the change (fractionation) in isotopic values that occurs between ingestion and incorporation into consumer tissues (written as Δ¹³C and Δ¹⁵N). General averages of TDFs are commonly used (e.g. ≤1 ‰ for Δ¹³C and 3.4 ‰ for Δ¹⁵N, [52]). However, these values often do not apply to consumption by heterotrophic microbes or to metazoan consumption of those microbes or algae. In our study, TDFs between basal organic matter sources and deposit-feeding macroinvertebrates were based on a survey of published values (Table 2 and S2 Table).

When peat layers in surrounding tundra thaw, the organic matter in dead plant tissue is metabolized and fractionated by bacteria to be released as DOM [62]. This DOM is then carried by pore water diffusing through subsurface material into the wetlands. Once in the wetlands, peat-derived DOM can be further fractionated by photolysis [63], and can sorb to the periphytic layer and diffuse into that matrix for use by bacteria [48, 64]. The activity of bacteria bound to surfaces is generally far greater than that of unbound bacteria in the water column [48]. By the time these bacteria or their exudates are subsequently ingested by deposit-feeding macroinvertebrates, organic molecules of the original peat have been fractionated multiple times. We acknowledged this repeated fractionation in estimating the TDFs between peat tissue and deposit-feeding macroinvertebrates (Table 2 and S2 Table).

The bacterial fraction of the periphytic community can ingest only DOM, which may be derived from lysed cells or exudates of algae, from leachates of macrophytes [65], or from peat DOM leaching into the wetland from surrounding tundra. Although predation on bacteria by protists may insert another trophic level, fractionation by protists consuming bacteria appears to be minimal [66, 67].

Peat OM enters the wetlands almost entirely as DOM, and bacteria generally can consume OM only as DOM [48]. Thus, we assumed that almost all peat DOM is ingested by detritivores in the form of bacteria or their flocculated exudates. Fresh algal cells can be consumed directly by detritivores, whereas we assumed that algal exudates and dead algal cells are assimilated by detritivores mainly in the form of bacteria [48]. These assumptions are consistent with the prevailing view that primary production in freshwater ecosystems, especially those dominated by detritivores, enters metazoan food webs mainly as detritus via bacterial and fungal intermediates [68, 69]. Fractionation of N isotopes between fresh algal cells and invertebrate consumers tends to be quite low, whereas we expect fractionation between peat DOM and dead algae to detritivores to be substantially higher (including bacterial or fungal intermediates) (Table 2).

We assumed that macrophyte tissue (litter) is assimilated by detritivores mainly as bacteria and especially as fungi growing on the tissue (Table 2) (cf. [16]). Particulate matter composed of plant tissues may be largely refractory to invertebrate digestion [70, 71], but bacteria colonizing the surface of particulate matter can provide readily accessible nutrients for invertebrate consumers [72]. Fungal biomass can far exceed bacterial biomass in attached communities, representing 70‒99% of microbial biomass on decomposing leaves (including emergent macrophytes) in fresh waters [73, 74], and 70% of total microbial carbon in sediments of a wetland in our study area [75]. Unfortunately, the role of fungi in detritivore diets is poorly studied in aquatic systems.

Mixing models

Although TDFs for δ¹⁵N were assigned for initial display purposes (Fig 3), very wide variation in δ¹⁵N indicated that variable numbers of trophic levels within the microbial loop confounded selection of consistent TDFs for δ¹⁵N [cf. 53]. Consequently, δ¹⁵N values were of limited value for discriminating organic matter sources, so we used single-isotope models for δ¹³C only. We used the Bayesian mixing model software MixSIAR [76] to estimate diets of each invertebrate taxon within each wetland type. Proportions of organic matter sources for different invertebrates were unknown, so we chose an uninformative prior for analyses. We ran models with the following Markov Chain Monte Carlo parameters: 3 chains; 300,000 samples; a 200,000 sample burn-in; and retention of each 100^th sample to minimize autocorrelation (a “long” run in MixSIAR).

Biomass estimates

We determined the carbon content of invertebrates from subsamples prepared for stable isotope analyses. We divided the C biomass by the number of individuals per subsample to yield an estimate of C biomass per individual for each taxon (S3 Table). The latter estimate was then multiplied by counts in field samples to yield g C m^‒2 for each taxon for cores and net sweeps separately in each wetland. We then calculated mean biomass in cores and sweeps for each invertebrate taxon among wetlands within each wetland type. Tipulidae and Oligochaeta were not included in biomass analyses. Tipulidae were generally rare in all wetland types and were not detected in Deep Open Lakes or Streams. Although samples of Oligochaeta were adequate for stable isotope analyses, degradation of Oligochaeta by freezing prevented accurate biomass determinations.

We used PRIMER version 7 with the PERMANOVA+ add-on to test for differences in biomass among wetland types for core and sweep samples separately [77]. We transformed the data by log (x + 2) before constructing Bray-Curtis dissimilarity matrices. For comparing invertebrate communities often dominated by a few taxa, this transformation is recommended to down-weight dominant taxa and enhance consideration of taxa with lower total biomass [78]. Using wetland type and individual wetland as factors, we ran a Permutational Analysis of Variance (PERMANOVA) with pairwise comparisons of wetland type for each invertebrate taxon. For comparisons with less than 100 unique permutations, we conducted Monte Carlo tests to obtain pseudo-F statistics and permutation-p values.

Organic matter sources of invertebrate taxa by wetland type

A single-isotope model cannot discriminate more than two endmembers. In such cases, one must use only the two most likely endmembers suggested by the isotope biplots [79], which were periphytic microalgae and macrophytes-peat (Fig 3). Cyanobacteria were important only to Physidae snails, and to a lesser extent Oligochaeta and Tipulidae, so these taxa were excluded from single-isotope models.

The continuous distribution of δ¹³C values within deposit-feeders (Fig 3) suggested variation in fractionation of algal inputs (multiple trophic levels) within the microbial loop, variation in dietary proportions of algae vs. macrophytes-peat, or both. Our mixing models could not distinguish these possibilities, attributing all intermediate values to mixtures of the endmembers rather than to variable fractionation of the algal endmember.

Results of mixing models (Fig 4) indicated that in Shallow and Deep Arctophila and Shallow Carex wetlands, microalgae were the dominant ultimate source of organic matter for all invertebrate taxa examined (50‒82%, mean 64%). (Statistics on mixing model convergence are in S5 Table, and means and SE of model estimates of the percentages of organic matter sources are in S6 Table.) In the same wetlands, macrophytes-peat accounted for only 18‒50% (mean 36%) of OM inputs for these invertebrates. In Deep Carex wetlands, microalgae represented 39‒64% (mean 46%) of OM inputs in invertebrates, with slightly larger contributions from macrophytes-peat (36‒61%, mean 54%) (Fig 4D). In Streams, microalgae accounted for 46‒70% (mean 58%) of OM inputs, compared to contributions of 30‒54% (mean 43%) from macrophytes-peat (Fig 4E). In contrast to the shallower wetland types, invertebrates in Deep Open Lakes reflected mainly macrophytes-peat as the dominant carbon source (38‒80%, mean 69%), with lesser contributions of microalgae (20‒62%, mean 31%) (Fig 4F). Inspection of Fig 3 indicates that cyanobacteria, which were not included in our two-source mixing models, were probably a minor dietary component for all taxa except Physidae snails.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Relative contributions (mean percentages) of periphytic microalgae and of peat and macrophytes combined to diets of invertebrates across six wetland types.

For data values see S6 Table. Acar = Acari, Crus = Crustacea, Chir = Chironomidae, Plec = Plecoptera, Tric = Trichoptera, Tipu = Tipulidae, Cole = Coleoptera, Olig = Oligochaeta, Phys = Physidae). (A) Shallow Arctophila (n = 34), (B) Deep Arctophila (n = 26), (C) Shallow Carex (n = 36), (D) Deep Carex (n = 40), (E) Streams (n = 11), and (F) Deep Open Lakes. Asterisks (*) indicate taxa that were not detected in a given wetland type.

https://doi.org/10.1371/journal.pone.0286368.g004

Invertebrate biomass

Although patterns of organic matter sources were mostly similar among invertebrate taxa in different wetland types, total and relative biomasses among taxa in cores were highly variable among wetland types (Fig 5; for numerical values and results of statistical tests, see S7 Table). Chironomidae had consistently high or the highest biomass of all taxa. Coleoptera (mainly Dytiscidae) had exceptionally high biomass in at least some Streams (although with very high variance among wetlands of this type), but relatively low or negligible biomass in other wetland types. Crustacea had very low biomass in some wetland types, and moderate biomass in others. Acari always had negligible biomass, while Plecoptera, Trichoptera, and Physidae generally exhibited low and variable biomasses among wetland types.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Mean ± SE of invertebrate biomass (mg C m^‒2) in benthic cores across six wetland types.

(Acar = Acari, Crus = Crustacea, Chir = Chironomidae, Plec = Plecoptera, Tric = Trichoptera, Cole = Coleoptera, Phys = Physidae,). (A) Shallow Arctophila (n = 14), (B) Deep Arctophila (n = 10), (C) Shallow Carex (n = 11), (D) Deep Carex (n = 7), (E) Streams (n = 6), and (F) Deep Open Lakes (n = 4). Total biomass of all invertebrates is annotated in each panel. Numerical values and results of statistical tests are in S7 Table.

https://doi.org/10.1371/journal.pone.0286368.g005

Net sweeps in the water column within emergent stands yielded more consistent and even representation of invertebrate taxa (Fig 6; for numerical values and results of statistical tests, see S8 Table). Chironomidae were again the dominant taxon in all wetland types, having especially high biomass in Shallow Carex. Trichoptera and Coleoptera were relatively more important to total biomass in sweeps than in cores for most wetland types. Shallow Arctophila and Shallow Carex had 2‒3 times the total invertebrate biomass of Deep Arctophila, and Shallow Carex had 54% higher biomass than Deep Carex. The Carex wetlands had 41 to 89% higher total biomass than Arctophila wetlands of comparable depth. Although Streams had relatively high biomass of Chironomidae, total biomass of invertebrates was relatively low. Except for Crustacea (mostly Copepoda with some Notostraca and Daphnidae), the biomass of invertebrates was generally far lower in the littoral zones of Deep Open Lakes than in the other wetland types.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Mean ± SE of invertebrate biomass (mg C m^‒3) in net sweeps through emergent vegetation across six wetland types.

Acar = Acari, Crus = Crustacea, Chir = Chironomidae, Plec = Plecoptera, Tric = Trichoptera, Cole = Coleoptera, Phys = Physidae). (A) Shallow Arctophila (n = 12), (B) Deep Arctophila (n = 10), (C) Shallow Carex (n = 11), (D) Deep Carex (n = 7), (E) Streams (n = 6), and (F) Deep Open Lakes (n = 4). Total biomass of all invertebrates is annotated in each panel. Numerical values and results of statistical tests are in S8 Table.

https://doi.org/10.1371/journal.pone.0286368.g006

Methane oxidation vs. fractionation during DIC uptake

For invertebrates with δ¹³C values intermediate between microalgae and macrophytes-peat (Fig 3), it is also possible that their carbon sources included a mixture of macrophytes-peat with relatively high δ¹³C values, and of methane-oxidizing bacteria with very low δ¹³C values. Indeed, our values of δ¹³C in deposit-feeding invertebrates were often in the “zone of contention” (‒40 to ‒30‰), meaning that these values might or might not represent dietary mixtures of moderately depleted foods with methanotrophic bacteria strongly depleted in ¹³C [80]. However, several lines of evidence suggest that the latter possibility is less likely in our tundra wetlands.

In wetlands in our study area, Throckmorton et al. [54] reported that most subsurface methane was routed upward through the stems and leaves of emergent plants, mainly bypassing the potential for methane oxidation. As a result, only 2.5% (July) and 0.8% (September) of methane produced was consumed by methanotrophic bacteria, accounting for <0.1% of dissolved inorganic carbon (DIC) production. Instead, DIC was produced mostly via respiration by non-methanotrophic bacteria. The δ¹³C value of DIC in porewater at the surface of the active layer was ‒12‰ [54], further indicating that little methane oxidation was occurring near the sediment surface. Methane oxidation has been important mainly in deeper, stratified lakes with anoxic sediments and bottom water [61]; however, anoxia favoring some methane oxidation can occur in organic, subsurface sediments of shallow lakes (< 2 m deep), perhaps especially those with dense macrophytes [81, 82]. In contrast, in our area and others with polygonal tundra, macrophytes appear to channel most methane past the zone of methane oxidation [23, 54, 83]. Moreover, a survey of 87 temperate lakes indicated that appreciable consumption of methanotrophic bacteria by chironomid larvae occurred only when oxygen concentrations near the sediment surface fell below about 2 mg O₂/L [61, 84]. Our tundra wetlands were generally <1 m deep with perpetually windy conditions, and past measurements in three wetlands yielded a minimum of 3.5 mg O₂/L within an emergent stand of Carex aquatilis [85]. Together with the findings of Throckmorton et al. [54], these aspects suggest that our vegetated tundra wetlands supported little biomass or production of methanotrophic bacteria, and yielded little respired CO₂ from methanotrophs available for uptake by benthic algae [61].

Alternatively, the low δ¹³C values in deposit-feeders could result merely from selective algal uptake of DIC with low δ¹³C value, given the very high partial pressure of CO₂ (pCO₂) in these wetlands. For mean summer temperature during mid-day at Barrow (⁓7.2°C), the CO₂ solubility coefficient (K_o) is 0.059 mol/kg atm. For a pCO₂ of 1.3336 atm measured in local wetlands (Table 2 in [22]), CO_2(aq) = pCO₂ / K_o = 22.6 μmol/L. Based on eqn (3) of Lennon et al. [86], concentration-dependent fractionation between CO_2(aq) and HCO₃^‒ during photosynthesis by microalgae is estimated as ‒ [25.4 × CO_2(aq)] / [3.7 + CO_2(aq)] = ‒21.8‰. Relative to δ¹³C values of ambient DIC at the sediment surface of ‒12‰ [54], benthic microalgae should have δ¹³C values of about ‒33.8‰, and primary consumers about ‒32.8‰ (assuming Δ¹³C ≈ 1‰). Given errors and assumptions in these estimates, isotopic fractionation during microalgal uptake of DIC at such high CO₂ concentrations appears adequate to explain most if not all of the values observed in deposit-feeders, without invoking consumption of methanotrophic bacteria. The main taxa reported to consume methanotrophic bacteria are chironomid larvae, in particular those living in irrigated tubes containing oxygenated water adjacent to anoxic surrounding sediments [61]. However, in our wetlands, diverse deposit-feeding invertebrates (Crustacea, Plecoptera, Trichoptera) overlapped in δ¹³C values with chironomids (Fig 3), suggesting that the most often-cited mechanism supporting consumption of methanotrophs was not necessary to produce the observed values.

In summary, we believe that direct algivory, and consumption of bacteria that have ingested algae, their exudates, or DOC leached from macrophytes-peat, are the most important carbon sources at the base of invertebrate food webs in these tundra wetlands. Consumption of methane-oxidizing bacteria can be appreciable in other areas, but is highly variable within and among systems [61] and in our case appears to be much less important. We note that Physidae snails appeared to be unique in consuming substantial amounts of OM from cyanobacteria (Fig 3).

Effects on endmember values

Across invertebrate taxa, δ¹³C values were quite low compared to literature values for available organic matter sources except microalgae (cf. Fig 3, Table 1 and S1 Table). Lipids exhibit more negative δ¹³C values compared to other tissues [40], but extraction of lipids from our samples prevented this effect. A microalgal endmember of –38‰ was needed to encompass most values in our samples (Fig 3). Values for biofilm as low as ‒40‰ in streams flowing through permafrost areas south of our study sites [42] (Table 1 and S1 Table) indicate that such a low endmember value is reasonable.

In these tundra wetlands mostly <1 m deep with almost continuous daylight and very high levels of N and P (Fig 1) [19–21, 87, 88], microalgae (mostly periphyton) likely have very high productivity and were an important carbon source for most taxa (Figs 3 and 4). The extraordinarily high pCO₂ in these wetlands (up to 5000 μatm, [22]) may result in proportionately strong selection of the light isotope of C during uptake of DIC by microalgae (see previous section), leading to very low δ¹³C values in microalgal consumers. In local wetlands, CO₂ levels in direct proximity to benthic microalgae were driven mainly by bacterial respiration [54], while very high DIC levels may in fact reduce diffusion of atmospheric CO₂ through the water column to the sediment surface [57]. In local wetlands in late July, DOC concentrations increased by almost 60% (about 11.6 to 18.5 mg L^‒1) and pCO₂ by 228% (about 688 to 1568 μatm) from 1971 to 2009‒2013, apparently associated with increased thawing of surrounding tundra [22] (see also [62, 89]). As warming and tundra thawing has continued since that time, this scenario may explain the low δ¹³C values in our invertebrates.

Within periphyton communities, cyanobacteria generally have substantially higher δ¹³C values than do other primary producers [90]. Cyanobacteria concentrate inorganic carbon more efficiently than other groups, and thus can take up higher amounts of HCO₃^‒ relative to CO₂ when CO₂ is less available [91]. Because the δ¹³C value of HCO₃^‒ is higher than that of dissolved CO₂, this difference may explain the generally higher δ¹³C value in cyanobacteria (Fig 3) [90]. However, except when CO₂ is in short supply as during intense blooms or in still waters, cyanobacteria normally take up CO₂ rather than HCO₃^‒ [92]. Another possible explanation for the much higher δ¹³C value in cyanobacteria than in eukaryotic microalgae is that some cyanobacteria are known to consume organic compounds as an alternative nutrient source when such compounds are abundant [93, 94]. Feng et al. [95] noted that N fixation is an energy demanding process, and that availability of labile, energy-rich C substrates to cyanobacteria eliminates the cost of C fixation to fuel the cost of N fixation, thereby allowing maximum growth. The value we used for δ¹³C of cyanobacteria was a measurement for mats of Nostoc spp. in our study region [43], and additional measurements are needed as pCO₂ increases in these wetlands [22].

The porosity and hydraulic conductivity of peat generally declines sharply at depths below 10‒20 cm, lateral flow through the peat in our area is mostly laminar, and deepening of the active layer above impermeable permafrost with thawing over the summer is gradual [31, 55]. Accordingly, much of the DOM leached into wetlands likely comes from relatively recently deposited plant material in the peat acrotelm, despite increasing contributions from thawing of the upper layer of permafrost containing older carbon [57]. Given that the acrotelm in peat formed mainly from Carex roots typically contains both living and dead tissue, we could not distinguish macrophytes from peat in surrounding tundra as OM sources for macroinvertebrates. ¹⁴C dating of peat from the active layer and upper layer of permafrost might help discriminate the supply and uptake of increased DOM from thawing permafrost vs. increased DOM from the active layer due to a longer annual period of thawing [42, 57, 64, 96].

An important caveat to our study is that the endmembers of epiphytic microalgae, peat, and cyanobacteria were based on averages of literature values, including those from in and near our study area, as opposed to direct measurements for our study sites (S1 Table). Detailed justification for this approach is provided in the Endmembers section of the Methods. Although we doubt that directly measured values would have differed substantially or shown greater repeatability, future studies should attempt to overcome the challenges of sampling in ways that reflect more accurately the local isotopic values of foods available to and selected by various invertebrate consumers.

Patterns of variation in invertebrates

The δ¹³C values of most invertebrate samples fell in a range between the very low endmember for fresh microalgae and the much higher endmembers for macrophytes and peat (Fig 3). These intermediate values could represent consumption of a combination of these materials, as mixing models between two endmembers are expected to indicate. However, these values might also reflect ingestion of microalgal phytodetritus that had been reworked and fractionated to varying degrees by bacteria and macrofauna [97–99]. Peat carbon leaching into the wetland as DOC must in large part be assimilated by bacteria and thereby converted to particulate form for macroinvertebrates to access that peat carbon in their diet [100, 101]; we attempted to account for such processes in the TDFs (Table 2, Fig 3). Deposit-feeders with different feeding modes perhaps ingest different mixtures of fresh microalgae and bacteria. Moreover, we expect shifts in the relative production of microalgae and bacteria depending on the relative supply of inorganic nutrients and DOC in particular wetlands [12, 102].

Compared to other taxa, Physidae snails appeared to assimilate larger amounts of cyanobacteria relative to eukaryotic microalgae (Fig 3) [cf. 103–105]. This difference may partly reflect differential dependence of snails on surface biofilms relative to subsurface deposits [cf. 104]. It is unclear how the scraping of surfaces by snail radula might differ in effectiveness for feeding on surface biofilms or “mats” of cyanobacteria [43] relative to feeding on eukaryotic microalgae that may be less consolidated and more amenable to deposit-feeding.

A number of the δ¹⁵N values of invertebrates (Fig 3) initially appear too high for the taxa to be mainly consumers of the endmembers. Thus, some of the invertebrates might be predators, meaning that fractionation values for δ¹⁵N we used in mixing models are too low by the number of trophic levels [106]. The only published values we found for fractionation (Δ¹⁵N) between fresh microalgae and insect predators of herbivorous macroinvertebrates were 1.2 to 1.9‰ (S2 Table), which are not distinguishable from values for detritivores consuming bacteria or fungi (1.70 to 1.93‰, Table 2). Moreover, N in algal- or peat-derived DOM may be fractionated repeatedly in the microbial loop of the complex periphytic community by heterotrophic microbes (Table 2, S2 Table). Variable numbers of trophic transfers in the microbial loop can lead to increases in δ¹⁵N in macroinvertebrate consumers of the periphytic community that are comparable in magnitude to increases expected in predators of macroinvertebrates [69, 107, 108].

Coleoptera were mostly predatory Dytiscidae, but their δ¹⁵N values largely overlapped those of deposit-feeders presumed to be their prey. The only taxa in our wetlands other than Dytiscidae that could function as predators are Chironomidae in the subfamily Tanypodinae. However, Tanypodinae were a small fraction of Chironomidae in our study area, and are often only facultative predators (see section on Sample collection and processing in Methods). Thus, the relatively high numbers of chironomid samples with high δ¹⁵N (Fig 3) appear to represent mostly non-predatory taxa that were deposit-feeding in patches with exceptional turnover and fractionation of δ¹⁵N in the microbial loop. The high δ¹⁵N values of some grazing Physidae snails (Fig 3) also indicate that high δ¹⁵N values among our samples of detritivores resulted mostly from spatial patchiness of N recycling by the microbial loop in periphytic communities, rather than predation. Our results suggest that high δ¹⁵N in portions of chironomid communities in other studies may in fact not reflect carnivory [cf. 106].

Microalgae were clearly a dominant or major OM source in all wetland types except Deep Open Lakes where peat was dominant. Submersed macrophytes (Arctophila), which provide surfaces for epiphytic growth, are rare in Deep Open Lakes (as well as in Deep Carex). Inputs of labile DOM from thawing permafrost are generally greater in deeper lakes, because thawing sediments underneath deeper lakes extend below the typical active layer which supplies most DOM to shallow wetlands [57, 96]. This greater input of DOM may facilitate bacteria which are probably an intermediate step in invertebrate consumption of peat carbon. Deep Open Lakes were also the only wetland type in which Physidae snails did not include cyanobacteria as a major source of organic matter (Fig 3). Thus, there may be less growth of benthic microalgae and cyanobacteria in deeper, more light-limited waters.

Relative biomasses of taxa among wetland types

Relative biomasses of invertebrate taxa varied substantially among wetland types. In cores (Fig 5), Chironomidae occurred consistently and often had the highest biomass, whereas Coleoptera ranged from having the highest biomass in Streams to being much less common or undetected in other wetland types. In net sweeps, the most striking overall pattern was that wetland types dominated by stands of macrophytes (Arctophila or Carex) had greater diversity and evenness among invertebrate taxa, as well as generally much higher total biomass of detritivores compared to Streams or Deep Open Lakes (Fig 6). These aspects emphasize the inordinate importance of shallow vegetated wetlands to the diversity and biomass of invertebrate prey available to avian consumers. Thus, although relative contributions of organic matter sources for different invertebrate taxa were largely consistent among wetland types, the greater invertebrate biomass and diversity in Arctophila and Carex wetlands, and their susceptibility to climate change [6], emphasize the importance of these wetland types in monitoring and conservation efforts [109].

Climate-driven trajectories of wetland food webs

With climate warming, inputs to tundra wetlands of DOM leached from thawing permafrost will continue to increase [20, 62, 110]. Although the nutritional quality of organic matter in the active layer above permafrost is degraded over time with repeated thawing, leaching, and refreezing, the quality of organic matter within permafrost is well preserved [20]. Thus, DOM released from thawing permafrost tends to have a higher labile component which is more readily taken up by bacteria [111–113]. The resulting increase in bacterial activity likely explains the very high and increasing pCO₂ levels in tundra wetlands near Utqiaġvik [22]. High pCO₂ levels in turn facilitate microalgal production [114]. In the 1970s and 2010s, epibenthic algae in local wetlands were not nutrient-limited [19, 21, 86, 88], and recent permafrost thawing is also releasing abundant N and P into the wetlands [20]. Reduced exposure to light owing to burial by bioturbation appeared to be the main limitation of benthic algal biomass [32], but long Arctic photoperiods and extended ice-free durations will maintain high levels of incident light. With abundant supply of nutrients, light, and CO₂, production of microalgae in these tundra wetlands is expected to be quite high (Fig 1; see also [10]).

Our results indicate that invertebrate food webs in these tundra wetlands depend strongly on direct herbivory of periphytic microalgae, and on ingestion of bacteria that consume microalgae or their exudates, vascular plant tissue or exudates, and peat DOM leached from surrounding tundra. Sources of organic matter to different invertebrate taxa in these tundra wetlands were also relatively constant across the six wetland types examined. Accordingly, we suggest that thawing of permafrost is unlikely to reduce or change these major organic matter sources or appreciably alter invertebrate diets. Thus, any climate-driven changes in invertebrate community structure may depend more on physiological responses or variations in life history relative to shifts in seasonal temperatures and phenology [115]. Indeed, macroinvertebrate assemblages in some deeper wetlands in this area showed little change between the 1970s and 2000s [11].

In our study area, shallow wetlands <1 ha declined by 30% in area and 17% in number over 65 years from 1948 to 2013 [6], a trend documented throughout circumpolar regions [5, 116, 117]. Declines in wetland area and number are accelerated by encroachment of emergent vegetation due to longer growing seasons [6, 118]. At the same time, increasing wetland water temperatures in summer [11] can lead to formation of taliks (belowground channels of thawed permafrost), resulting in rapid wetland drainage [117]. The latter effect may have already begun crossing a thawing threshold about 70 years before terrestrial permafrost thaw is predicted to become widespread in northern Alaska [119]. Shallow Arctophila and Carex wetlands, which have the highest invertebrate diversity and total biomass, are the wetlands most susceptible to these changes. Thus, climate-driven effects on overall invertebrate biomass or diversity available to avian consumers will likely depend not on shifts in OM sources, but more on reductions in overall number or area of these shallow emergent wetlands.

Conservation implications for waterbirds

Tundra wetlands in the North American Arctic support a multitude of breeding waterbirds that migrate there from eastern Asia and throughout the western hemisphere [1–4]. These wetlands provide essential invertebrate foods for adults and young of a range of species, including sensitive populations of sea ducks (Mergini) and loons [120–122]. In our study area, sea ducks that wintered on the ocean relied heavily on nutrients from freshwater sources for egg production (89–95%), and either partly or entirely on freshwater nutrients for body maintenance during incubation (58–99%; [123]). For spectacled eiders (Somateria fischeri), duckling survival is correlated with duckling growth rates; and although predation may be the main proximate source of mortality, habitat conditions including food availability are likely the ultimate factor affecting duckling survival [124]. When northern pintail (Anas acuta) ducklings were introduced into individual tundra wetlands of the Yukon-Kuskokwim Delta of western Alaska, the ducklings quickly depleted invertebrate foods so that access to multiple alternative wetlands for feeding was critical (MWC Miller, pers observ).

Relative use of the different wetland types by sea ducks varies substantially [1, 125] (Miller et al. [Unpublished]). Arctophila and Carex wetlands have the highest invertebrate diversity and total biomass, and are by far the wetland types most used by breeding sea ducks in our study area. Despite lack of major shifts in organic matter sources and invertebrate assemblages within wetland types, the availability of shallow emergent wetlands as feeding areas for breeding birds may continue to decline. Reductions in these shallow wetland types has been concurrent with long-term decreases in breeding populations of king eiders (Somateria spectabilis), spectacled eiders, and long-tailed ducks (Clangula hyemalis) across the Arctic coastal plain [126]. Accordingly, documenting trends in the occurrence and types of Arctic tundra wetlands should be an important priority for conservation [6, 127].