Abstract

Lagoons are a prominent feature of Arctic coastlines, support diverse benthic food webs, and provide vital feeding grounds for fish, migratory birds, and marine mammals. Across the Arctic, loading of terrestrial/freshwater-derived organic carbon (CT) from watershed runoff and coastal erosion is predicted to increase with global warming, and may subsidize marine organic carbon as an energy source. To assess the importance of CT, we analyzed the trophic links and carbon assimilation pathways of twenty genera in five trophic guilds (suspension and filter feeders (Su/FF), surface and subsurface deposit feeders (Ss/De), epibenthic omnivorous invertebrates (Ep/Om), omnivorous fishes (Fish), and mammalian carnivores (Mam/Carn) as well as end-member organic matter (OM) sources. Because end-members had distinct carbon and nitrogen isotopic ratios, we employed a Bayesian stable isotope mixing model (simmr) to determine the contributions of CT, shelf OM, and marine microphytobenthos, to the diets of resident fauna. Ss/De and Ep/Om mainly assimilated marine-derived OM end-members. Su/FF, Fish, and beluga whales derived large portions of their diet from CT (>40%). We conclude that (1) coastal food webs are characterized by a high degree of omnivory and plasticity, (2) CT is an important OM subsidy to food webs, and (3) omnivorous fish transfer CT from lower to upper trophic levels.
  Do high Arctic coastal food webs rely on a terrestrial carbon subsidy?
Carolynn M. Harris
a,
,1
 ,Nathan D. McTigue
b
 ,James W. McClelland
a
 , DuntonKenneth H.
a
a
  The Universityof Texas at Austin - Marine Science Institute, 750 Channel View Drive, Port Aransas, TX 78373, USA
b
    NationalOceanographic and AtmosphericAdministration, National Centers for CoastalOcean Science, 101 Pivers IslandRd, Beaufort, NC 28516, USA
a b s t r a c ta r t i c l e i n f o
Article history:
  Received 19 March 2018
  Accepted 19 March2018
 Availableonline xxxx
  Lagoonsare a prominent feature of Arctic coastlines, support diversebenthic food webs, and providevital feeding
 grounds for sh, migratory birds, and marine mammals. Across the Arctic, loading of terrestrial/freshwater-
 derived organic carbon (C
T
  ) from watershed runoff and coastal erosion is predicted to increase with global
   warming, and may subsidize marine organic carbon as an energy source. To assess the importance of C
T
, we an-
    alyzed the trophic links and carbon assimilation pathways of twenty genera in ve trophic guilds (suspension
  and lter feeders (Su/FF), surface and subsurface deposit feeders (Ss/De), epibenthic omnivorous invertebrates
    (Ep/Om), omnivorous shes (Fish), and mammalian carnivores (Mam/Carn) as well as end-member organic
  matter (OM) sources. Because end-members had distinct carbon and nitrogen isotopic ratios, we employed a
   Bayesian stable isotope mixing model ( ) to determine the contributions of Csimmr
T
, shelf OM, and marine
    microphytobenthos, to the diets of resident fauna. Ss/De and Ep/Om mainly assimilated marine-derived OM
  end-members. Su/FF, Fish, and beluga whales derived large portions of their diet from C
T
 ( 40%). We concludeN
  that (1) coastal food webs are characterized by a high degree of omnivoryand plasticity, (2) C
T
 is an important
  OM subsidy to food webs, and (3) omnivorous sh transfer C
T
  from lower to upper trophic levels.
  © 2018 Elsevier Inc. All rights reserved.
Keywords:
Food webs
Terrestrialsubsidies
  Stablecarbon and nitrogen isotopes
Simmr
BeaufortSea
1. Introduction
  Arctic estuarine ecosystems are threatened by dramatic climate-
  driven changes that are occurring on land and at sea (McClelland
    et al., 2012). These changes include warming temperatures, thawing
  permafrost, increasing river discharge, and loss of shore-fast ice,
  which when coupled with enhanced storm activity, has resulted in
 greatly increased coastal erosion rates over the past 15 years
      (Petersonet al., 2006 Lantuit et al., 2012; ;Barnhart et al., 2014). In addi-
  tion to erosion rates ofthe bordering coastline, the quantity and concen-
    trations of organic matter (OM) in freshwater discharge from Arctic
   rivers control the magnitude of terrestrial carbon input to the coastal
      Beaufort(McClellandet al., 2012;Goñi etal., 2013) Inthe AlaskanBeau-.
  fort Sea, the annual load of organic carbon delivered from both uvial
  transport and coastal erosion is roughly equal to the integrated marine
  primary production within 10 km of the coast (Schell, 1983). Because
  light limitation results in a short growing period for autochthonous
  algal sources, such as phytoplankton and benthic microalgae, Arctic
  nearshore ecosystems may rely on terrestrial and freshwater-derived
  OM inputs (collectively referred to as C
T
  herein) to sustain their
 characteristically productive and diverse food webs (Dunton et al.,
2006, 2012;McClelland et al., 2012, 2014).
  Coastal lagoons, which are bounded by barrier islands to the north
    and Alaska's Arctic slope to the south, span over 50% of the Alaskan
  Beaufort Sea coastline. These lagoons link marineand terrestrial ecosys-
  tems and support productive biological communities that provide valu-
  able habitat and feeding grounds for many ecologically and culturally
  important species.Over 150 species of migratory birds from six conti-
  nents,including waterfowl(Brown,2006),and marine and anadromous
  sh ( ), rely on Arctic lagoons and nearby rivervon Biela et al., 2011
 deltas for summer habitat and feeding grounds (Churchwell et al.,
    2016). Beluga ( ) and bowhead whales (Delphinapterus leucas Balaena
  mysticetus), which are important to the subsistence and cultural heri-
   tage of communities living on the Beaufort Sea coast, forage in the
  open water outside of lagoons during summer months (Pedersen and
Linn, 2005).
  Food webstudies in the coastal Beaufort Sea have shownthat lagoon
  benthos provide a concentrated area of preferred invertebrate prey
    items to upper trophic consumers (i.e. ;Craig et al., 1982b Dunton
   et al., 2006, 2012). Lagoon invertebrate communities are dominated
  by opportunistic omnivores and detritivores (Craig, 1984) and include
 bivalves, gastropods, polychaetes, ascidians, sponges and crustaceans
  ( ). These invertebrates employ a range of feedingDunton et al., 2012
  modes, such as deposit-feeding, lter feeding, and omnivory, and in-
  habit several habitats, including living in and on benthic sediments
    (Dunton et al., 2006;Macdonald et al., 2010). Epibenthic organisms, in
  Food Webs xxx (2018) e00081
Corresponding author.
   E-mail address:carolynn.harris@utexas.edu. (C.M. Harris).
1
  Currentaddress: Departmentof Land Resources and Environmental Science,Montana
  State University, Bozeman, MT 59717, USA.
 Contents lists available at ScienceDirect
Food Webs
 j o u r n a l h o m e p a g e : ww w . j o u r n a l s . e l s e v i e r . c o m / f o o d - w e b s
https://doi.org/10.1016/j.fooweb.2018.e00081
  2352-2496/©2018 Elsevier Inc. All rights reserved.
  particular polychaetes, mysids, and amphipods, are important prey
  items for larger consumers such as sh and birds during the brief
    open water period (Brown et al., 2012).
  Some lagoon fauna, such as mysids, amphipods, and shes, season-
    ally migrateor are advectedto adjacent nearshoreareas outsidethe bar-
    rier islands and contribute to the diets of shes and marine mammals
   living within the shelf domain ( ;Dunton et al., 2006 Fechhelm et al.,
    2009). Dominant sh species include Arctic cod (Boreogadussaida), Arc-
    tic ounder (Plueronectes glacialis), Arctic char (Salvelinus alpinus), and
     several sculpin species ( spp.) (Myoxocephalus Craig et al., 1982a,
    1982b). Seals, toothed whales, and polar bears, as well as local subsis-
   tence hunters, consume these shes (Kruse, 1991).Arctic cod, in partic-
      ular, are one of the most abundant shes in the Beaufort Sea and maybe
  a key link that transfers energy from primary consumers to upper tro-
  phic levels ( ; ;Bradstreet and Cross, 1982 Craig et al., 1982b Hop and
Gjøsæter, 2013).
  It is importantto understand what carbon/energysources are assim-
  ilated by lagoon-dwelling biota because of expected climate-driven
    changes in both terrestrial inputs and marine production in the coastal
 Beaufort Sea. Knowledge of C
T
  assimilation by a variety of consumers
  will improve our ability to predict how nearshore food webs may be af-
  fected by climate change. Stable isotope analysis (SIA) is one method of
  assessing the relative contribution of isotopically distinct OM sources to
        an organism'sdiet (Hobson et al., 1995;Phillips et al., 2014). In the east-
  ern Alaskan BeaufortSea Coast, depleted δ
13
  C values may imply greater
   reliance on terrestrially-derived organic matter and/or organic matter
 produced within freshwaters, re ecting major inputsfrom the Macken-
    zieRiver in Canada aswell as inputs from numerous smaller rivers along
        the coast (Dunton et al., 2006 Iken et al., 2010; ; Casper et al., 2015;Bell
    et al., 2016). Two other organicmatter sources arepresent in the coastal
    Beaufort Sea during the open water period: marine phytoplankton and
    benthic microalgae, or microphytobenthos (MPB) ( ).Glud et al., 2009
 These autochthonous sources are more
13
C- and
15
 N-enriched than C
T
(Gradinger, 2009).
  Previous work on Beaufort Sea ecosystems concluded that benthic
 invertebrates assimilate C
T
 during the open water period, as evidenced
by depleted
13
      C values ( ;Schell,1 983 Dunton et al., 2012). It followsthat
C
T
  may be important to apex consumers like adult shes, seals, toothed
 whales, and polar bears, though this transfer of C
T
 to upper trophic
  levels is inferred and has not been quantitatively explored. Previous
  work also showed increasingly depleted δ
13
 C values with decreasing
     longitude (west to east along the coast) in sediments (21 to 27 ;
      Dunton et al., 2012), copepods ( 20.9 to− −26.7; ),Saupe et al., 1989
   benthic invertebrates ( 19 to 26 for suspension feeders, Dunton
     et al., 1989), and bowhead whales (18.8 to 20.7in muscle tissue,
   Schellet al., 1989). This eastward
13
  C depletion is likely driven by differ-
   ences in freshwater inputs and implies the importance of C
T
 as a basal
  OM/energy source may vary spatially over the Beaufort Sea.
  The purpose of this study is to determine the reliance on OM end-
  members for twenty genera representing ve distinct trophic guilds
  and multiple trophic levels in a high Arctic coastal ecosystem. To this
 end, we quanti ed the importance of C
T
  as an OM source, relative to
  phytoplankton and MPB, to genera within each guild using the stable
 isotope mixing model simmr. We hypothesize (1) C
T
assimilation will
  occur within all feeding guilds but deposit feeders and omnivores
 would assimilate more C
T
 than lter feeders, and (2) C
T
assimilation
   will vary among lagoons, where fauna inhabiting lagoons with greater
  freshwater inputs will assimilate more C
T
.
2. Methods
  2.1. Regional setting and study area
 The Beaufort Sea coast experiences weak lunar tides (mean 10 cm)
  ( uenced by river runoff. The mouth ofNOAA, 2010) and is strongly in
    the Mackenzie River, the largest riveremptying into the Beaufort Sea,
  is 400 km east of Barter Island and discharges ~380 km
3
of freshwater
  annually ( ). The Colville River, 200 km west ofMacdonald et al., 2004
  Barter Island, is the second largest source of freshwater input to the re-
 gion (~20 km
3
y
1
   ) (McClelland et al., 2014). Numerous smaller rivers
  and tundra streams that ow directly into the lagoonsalso greatly infl flu-
  ence the physical and chemical environment, especially during the
      spring freshet period (McClelland et al., 2014 Harris et al., 2017; ). Ex-
  change between lagoons and the nearshore environment occurs via
  shallow channels between barrier islands. The shallow lagoons (b4 m)
  are warm (11 ± 2 °C (mean ± SD)) and estuarine (21 ± 7) during
    the ice-free months of the summer (July to early-September), and
 their salinity regimes vary based on rate and magnitude of freshwater
  inputs (Harris et al., 2017).
 Our study area included four shallow sites bounded by barrier
  islands (lagoon), three sites in adjacent shelf waters (nearshore), and
  four offshore sites (marine) in the eastern Alaskan Beaufort Sea coast,
 near Barter Island ( , ). Kaktovik Lagoon (KA) is almostFig. 1 Table 1
  fully enclosed and only receives freshwater inputs from small tundra
       streams and runoff(Dunton et al., 2012;Harris et al., 2017). Jago Lagoon
   (JA), a moreopen lagoon east of KA, is separated from KA by a peninsula
  and receives direct freshwater inputs from the Jago River. Angun (AN)
 and Nuvagapak (NU) Lagoons, located further east, are semi-enclosed,
  and receive inputs from smaller rivers. All nearshore sites are located
  outside of barrier islands within 1 km of the coast.
 2.2. End-member collection
 We collected suspended particulate organic matter (SPOM) for use
  as end-members from lagoon, nearshore, marine, and river sites during
  mid-August 20112014. We also collected benthic particulate organic
  matter (BPOM).Lagoon and nearshoresites were sampledevery August
  from 2011 to 2013. The four marine sites, located 20 km offshore at
    3537 m water depth,were sampled once in 2014.Two North Slope riv-
  ers were sampled for SPOM, the Jago River (six times in 2012) and the
  Hulahula River (three times in 2011). The Jago River empties into JA
  and the Hulahula River empties into Camden Bay, located 10 km west
 of Kaktovik lagoon.
   Subsurface water samples were collected using 4 L carboys sub-
    merged to approximately ~0.5 m depth (lagoon and river stations),
    3 m (nearshore), and 10 m (marine). Water samples were sieved (180
    μm) to remove zooplankton, detritus, or other large particles. Within
     hours of collection water samples were ltered in duplicate to collect
    SPOM onto pre-ashed 25 or 47 mm Whatman GF/F lters and dried at
  60 °C in petri dishes. BPOM samples were collected viasurface sediment
  syringe cores (2 cm deep) from Ekman grabs (0.023 m
2
 ), which were
  repeated twice at every lagoon site. These sediments were stored in
  plastic vials and dried at 60 °C.
  Microphytobenthos (MPB) are a potentially important food web
  end-member, so they were included in the stable isotope mixing
   model despite not being sampled directly. MPB production can equal
  or exceed phytoplankton productionin shallow ( 30 m) Arctic environ-b
  ments ( ) and is evidencedby high sediment chlorophyllGlud et al., 2009
  content at our lagoon and nearshore sites (Dunton and McClelland,
  unpublished data). MPB can possess variable stable isotope signatures
  depending on their growth rate and source of dissolved inorganic car-
      bon (Oxtoby et al.,2015 Lebreton et al.,2016; ). To encapsulate the pos-
  sible variation in stable isotope values for the Beaufort Sea lagoon MPB,
 we incorporated average values (δ
13
     C mean ± SD = 16.9 ± 1.2 ;
δ
15
  N = 6.4 ± 1.2) derived from a literature survey that has been suc-
    cessfully applied to mixing models in the Chukchi Sea (McTigue and
Dunton, 2017).
   Though Oxtoby et al. (2015) estimatedδ
13
  Cvalues for three clades of
    microalgae that might be components of MPB in the Beaufort Sea shelf
  sediments using models of
13
  C-discrimination of bottom water DIC,
   they did not estimate δ
15
  N and our mixing model requiresboth isotope
  values. Our estimated δ
13
  C values for lagoonal MPB, however, are
  2C.M. Harris et al. / Food Webs xxx (2018) e00081
slightly more
13
  C-enriched than those estimated for the Beaufort Sea
  Shelf MPB, which is reasonable assuming that growth rates are higher
 in the lagoons where photosynthetically active radiation is higher
  than at shelf sediments. The large standard deviations around the
 pooled MPB δ
13
C and δ
15
  N value reect the uncertainty in this end-
  member value that was not directly measured.
  Ice algae, comprised mainly of pennate diatoms and
  micro agellates, live on and in the sea ice matrix, and is another poten-
    tial OM source inthe nearshore regionsof the Beaufort Sea (Horner and
      Schrader, 1982;Pineault et al., 2013). It is considered a minor sourceof
  autochthonous OM in the lagoons relative to phytoplankton and MPB,
  however. Though ice algae is present during the winter and spring,
    blooms of phytoplankton and MPB in lagoons and nearshore waters
    during and shortly after ice break-up make ice algae contributions pro-
    portionally negligible (Horner and Schrader, 1982). This contrasts with
  deeper offshore waters of the Arctic Ocean, where ice algae production
    can represent 50% of water column production (NLegendre et al., 1992;
   Gradinger, 2009). It is therefore important to consider ice algae as an
  OM source to offshore consumers that inhabit the Beaufort Sea slope
  in summer months.Sea ice algae in the Beaufort is
13
C-enrichedrelative
  to water column POM, but has similar δ
15
  N values (Gradinger, 2009).
Our δ
13
C and δ
15
  N values for ice algae were observed in the Chukchi
  Sea in 2012and have large standard deviationsto reectthe uncertainty
  in this end-member thatwas not directly sampled in our study region.
 2.3. Consumer collection
  Benthic invertebrates thatlive in or on the sediments were collected
  via Ekman grabs (0.023 m
2
  ) at lagoon and nearshore sites. We obtained
     two or three grabs per station and sieved (1 mm) the contents to re-
  move sediment and retain invertebrates. Epibenthic invertebrates and
   small shes were collected via a 1 m beam trawl (1 mm mesh size)
  that was towed for ~10 min at each station. Copepods were collected
  via a vertical plankton net tow (250 m mesh). Organisms recoveredμ
    Fig. 1. Lagoon (blue symbols), nearshore (redpoints), marine (yellow symbols), and river (greensymbol) sites along the coast of the eastern Alaskan BeaufortSea (see for siteTable 1
  codes). Organic matter samples were collected from all sites and animal samples were collected from nearshore and lagoon sites. All sampling occurred during the open water period
 between 2011 and 2016.
Table 1
    Locationand hydrographic and chemicaldata (mean ± SE ( )) from coastal sites along the Alaskan eastern BeaufortSea in summer 2011n2014.Hydrographic parametersare meansof
        surface(b1 m) measurementsobtained in August at two-threestations in each lagoonand nearshoresite. Marine sites area single measurementobtained in August2014. Several stations
     at each riversite were visited in August2011 (Hulahula River)or August 2012 (Jago River),though hydrographic and chemicaldata was not obtained. indicates datawas not collected.“–”
    Site type Site Site code Latitude (°N) Longitude (°W) Temp (°C) Salinity pH Bottom depth (m) Years sampled
    Marine STA 21 70.28 143.91 36 (1) 37 2014– –
    STA 22 70.19 142.9 36 (1) 35 2014– –
    STA 24 70.26 141.76 35 (1) 52 2014– –
    STA 25 69.85 141.72 34 (1) 23 2014– –
      Nearshore Hulahula Delta HU 70.07 144.19 6.2 ± 0.7 (4) 27.7 ± 1.9 (4) 7.9 ± 0.02 (4) 4 2011, 2012
       Bernard Spit BE 70.16 143.58 6.4 ± 0.5 (6) 26.6 ± 1.6 (6) 8.0 ± 0.04 (6) 10 2011, 2012, 2013
       Demarcation Point DP 69.7 141.31 4.1 ± 1.1 (4) 23.7 ± 4.4 (4) 7.9 ± 0.03 (4) 6 2011, 2012
      Lagoon Kaktovik KA 70.09 143.61 11.6 ± 0.4 (9) 23.2 ± 0.8 (9) 7.9 ± 0.1 (9) 4.5 2011, 2012, 2013
       Jago JA 70.11 143.5 9.9 ± 0.4 (9) 23.5 ± 1 (9) 8.0 ± 0.03 (9) 4 2011, 2012, 2013
      Angun AN 69.96 142.49 10.3 ± 1.1 (6) 22.6 ±2 (6) 7.9 ± 0.01 (6) 2.5 2011, 2012, 2013
    Nuvagapak NU 69.86 142.19 10.1 ± 0.6 (6) 11.0 ± 3.5 (6) 7.9 ± 0.1 (6) 2.5 2011, 2012
         Demarcation Bay DE 69.66 141.27 8.6 ± 0.5 (6) 21.0 ± 2.3 (6) 8.0 ± 0.4 (6) 4 2011, 2012, 2013
  River Hulahula River HU-R1 69.49 144.36 2011––––
  HU-R2 69.76 144.16 2011– – – –
  HU-R3 69.98 144.02 2011– – – –
  Jago River JA-R 69.72 143.6 2012– – – –
  3C.M. Harris et al. / Food Webs xxx (2018) e00081
  from grab, trawl, and net samples were rinsed in ltered seawater,
  sorted into separate vials, and stored in ltered seawater at 5 °C for
later identication.
   Within one day of collection, all organisms were identi ed to the
  lowest taxonomic unit possible, usually species. All organisms were
  sampled in triplicate when possible, and dried to a constant weight at
  60 °C in aluminum dishes. Small organisms were dried whole
  (e.g., polychaetes, priapulids, and small crustaceans); muscle tissue
  was isolated from large organisms (bivalves, gastropods, large crusta-
  ceans, and sh) prior to drying.When taxa were small, multiple individ-
  uals of the same species were pooled to collect suf cient biomass for
  isotope analysis. All sh were handled in accordance to IACUC permit
#AUP-2012-00103.
   Most large animal tissue samples (mammals and 10 cm) weresh N
 graciously donated by local Inupiat hunters and shers or contributed
  by U.S. Fishand Wildlife Servicebiologists working withinthe Arctic Na-
  tional Wildlife Refuge. Samples were provided to us in cleansealed bags,
 labeled with the organism's common name, date and method of cap-
  ture, and approximate location. Muscle tissue was isolated from all
       large animal samples prior to drying to a constant weight at 60 °C in
 plastic snap-cap vials.
    Dried samples were transported to The University of Texas Marine
  Science Institute (UTMSI) in Port Aransas, TX for stable isotope prepara-
  tion and analysis. Over 1200 organisms were collected, from which 250
  organisms from seven phyla and twenty-four genera were selected for
  analysis based on their replication and distribution among sampling
locations.
 2.4. Trophic guild assignment
  Invertebrate species were classied a priori into trophic guilds (sus-
    pension and lter feeders (Su/FF), surface and subsurface deposit
  feeders (Ss/De),or epibenthic omnivores (Ep/Om))based on taxonomic
   data from and the World Register of MarineMacdonald et al. (2010)
  Species ( ). Data for the genus was used ifAppeltans et al., 2012
 species-speci c data was unavailable. All shes are considered to befi fi
 omnivorous ( sh). Carnivorous marine mammals (Mam/Carn) are con-
  sidered as a separate guild.
  2.5. Stable isotope analysis (SIA)
  SIAis routinely used to examinetrophic structure andcarbon assim-
  ilation pathways, and has been successfully applied to the Arctic marine
       environment (e.g. Hobson et al., 1995;Iken et al., 2005;von Biela et al.,
    2011 Dunton et al., 2012 Connelly et al., 2014 Bell et al., 2016; ; ; ;
    McTigue and Dunton, 2017). SIA provides a longer-term estimate
    (weeks to months) of diet than gut content analysis and provides infor-
  mation on what food sources are assimilated after ingestion. Typically,
δ
13
  C values experience little enrichment (~0– ‰1 ) with increasing tro-
  phic level and indicate the basal OM source ( ;Fry and Sherr, 1984
  McCutchan Jr et al., 2003). δ
15
  N values also provide information about
  basal organic matter sources, but with a ~2 to 4enrichment between
  food source and consumer that is useful for quantifying tropic position
(level).
  All end-member and consumer samples were analyzed for stable C
 and N isotope ratios. SPOM lters were analyzed separately for C and
  N isotope ratios. The C lters were triple acidi ed with 6% sulfurousfi fi
    acid to remove inorganic carbon prior to analysis; the N lters were
  not. All BPOM and consumer samples were dried and homogenized
  with a mortar and pestle. For BPOM, calcifying animals, and animals
  from which muscle tissue was not isolated, two subsamples were ana-
  lyzed. One subsample was treated with acid to remove carbonates by
  soaking in 1 N HCl until bubbling ceased, rinsed twice in de-ionized
  (DI) water, re-dried at 60 °C, and analyzed for C isotope ratios.N isotope
 ratios were determined from separate, non-acidi edfi filters. All animal
 samples that did not require acidi cation were analyzed once for dual
  C and N isotope ratios.
   Dried samples were weighed into tin capsules and analyzed for C
  and N content and isotopic composition on a Finnigan MAT Delta Plus
    continuous ow isotope ratio mass spectrometer (CF-IRMS) coupled
  to a Carlo Erba 1500 elemental analyzer (EA) at The University of
 Texas Marine Science Institute. Isotope values are expressed in delta
(δ) notation:
δ13C or δ15NR
sample Rstandard 1 x 1000
where Ris
13
C/
12
C (or
15
N/
14
  N) and the standard reference is Vienna Pee
  Dee Belemnite or atmospheric nitrogen (N
2
 ), respectively. Based on in-
  ternal standards, which were run every 12th sample, instrumental ana-
  lytical error was ± 0.2for δ
13
C and δ
15
 N. MolarC:N ratios for samples
  analyzed as acidi ed and non-acidi ed subsamples were calculatedfi fi
  using the carbon and nitrogen content in micromoles from the acidied
subsample only.
  All sampleswere collected from late-July to early-September. There-
  fore, althoughthe process of assimilating carbon intoan animal's tissues
   may take several weeks to several months (half-life for C is ~1370 days
      for an Arctic amphipod; Kaufman et al., 2008, four weeks for an Arctic
    bivalve; McMahon et al., 2006), we assume that the isotopic composi-
      tion of whole-body and muscle tissueof the animals in this study reect
 their diet post-ice break up (which typically occurs in mid-June).
  Because lipids are important energy reserves for Arctic animals
  ( ), and the process of lipid extraction mayMøller and Hellgren, 2006
  also compromise other tissue constituents, no lipid extractions were
  performed on our samples to avoid the potential loss of critical informa-
  tion and introduction of error into the food web analysis. This allows us
  to compare our data to previous work from the Beaufort Sea (Dunton
      et al., 2006, 2012 Bell et al., 2016; ) and other areas in the Arctic region
  ( ; ). Though lipidsIken et al., 2010 McTigue and Dunton, 2014, 2017
  are known to be inherently depleted in
13
  C relative to other tissues
  ( ) and may be a confounding variable inDeNiro and Epstein, 1977
  food web studies that utilize stable carbon isotopes, previous work
 shows Arctic benthic consumers have low lipid content (Graeve et al.,
     1997 Mohan et al. (2016)). A recent study by examined the δ
13
C of
  the lipid fraction of the amphipods, andCalanus copepods, Onisimus
    mysids ( ) collected in this study. Mohan et al. found that theMysis
δ
13
    C values of bulk lipids varied by ~6among the crustaceans exam-
  ined, which questions the utility of a one-size- ts-all mathematical
    equation to correct for lipid content (2016). Consequently, adjust-“ ”
  ment of the original isotopic value may introduce more bias than it
removes.
  2.6. Stable isotope models
   We quantied the contribution of various OM sources to consumers'
  diets using a stable isotope mixing model, which determines possible
  combinations of food web end-members to the consumers' diet based
  on stable isotope data. We chose to use the R package Stable Isotope
    Mixing Models in R ( ) for this analysis since it is the upgrade tosimmr
     Stable Isotope Analysis in R (SIAR) (Parnell et al., 2013). Like SIAR,
  simmr incorporates the variability in both end-member and consumer
      stable isotope values as well as the uncertainty in trophic enrichment
  factors (TEF) to provide credible intervals of possible dietary solutions
  (Parnell et al., 2010, 2013).
  TEFs have only been directly measured for a few Arctic organisms
    (i.e. McMahon et al., 2006 Vander), and are known to be variable (
    Zanden and Rasmussen, 2001 Post, 2002; ; ).McCutchan Jr et al., 2003
  Therefore,we used literaturevaluesthatencompassthese observations:
  for all invertebrate guilds (Su/FF, Ss/De, and Ep/Om), we used Δδ
13
C =
  1.0 ± 1.0and Δδ
15
   N = 3.4 ± 1.0to reect their position as primary
     consumers. The sh trophic guild used TEFs of 2.0 ± 2.0 and 6.8 ±
  4C.M. Harris et al. / Food Webs xxx (2018) e00081
2.0for Δδ
13
C and Δδ
15
  N, respectively, to reect their position as sec-
  ondary consumers compared to basalOM end-members. For the mam-
 malian carnivoregroup, Δδ
13
C and Δδ
15
   N were increased to 3.5 ± 3.5
      and 11.9± 3.5 to reect their higher trophic position as well as thein-
    crease in uncertainty in TEFs over multiple trophic steps. These TEFs
  represent expected enrichments in consumers' isotope values relative
  to basal OM sources. We performed this analysis at the genus level
within each trophic guild when replication was suf 4).ficient (nN
  simmr relies on a Markov Chain Monte Carlo to nd possible solutions
  and disregards those not probabilistically consistent with the data. The
  iterations run were 10
4
  , the burn-in was 10
3
  , the posterior was thinned
  by 10, and the number of chains t was 4.
  The model applies a geometric TEF correction to all end-simmr
  member isotope values; to graphically represent this extrapolation,
  minimum and maximum end-member isotope value for each trophic
  step are represented as:
meanendmember SDendmember mean
TEF SDTEF
 Gelman-Rubin statistics for analyses were 1.03, which indicatesb
 that the chains were sufciently long(since values 1.1 diagnose unsat-N
      isfactory runs) (Gelmanet al., 2004). The Rscripts and raw data used for
    analysis are available in the Supplemental Methods or available at
 b Nhttps://github.com/nathanmct/Beaufort-Sea-Lagoons . All analyses
  were performed in R 3.4.0 (R Core Team, 2017).
  Regression analysis relating δ
18
 O-derived meteoric water content
 of the lagoons (i.e. fresh water from rainfall/ runoff) to δ
13
C values of
  POM and con sumers was used to further elucidate potential linkages
   between terrestrial inputs and sta ble carbon isotope values . The δ
13
C
  values of water column POM and consumers, grouped by genera,
  were averaged for each lagoon and plotted as a function of meteoric
  water estimates from Harris et al. (2017). Previous studies have
shown an eastward deple tion in δ
13
C value s along the Beaufort
    coast (Dunton et al., 1989, 2012;Saupe et al., 1989), so we also
  used regression analysis to relate δ
13
C of POM and consumers to lon-
  gitude. For bot h analyses, genera were only included where at le ast 5
  samples from at least 3 lagoon sites were available. Because con-
sumer δ
13
 C data was pooled across several sampling years and sev-
  eral sampling stations within each lagoon, mean meteoric water
  values were likewise generated by pooling data across sampling sta-
  tions and over the three sampling years (Table 5).
3. Results
  3.1. End-member stable isotope compositions
  River OM, lagoon SPOM, and lagoon BPOM werenot isotopically dis-
  tinct end-members; therefore, a conglomerate of river OM, lagoon
  SPOM, and lagoon BPOM, which we hereafter refer to as terrestrial/
  freshwater-derived OM, was used as a single end-member for the
  food web ( ). Nearshore SPOM was moreTable 2
13
 C depleted in 2012
  and 2013 than in 2011, likely reecting greater C
T
  inputs in 2012 and
  2013. For the nearshore SPOM end-member in the model, wesimmr
  chose to use the 2011 mean for δ
13
C and δ
15
  N (23.0 ± 0.3 and 7.5
  ± 0.4, respectively), as this better represents typical nearshore
  SPOM composition without the bias of a potentially large C
T
input
  masking it. The marine SPOM end-member possessed slightly more
 enriched in δ
13
 C and δ
15
    N values ( 21.9 ± 0.6 and 8.9 ± 1.9 , re-‰ ‰
  spectively) than nearshore SPOM. The MPB δ
13
 C value used was
  16.9 ± 1.2, which represents a mean of benthic diatom isotope
     values reported in the literature (see McTigue and Dunton, 2017 for de-
  tails). The ice algae δ
13
   C value 17.7 ± 0.2was observed ~40 km off-
   shore in the Chukchi Sea (Dunton, pers. comm.) and was used as an
  end-member for the non-lagoonal, uppertrophic level organisms.
  3.2. Consumer stable isotope compositions
Mean δ
13
  C values for consumers ranged from (25.0Monoporeia,
    amphipod) to (17.1Weyprechtia, amphipod) and mean δ
15
N values
    ranged from 6.6 ( peracarid crustacean) to 20.3 (Mysis, Ursus
 maritimus, polar bear) (Table 3 Fig. 2, ). The δ
13
C and δ
15
N values of
  end-members (Table 2) overlapped with some consumers in the inver-
  tebrate guilds. Su/FF was the most depleted guild in
13
C and
15
N. The
      other two invertebrate guilds (Ss/De, and Ep/Om) had similar δ
13
C
and δ
15
N ranges, and were more enriched in
13
C and
15
N relative to
 the Su/FF group. The Fish guild had the largest range in δ
13
C values
 and intermediate δ
15
   N values. The Mam/Carn guild had intermediate
δ
13
  C values and was the most enriched in
15
N (except for bowhead
  whale ( )). Although every one of the ve trophic guilds over-Baleana
  lapped with at least one other guild,the most overlap occurredbetween
  the Ep/Om and Ss/De groups, both of which had intermediate δ
13
C
  values and lower δ
15
N values (Fig. 3).
  3.3. End-member OM assimilation (simmr)
  The simmr mixing modelwas used to estimatethe possible contribu-
    tion of each end-member OM source to the diet of consumers. As an in-
  dicator of the model's ability to discriminate between end-members,
  the model reports correlations of end-member probabilities for each
  consumer ( ). Large negative correlations may indicate theTable 4
  model cannot discriminate dependably between OM sources. Overall,
  the model could distinguish very reliably between terrestrial/freshwa-
  ter POM and MPB (mean correlation = 0.12), and fairly reliably be-
  tween terrestrial/freshwater POM and shelf POM ( 0.72) and
  between shelf POM and MPB (0.72).
  The proportion of end-member assimilation is represented as the
    range of the 95% credible interval(analogous to the condence interval
  in frequentist statistics) of mixing model solutions. Genera within all
  trophic guilds assimilated multiple end-member OM sources and
Table 2
   End-member organicmatter (OM) mean (±SE) δ
13
C,δ
15
 N, and C:N (molar ratio) for each
site type along the Alaskan Beaufort Sea: lagoon, nearshore, marine and river. is then
   number of samples analyzed for each parameter. Bold valueswere used as end-members
    in the simmr mixing model. *The terrestrialPOM end-member is a grand mean oflagoon
  SPOM, lag oon BPOM, and riv er SPOM. **Micro phytobenthos (MP B) data is a mea n
  resulting from a literature survey of benthic microalgae and benthic diatoms (McTigue
  and Dunton, 2017). ***Ice algae data are from the Chukchi Sea collected in 2012 (see
 Methods Section 2.2).
nδ
13
C ()δ
15
 N () Molar C:N
Marine SPOM
    2014 4 21.9 ± 0.6 8.9 ± 1.9
Nearshore SPOM
     2011 3 23.0 ± 0.3 7.5 ± 0.4 6.3 ± 0.2
     2012 3 28.5 ± 0.3 5.8 ± 1.1 7.1 ± 0.3
  2013 1 27.5 9.4 5.9
     Mean 7 26.0 ± 1.1 7.1 ± 0.7 6.6 ± 0.2
Lagoon SPOM
     2011 4 24.8 ± 0.7 6.7 ± 0.4 6.6 ± 0.3
     2012 4 28.9 ± 0.6 5.5 ± 0.9 7.2 ± 0.1
     2013 3 28.3 ± 0.3 7.0 ± 0.8 7.7 ± 0.1
     Mean 11 27.2 ± 0.7 6.3 ± 0.4 7.1 ± 0.2
Lagoon BPOM
     2011 4 27.2 ± 0.3 3.4 ± 0.2 14.1 ± 0.7
     2012 4 27.7 ± 0.8 3.0 ± 0.2 14.9 ± 1.5
     2013 3 27.0 ± 0.4 3.0 ± 0.4 14.0 ± 0.2
    Mean 11 27.3 ± 0.3 3.2 ± 0.2 14.4 ± 0.6
River POM
    2011 3 29.3 ± 0.4 3.6 ± 0.5 11.4
  2012 1 28.2 2.5 10.9
     Mean 4 29.0 ± 0.5 3.3 ± 0.5 11.1 ± 0.2
       Terrestrial POM 12 27.3 ± 1.5 4.2 ± 1.9 11.4 ± 3.8
⁎⁎     MPB 16.9 ± 1.2 6.4 ± 1.2
⁎⁎⁎     Ice algae 3 17.7 ± 0.2 4.2 ± 1.9
  5C.M. Harris et al. / Food Webs xxx (2018) e00081
Table 3
Consumer δ
13
C, δ
15
    N, and C/N (mean ± SD) for taxa collected in 20112016 from the eastern Alaskan Beaufort Sea. Organisms with n 4 were included in theNsimmr mixing model.
  Trophic guild Genus Type Total length nδ
15
N ()δ
13
 C () C/N (molar)
   Su/FF Alcyonidium Bryozoan 12 10.34 ± 0.75 22.89 ± 1.98 6.08 ± 3.19
     Calanus Copepod 5 10.69 ± 1.68 23.74 ± 1.35 5.79 ± 2.16
   Eucratea Bryozoan 5 9.20 ± 2.58 23.67 ± 2.84
     Liocyma Bivalve 16 8.87 ± 0.95 22.58 ± 1.11 5.52 ± 0.54
  Molgula Ascidian 1 8.11 23.51 5.96
     Ss/De Ampharete Polychaete 2 8.15 ± 0.95 21.42 ± 1.57 5.58 ± 0.19
  Cistenides Polychaete 1 8.94 20.18 6.71
     Cylichna Gastropod 14 10.59 ± 1.15 17.5 ± 1.47 5.89 ± 0.72
   Halicryptus Peanut worm 8 12.00 ± 1.47 19.77 ± 1.56
  Monoporeia Amphipod 1 9.58 25.02 9.50
   Pontoporeia Amphipod 10 9.40 ± 0.57 18.02 ± 1.66
   Portlandia Bivalve 6 9.24 ± 0.16 18.44 ± 1.68
     Priapulus Peanut worm 17 12.47 ± 1.36 18.28 ± 1.25 4.65 ± 0.32
   Prionospio Polychaete 4 10.19 ± 0.52 20.59 ± 1.24
   Ep/Om Atylus Amphipod 2 9.35 ± 0.29 21.46 ± 0.12
   Diastylis Cumacean 2 12.42 ± 0.46 21.6 ± 0.15 5.56 ± 0.14
  Gammaracanthus Amphipod 1 11.29 20.40 5.75
   Gammarus Amphipod 9 9.24 ± 1.00 20.23 ± 2.24
   Macoma Bivalve 3 9.00 ± 0.34 18.28 ± 3.59
  Mysis Mysid 13 10.49 ± 1.44 20.34 ± 1.74
     Mysis-NU Mysid 5 6.68 ± 0.60 17.52 ± 0.19 5.62 ± 0.82
   Nereimyra Polychaete 5 10.29 ± 1.06 20.73 ± 1.48
     Onisimus Amphipod 8 10.30 ± 1.86 19.91 ± 2.34 7.12 ± 2.00
     Saduria Isopod 5 11.97 ± 1.27 18.99 ± 1.08 6.67 ± 0.85
   Terebellides Polychaete 7 10.14 ± 0.64 20.44 ± 1.00
  Weyprechtia Amphipod 1 12.41 17.11 6.43
     Fish Boreogadus (Lagoon) Arctic cod 5 cm 6 12.00 ± 0.95 23.63 ± 1.03 4.70 ± 0.60b
     Boreogadus (Shelf) Arctic cod 3 8 cm 3 12.42 ± 0.77 24.14 ± 0.96 4.74 ± 0.42
     Coregonus Arctic cisco 15 34 cm 4 13.21 ± 0.59 21.91 ± 0.69 4.26 ± 0.39
     Eleginus Saffron cod 35 40 cm 3 15.23 ± 0.76 18.87 ± 0.21 3.74 ± 0.07
    Lumpenus Slender Eelblenny 3 11 cm 12 12.06 ± 0.92 23.71 ± 2.19 4.78 ± 0.25
     Myoxocephalus (Lagoon) Sculpin 7 23 cm 12 13.75 ± 1.96 20.55 ± 1.60 4.05 ± 0.21
     Myoxocephalus (Shelf) Sculpin 5 cm 6 12.37 ± 1.07 24.12 ± 0.91 4.89 ± 0.24b
     Pleuronectes Arctic Flounder 8 12 cm 11 12.67 ± 2.00 19.47 ± 1.58 4.22 ± 0.71
     Salvelinus Arctic char 25 55 cm 14 13.39 ± 1.88 23.55 ± 2.58 4.93 ± 1.97
     Mam/Carn Balaena Bowhead 3 13.07 ± 0.31 21.03 ± 0.75 4.51 ± 0.26
 Delphinapterus Beluga 7 17.85 ± 1.82 20.35 ± 1.80
  Erignathus Bearded Seal 1 17.29 19.13 4.23
  Ursus Polar Bear (male) 1 20.25 20.41 4.15
    Fig. 2. Jitterplot of (a) δ
13
C and (b) δ
15
    N isotope rangefor eachtrophic guild (indicatedby color). Sus/FF = suspensionand lter feeders; Ss/De = surface and subsurfacedeposit feeders;
      Ep/Om = epibenthic omnivorousinvertebrates; Fish = omnivorous shes, Mam/Carn = mammalian carnivores.
  6C.M. Harris et al. / Food Webs xxx (2018) e00081
  variation within most trophic guilds and individual genera was high,
  though this was not correlated with lagoon site or sampling year (ex-
      cept in the case of Mysis Myoxocephalusand ; see explanations below)
  ( a b, ). Overall, genera in the Su/FF guild derived most of theirFigs. 4 5
  diet from shelf POM (28 69%) and terrestrial/freshwater POM
  (2049%) and relatively small amounts from MPB. Genera in the Ss/
  De, and Ep/Om guilds derived most of their diet from MPB (30 76%)
  and small amounts of terrestrial/freshwater POM ( 20%). Fish generab
  derived most of their diet from terrestrial/freshwater POM and shelf
 POM (Fig. 5).
  Within the Su/FF trophic guild, Eucratea, Calanus and Liocyma con-
    sumed the most terrestrial/freshwater POM and the least MPB
  ( a, ). There was little variation among genera in the Ss/DeFigs. 4 5
  guild: all genera derived the majority ( 60%) of their diet from MPBN
  and appeared to assimilate little to no terrestrial/freshwater POM, ex-
  cept for , which consumed large amounts of shelf POMHalicryptus
   ( a, ). Genera within the Ep/Omgroup showed the most variationFigs. 4 5
 ( a, ). Most Ep/Om genera derived the majority ( 50%) of theirFigs. 4 5 N
  diet from shelf POM. Mysids from NU were isotopically distinct (more
depleted in
13
 C) thanmysids from other sites and were treatedas a sep-
   arate group in the model. The NU-mysids were unique in theirsimmr
   apparent assimilation of large (~70%)amounts of MPB.
  Though the sh and Mam/Carn guilds do not consume the end-
   member OM sources directly, their assimilation of basal OM sources
  was modeled by using larger TEFs to re ect the multiple trophic steps
  between them and the end-members in the food web. Within the sh
  guild, sculpin spp. (Myoxocephalus) from lagoon sites were isotopically
  distinct (more enriched in
13
C and
15
  N) from those at nearshore sites,
      and were treatedas separate groupsin the model.Most shes de-simmr
  rived the majority of their diet ( 50%) from terrestrial/freshwater POMN
  and shelf POM (20 30%), with MPB being a minor component of the
   diet ( 15%) ( a, ). LagoonbFigs. 4 5 Myoxocep halus and Pleuronectes
  showed a different pattern, deriving the majority of their diet from
  shelf POM (N40%) and small amounts of terrestrial/freshwater POM. In
 the Mam/Carn group, only beluga (Delphinapterus) was sampled in suf-
     cient replication to be included in the model. Thoughbeluga mi-simmr
  grate eastward on the Beaufort coast and feed in the nearshore
 environment, they do not typically enter the shallow lagoons.To reect
  their more offshore habitat, their end-members for included ter-simmr
 restrial/freshwater POM, marine SPOM (from marine sites), and ice
 algae. assimilated large proportions (50%) of terres-Delphinapterus
  trial/freshwater POM, and 2030% of the other OM sources ( b, ).Figs. 4 6
 3.4. Spatial differences
   Linear regression analysis revealed that % meteoric water was a sig-
 ni cant predictor of POM-δ
13
 C values (r
2
 = 0.85, 0.0001) ( ),pbFig. 7
  but not for any consumer genera δ
13
 C values (Fig. S1). Longitude was
  not a signicant predictor of δ
13
  C values for POM or any consumer gen-
 era (Figs. S2, S3).
4. Discussion
  4.1. Sources of organic matter
  Analysis of easternBeaufort lagoonand nearshoresuspended (water
    column) OM suggest there are at least two distinct sources of OM pres-
  ent in the coastal environment that are distinguishable using C and N
  stable isotopes: marine primary production (more enriched in
13
C and
15
    N) and terrestrial/freshwater-derived OM inputs. Our terrestrial/
  freshwater POM end-member encompasses the most depletedlagoon
   OM value observed in the present study ( 28.9 in 2012), tundra
  peat values reported by , and falls within the extremesSchell (1983)
  ( 23 to 31 ) observed for North Slope river SPOM by McClelland
   Fig. 3. All consumersamples plotted in δ
13
C and δ
15
 N bivariate space.Points are individual
  values and ellipses are standard ellipse areas (SEA) including 40% of the observed values.
Table 4
    Summaryof model capabilityto discriminate betweenend-members (whencalcu-simmr
    lating contributions to consumer diets).Values are correlations of end-member probabil-
 ities for incorporationinto consumer diets for individualgenera and overall (mean ± SE).
    Large negative correlationsindicate the model could not distinguish between sources.
 Guild Genus MPB ~ shelf
POM
MPB ~
terrestrial POM
 Shelf POM ~
terrestrial POM
  Sus/FF Alcyonidium 0.65 0.07 0.80− −
 Calanus 0.53 0.12 0.91− −
 Eucratea 0.38 0.28 0.78− −
 Liocyma 0.87 0.53− −0.88
 Dep/SS Cylichna 0.90 0.14 0.55− −
  Halicryptus 0.91 0.45 0.78− −
  Pontoporeia 0.76 0.21 0.47− −
  Portlandia 0.76 0.42 0.27− −
 Priapulus 0.89 0.46 0.81− −
  Ep/OM Gammarus 0.74 0.16 0.53− −
 Mysis 0.90 0.45 0.8− −
  Mysis-NU 0.69 0.50 0.28− −
 Nereimyra 0.84 0.28 0.75− −
 Onisimus 0.86 0.13 0.61− −
 Saduria 0.88 0.19 0.64− −
  Terebellides 0.88 0.47− −0.84
Fish Boreogadus
(Lagoon)
 − − 0.50 0.04 0.84
Lumpenus 0.51 0.03 0.85− −
Myoxocephalus
(Lagoon)
 − −0.87 0.60 0.91
Myoxocephalus
(Shelf)
 − − 0.34 0.2 0.84
  Pleuronectes 0.84 0.45 0.86− −
 Salvelinus 0.43 0.08 0.93− −
     Grand Mean 0.72 ± 0.19 0.12 ± 0.32 0.72 ± 0.19
 Ice Agae ~
marine POM
 Ice Agae ~
terrestrial POM
 Marine POM ~
terrestrial POM
  Mam/Carn Delphinapterus 0.62 0.09 0.72− −
Table 5
H
2
O-δ
18
    O (mean± SE)values from water samples collected from 1 to 3 m at each site in
  August 20112013. Meteoric water is the modeled percent of each lagoon determined to
 be from river inputs or precipitation basedon a δ
18
  Oand salinity mixing model.Data from
  Harris et al. (2017).
  Lagoon site Site Code n δ
18
 O () Meteoric water (%)
   Kaktovik KA 19 6.8 ± 0.2 28.8 ± 1.1
  Jago JA 13 6.0 ± 0.3 24.7 ± 1.7
  Angun AN 10 6.5 ± 0.7 27.4 ± 3.5
   Nuvagapak NU 11 11.4 ± 1.7 50.8 ± 8.0
  Demarcation Bay DE 8 6.7 ± 1.0 27.5 ± 4.9
  7C.M. Harris et al. / Food Webs xxx (2018) e00081
  8C.M. Harris et al. / Food Webs xxx (2018) e00081
    et al. (2014). We chose to analyze lagoon OM and terrestrial/freshwater
  POM as a single food web end-member because lagoons contain large
   amounts of terrestrial/freshwater-derived material in August
   (Connelly et al., 2015) and they were not isotopically distinct for treat-
  ment in the mixing model.
C
T
  reaches the coastal environment from watershed runoff and
  coastal erosion as the land margin of the lagoons actively recedes
 ( ). Erosional inputs and runoff of terres-Jorgenson and Brown, 2005
  trial/freshwater sources mix with marine-derived primary production
  in the coastal environment to produce the intermediate OM δ
13
C and
δ
15
    N values observed at some locations ( ,Table 2 Connelly et al., 2015).
  Although riverson the North Slope delivermore dissolved OMthan par-
    ticulate OM (unless inputs from the Colville River are considered), POM
   inputs from smaller rivers may be an important OM source for lagoon
     consumers (Macdonald et al., 1998;McClelland et al., 2014). In addition
to C
3
   plant-derived organicmatter from the tundra environment, the C
T
   pool includes freshwater algae carried in river water (Churchwell et al.,
  2016) and phytoplankton producedwithin the lagoonsunder low salin-
  ity conditions. Freshwater algae are typically more
13
C-depleted than
  marine phytoplankton because a larger proportion of the inorganiccar-
  bon that they use for photosynthesis comes from CO
2
 as opposed to bi-
  carbonate (Marty and Planas, 2008).
  Marine primary production includes phytoplankton and
  microphytobenthos, which can be further distinguished based on
13
C
  content [literature values of MPB are more
13
 C enriched (e.g. Newell
      et al., 1995,McTigue and Dunton, 2017) than marine SPOM values re-
  ported here]. Though lagoonand nearshore SPOMboth contained C
T
, la-
  goon SPOM is characterized by a higher proportion of dino agellates
  relative to diatoms and has a higher polyunsaturated fatty acid content
  than SPOM from nearshore sites ( ). This ndingConnelly et al., 2015
  suggests lagoon SPOM is likely an important food source for consumers
    because of its high nutritional value (Connelly et al., 2015).
  While primarily comprised of marine phytoplankton, our marine
  SPOM samples ( m size fraction) represent a mixture of materialb180 μ
   that could include resuspended benthic organic matter, small zooplank-
     ton, heterotrophic protists, and fecal pellets and other detritus. The in-
  clusion of MPB and/or resuspended benthic organic matter that has
  been microbially reworked may contribute to our enriched
13
C values
    ( ). Mid- to late-summer phytoplankton productionMcTigue et al., 2015
    is low in lagoons, because of the low inorganic nitrogen concentrations
  ( ), which implies that MPB may be an importantDunton et al., 2012
 source of nutrition for biota. Moreover, MPB is a
13
C-enriched end-
  member that explains the enriched δ
13
  C values of some consumers
  like , ,Cylichna Pontopo reia Portlandia, and Priapulus. The δ
13
C value
    ( ) of the nearshore SPOM end-member agrees well23.0 ± 0.3
    with the mid-water column POM value from the nearshore Beaufort
      Sea shelf (24.8) reported in Dunton et al. (2012). This value repre-
  sents a mixture of the more
13
 C-enriched marine SPOM and more
13
C-
 depleted lagoon/freshwater POM. Nearshore SPOM sampled in the
  years 2012 and 2013 contained a higher proportion of lagoon/freshwa-
   ter POM, as reected in their
13
  C-depleted values, relative to the sam-
   ples from 2011 (Table 2).
  Ice algae, which are present in the Beaufort Sea duringnon-summer
  months, are a distinct fourth carbon source available in the nearshore
    environment that can have similar isotope values to that of MPB
   (Gradinger, 2009;Pineault et al., 2013). Our estimated value is similar
  to the wide range of values observed in the Beaufort Sea ( 25 to
    14, (Gradinger, 2009 Pineault et al., 2013 Bell et al., 2016, , ). Ice
  algae are not typically found in coastal environments past June,
  however, and our sampling occurred during the open water period in
      August. In addition, no ice algae were observed in any of these lagoons
    during spring 2012 and 2013 (Dunton and McClelland, pers. obs). Con-
  sequently, its role as a carbon source in the nearshore environment in
     August is likely minimal, though itis more important indeeper, offshore
    waters (Legendre et al., 1992 Connelly et al., 2015; ).
 Although there was high interannual variability in δ
13
C and δ
15
N
 values for lagoon and nearshore OM reported here (as in Connelly
    et al., 2015), the means for all years are similar to those reported for
  other studies located in the coastal eastern Alaskan Beaufort Sea
      (Dunton et al., 2006,2012 Bell et al., 2016; ). Therefore, we are condent
  that the lagoon and nearshore POM isotopic means reported here are
  robust and representative of typical open water values. Interestingly,
       the agreement of our valueswith those of Dunton et al. (2012) alsosug-
  gests that despite the changing coastal environment over the last de-
  cade, the bulk composition of OM during summer, as re ected in
     stable isotope values along the coast, has not changed substantially.
  4.2. Food web structure of the eastern Alaskan Beaufort Sea
  Omnivory is a common feeding strategy in the Arctic (Dunton and
   Schell, 1987) and, as in other marine environments, true herbivores
  are rare. All three invertebrate feeding groups had similar δ
15
N ranges,
  indicating the organisms they contain occupied similar trophic levels
     near the base of the food web but assimilated different proportions of
  OM sources. In particular, most overlap occurred between the Ep/Om
  and Ss/De trophic guilds, which re ects the plasticity of feeding behav-
   ior in these organisms and illustrates the substantial degree of trophic
  redundancy within these estuarine systems. The stable isotope values
  for the invertebrate feedingguilds Su/FF, Ss/De,and Ep/Om were encap-
  sulated by those of the terrestrial/freshwater-derived OM andMPB end-
   members, which not only indicates that consumer isotope values can be
  explained by therepresentativeend-members butalso meets the condi-
    tions to run a stable isotope mixing model (Phillips et al., 2014).
  Other Arctic studies, including those in the Beaufort Sea, have also
  found benthic invertebrates to be more
13
 C enriched than marine
    POM ( ; ;Dunton et al., 1989 Hobson et al., 1995 McTigue and Dunton,
    2014 Connelly et al., 2014; ). In our model, the
13
 C enrichment of con-
    sumers relative to shelf POM indicates the assimilation of MPB. This en-
 richment, however, may also re ect assimilation of microbially-
  reworked organic matter, because microbial processes often result in
  isotopically-enriched OM substrate (Macko et al., 1987) or
13
C-
  enriched bacteria themselves ( ). This explanationOakes et al., 2016
 has been proposed to explain
13
   C enrichmentof fauna relative to phyto-
  plankton in the Bering and Chukchi Sea (McConnaughey and McRoy,
    1979 Hobson et al., 1995 Lovvorn et al., 2005 McTigue and Dunton,; ; ;
     2014 North et al., 2014; ; McTigue et al., 2015) and for
13
C enrichment
  of benthic POM relative to suspended POM in near-bottom waters
 near Spitsbergen and on the Beaufort Sea shelf (Tamelander et al.,
  2006;Connelly et al., 2012).
  Overall, genera in the Su/FF guild were the most depleted in
13
C and
15
  N, reecting the assimilation of large amounts of C
T
 . The individuals
      analyzed of the bivalve Liocyma n( = 16), a surface suspension-feeder
    ( ), exhibit little variation in their stable isotopeDunton et al., 2012
  values across three sampling seasons and two sampling locations,
    which is re ected in their small range of range of diet propor-simmr
  tions. The dietary contribution of terrestrial/freshwater POM for
  Liocyma is likely preferential consumption of freshwater phytoplankton
  since these bivalves are suspension feeders of microalgae; it is possible
  Fig. 4. Faceted δ
13
C and δ
15
  N bivariate plots showingall consumer samples by feeding guild (a: Sus/FF, Dep/Ss, Ep/Om, Fish, b: Mam/Carn). Points are individual values; symbol color
      indicates sampling location, and symbol shape indicated sampling year (note: colors repeat in each guild). Only genera with 4 that were included in thesimmr model are shownnN
  (except for Mar/Carngroup, all data shown). Boxes represent the position of end-member carbon sources (a: terrestrial POM, shelf POM, and MPB; b: terrestrial POM, marine POM,
     and ice algae) in isospace after projecting for different trophic levels (see for more details). (a) from Nuvagapuk Lagoon (NU) were distinct from from all otherTable 2 Mysis Mysis
   sites, so individuals from this site were analyzed separately. Because Myoxocephalusfrom lagoon and nearshoresites differed in δ
13
C and δ
15
  N, this genus is plotted separately for each
site type.
  9C.M. Harris et al. / Food Webs xxx (2018) e00081
  they are indiscriminately consuming and assimilating suspended ter-
    restrial detritus, but seems unlikely. Similarly, our results indicate that
  Calanus copepods, which are herbivorous and consume microalgae
   and phytoplankton (Macdonald et al., 2010), assimilate both terres-
  trial/freshwater POM and shelf POM. This suggests consumeCalanus
  both freshwater and marine-derived phytoplankton. It is also possible
  that Su/FF genera are assimilating carbon from semi-labile river-
  supplied DOM that has been remineralized microbially into dissolved
   inorganic carbon (DIC) and taken up by phytoplankton that retain the
lighter δ
13
    C signal (Sipler and Bronk, 2015). Though previous research
   showed the bryozoan Alcyonidium consumes POM and phytoplankton
    and the bryozoan graze mainly on microalgae (Eucratea Macdonald
   et al., 2010), the range of δ
13
  C values for these genera indicate assimila-
  tion of MPB and terrestrial/freshwater POM, suggesting they are highly
    plastic in thetypes of microalgaethey assimilate(i.e. assimilatemultiple
  isotopically-distinct OM sources) and are opportunistic feeders.
    The Ss/De and Ep/Om groups had similar δ
13
C and δ
15
N ranges
  (Fig. 3) despite feeding primarily on different sediment horizons (sub-
    surface vs. surface) ( ;Macdonald et al., 2010 Appeltans et al., 2012).
  Studies in the adjacent Beringand Chukchi Seas have found that surface
  andsub-surface feedershave similar C and N isotopic composition,indi-
  cating OMdeposited on the sediment surfaceis mixed to sub-surface via
    bioturbation (McTigue and Dunton, 2014, 2017 North et al., 2014; ). Ss/
  De genera showed little reliance on terrestrial POM and were more
13
C-
  enriched than other consumers, which is surprising given the depleted
δ
13
    C values of sediment observed in this study. This nding suggests
  that these deposit-feeders may be preferentially consuming and/or as-
 similating carbon from benthic microalgae (as opposed to whole
     Fig. 5. Boxplotsshowing modeled proportionsof end-member OM sources in consumer diets for each generawithin each lower-trophic feedingguild. The centerline is the median of all
  model solutions, the box encloses 25th and 75th percentiles, and the whiskers extend to the 2.5th and 97.5th percentiles. Outliers were omitted for clarity.
     Fig. 6. Boxplots showing modeled proportions of end-member carbon sources in beluga
    whale (Delphinapterus) diet. The centerline is the median, the shaded box encloses 25th
    and 75th percentiles, and the whiskers extendto the 2.5th and 97.5th percentiles. Only
    beluga was included in this analys is because of low sample size for other mammals.
  Becausebeluga whales are a top predator, the TEFs for this analysis were modied (δ
13
C
  = 3.5 ±3 .5, δ
15
   N = 111.9 ± 3.5).
   Fig. 7. Relationship between mid-watercolumn POM δ
13
  C valueand % meteoric water for
  16 samples collected from ve lagoon sites in the Eastern Alaskan BeaufortSea. POMδ
13
C
   value decreased signi cantly with increase % meteoric water (as estimated from aδ
18
O
    and salinity mixing model in Harris et al., 2017).
  10 C.M. Harris et al. / Food Webs xxx (2018) e00081
       sediments) or the labile OM that is stored in nearshore sediments in the
    Arctic Ocean ( ; ). Labile OM isMincks et al., 2005 McTigue et al., 2015
typically more
13
  C-enriched than other components of the sediment
 ( ). performed a similarMcMahon et al., 2006 Dunton et al. (2012)
  study on lagoon and nearshore consumers in the eastern Alaskan Beau-
  fort Sea and also found that deposit-feeders and detritivoreswere more
13
    C-enriched relative to suspension-feeding or omnivorousinvertebrate
   consumers.Similarly,McTigue and Dunton(2014, 2017) examinedcon-
  sumers in the Chukchi Sea and found that surface and sub-surface de-
  posit feeders were consistently enriched in
13
C and concluded they
    assimilated more ice algae or MPB relative to particulate OM.
 The intermediate δ
13
  C values of the Ep/Om guild agree well with
    other benthic studies in the Beaufort ( ) andDunton et al., 2006, 2012
  Chukchi Seas ( ; ). Ep/OmIken et al., 2010 McTigue and Dunton, 2017
    genera may consume POM directly (e.g. amphipod;Onisimus Carey
  and Boudrias, 1987), preferentially consume surface detritus (ex.
    Terebellides worm; (Macdonald et al., 2010)), or switch between preda-
  tory and scavenging feeding modes (e.g. isopod;Saduria Kvach, 2009,
    Dunton et al., 2012). These differences in feeding mode (and trophic
 level) likely cause the range in δ
15
 N values observed (Bunn et al.,
  2013; ).McTigue and Dunton, 2014
 Curiously, from Nuvagapuk exhibitedMysis δ
13
C values 2 more
13
     C-enriched than mysids from any other lagoon site and were more
15
  N depleted than the MPB pool (after adjustment for TEF) despite hav-
 ing similarδ
13
  C values (Fig. 4a). Their position might reectassimilation
  of microbially reworked terrestrial POM, which may retain a light δ
15
N
  value but become enriched in
13
   C (Macko et al., 1987 Connelly et al.).
  (2015) reports that invertebrates contained large contributions from
  bacterial fatty acids, suggesting that a depleted δ
13
  C signal in lower tro-
  phic invertebrates (i.e. TL 3) may re ect eitherthe direct consumptionb
    of terrestrially-derived OM or indirect terrestrial OM consumption via
  bacterial food webs. The depleted isotope values may also indicate the
  mysids were assimilating cyanobacteria, which is present in lagoon
 margins and has a δ
15
N value of ~0(Dunton, pers.comm.).
  The sh guild similarly displayed a wide range ofδ
13
 C values, sug-
  gesting Beaufort coast sh are opportunistic generalists who derive
  their diet from many basal OM sources. This group is comprised of om-
  nivorous shes of many sizes (from 5 cm (total length) sculpin
    (Myoxocephalus) and eel blennies (Lumpenus) to N120 cm Arctic char
 (Salvelinus)). Many of these shes are gape-limited (i.e. prey size is lim-
  ited by how wide they can open their mouths) and the large δ
15
N range
  of this guild reects the different trophic positions of individual sh (i.e.
 smaller shconsume copepods,polychaetes,and amphipods,and larger
  fi fish consume larger crustaceans and small sh) ( ;Harris, 1993
      Macdonald et al., 2010;Walkusz et al., 2011). For example,mature Arc-
  tic cod ( ) feed mainly on amphipods and mysids, thoughBoreogadus
  zooplankton are a substantial food source for juveniles (Craig et al.,
     1982a, 1982b Walkusz et al., 2011 Dunton et al., 2012; ; ). Arctic char
  (Salvelinus) derived ~40% of their diet from shelf POM and ~40% from
  terrestrial/freshwater POM. These ndings imply thatSalv elinus are
    plastic in their assimilation of basal OM sources, which might reect a
    generalist, omnivorous style of feeding. The size range of this genera
  was relatively large (25 55 cm), however, so this range in OM source
 assimilation and large isotopic niche may re ect the inclusion of indi-
  vidual shes feeding at different trophic levels due to ontogenetic diet
shifts.
    Because many of these shes derived large portions of their diet
  from terrestrial OM, they likely consumed prey items that primarily as-
  similated terrestrial OM. Su/FF invertebrates,such as bryozoans and co-
  pepods, may be important components of sh diets and create one
    pathway for terrestrial OM to propagate to intermediate trophic levels.
   Calanus copepods, who appear to assimilate large portionsof terrestrial
 OM (likely because they consume freshwater phytoplankton), are resi-
  dents of the nearshore ecosystems and can be advected into lagoons
  where they are consumed by larger invertebrates and small shes
  ( ). Though Ep/Om invertebrates, such as worms,Dunton et al., 2012
  amphipods, and mysids, are some of the most abundant prey items in
    the coastal Beaufort Sea ( ; ), theseDunton et al., 2006 Bell et al., 2016
  genera consumed small amounts of terrestrial/freshwater OM and
  therefore are likely not the primary prey items of coastal shes.
  The Mam/Carn guild has a wide δ
15
 N range, reecting the inclusion
  of consumers that span several trophic levels. Bowhead whales
    (Baleana) feed on lower trophic level crustaceans (Lowry et al., 2004)
  whereas large belugas (Delphinapterus Erignathus) and seals ( ) feed on
    large shes ( ; ). Polar bears,Frost et al., 1993 Quakenbush et al., 2015
    apex predators, feed on large shes and seals (Frost and Lowry, 1984;
    Cherry et al., 2011). Gut content analysis of migratory belugas that in-
  habit the coastal eastern Alaskan Beaufort Sea in summer shows these
   whales are piscivorous and that saffron cod and Arctic cod are major
     diet components (Quakenbush et al., 2015). Consumption of omnivo-
  rous shes, in particular Arctic cod, may explain why terrestrial/fresh-
    water OM was found to be a major basal OM source to beluga diet in
this study.
 4.3. Spatial differences
The δ
13
  C of water column POM showed signicant spatial variance
      related to differencesin the magnitude of freshwater inputs(% meteoric
    water) among lagoons (Fig. 7). In contrast, none of the consumer genera
  that met our analysis criteria showed signicant spatial variance related
     to % meteoric water or longitude (Figs. S1, S3). Consumers' δ
13
C values
  integrate food sources over several months whereas the POM samples
  represent a shorter temporal window. This spatial homogeneity in con-
sumer δ
13
  C values across severallagoons further justies our decision to
  pool consumers from all lagoon sites for our simmr mixing model,
   though it is important to note the low sample size for each consumer
 genera and the lack of coverage across the full range of % meteoric
   water spectrum reduced the power of these analyses. It is also possible
 that the near-bottom salinities and the source of terrestrial/freshwater-
  derived organic matter are uncoupled because C
T
  can rain down from
    fresher waters near the surface of the lagoon. So animals could be feed-
  ing in a layer ofsalty bottom water frommarine intrusion andreceive C
T
 from fresher waters above.
   Sculpin genera ( ), however, showed spatial differ-Myoxo cephalus
 ences in δ
13
C and C
T
  assimilation between lagoon and nearshore sites.
  Lagoon sculpin were isotopically distinct from those collected from
  nearshore sites ( ). Lagoon sculpin assimilated more marineFigs. 5, 8
    OM sources (MPB and shelf POM) and less terrestrial/freshwater OM.
  This nding is consistent with previous isotope studies that concluded
  sculpin rely on a marine phytoplankton carbon source acrossthe Beau-
  fort Sea (sculpin values range from 20.5 23.0– ‰; ;Schell, 1983 Loseto
      et al., 2008;Dunton et al., 2012). Nearshore sculpin, however, derived
  N50% of their diet from terrestrial/freshwater carbon.
  This difference in stableisotope values for the two groups can be ex-
  plainedin two ways: 1) OM source assimilation is related tothe size and
   life history stage of individuals sampled (e.g., ontogenetic shifts in tro-
    phic habits),or 2) the location of their capture drives this trend with dif-
  ferent OM sources available at different sites. Nearshore sculpin were
  uniformly smaller (total length: 5 cm) than those collected from la-b
 goons (total length: 7 23 cm). Linear regression analysis revealed a sig-
  nicantpositive linear relationship between bothC and N isotopes and
    total length in sculpin (Fig. 8), which is indicative of ontogenetic diet
  shifts and suggests that smaller sculpin assimilate more terrestrial/
   freshwater OM. It is also possible that higher lipid-content may cause
  smaller sh to be more
13
 C depleted. Younger shes tend to contain
    more fats in their muscles than adult organisms ( ;Jobling et al., 1998
    Kiessling et al., 2001), which agrees with C:N values reported here
     (4.9 for small nearshore sh, and 4.1 for larger, lagoon sh). Juvenilefi fi
 sculpin may also feed more on lipid-rich copepod nauplii (Scott et al.,
  2000), whereas adult sculpin feed mostly on peracarid crustaceans,
  such as mysids and amphipods, which store less lipids than copepods
    such as Calanus hyperboreus (Connelly et al., 2012).
  11C.M. Harris et al. / Food Webs xxx (2018) e00081
  Motile organisms can travel between lagoon and open water areas,
  mysids and amphipods are known to migrate between lagoons and
 nearshore waters ( ). Phytoplankton may also be advectedCraig, 1984
 into lagoons through channels. These processes help explain the high
   invertebrate biomass observed in lagoons (Craig, 1984). These move-
  ments would further blur any differences in terrestrial/freshwater- vs.
  marine-derived OM assimilation, should they exist, among lagoons
  and between lagoon and nearshore sites.
  4.4. Conclusions and implications for a changing Arctic
  While terrestrial/freshwater contributions to food webs have been
    well dened in temperate estuaries, their role in Arctic food webs has
  been given far less attention. This study builds on work from the past
     decades to assess the role of terrestrial/freshwater carbon in Arctic
  coastal food webs. Genera within every trophic guild derived 10% ofN
   their diet from terrestrial and/or freshwater OM, though this subsidy
  was assimilated in greater proportions by suspension feeding inverte-
  brates and omnivorous shes. This nding corroborates previous stud-
  ies that found benthic food webs are supported by terrestrial carbon
    subsidies ( ; ). These ndingsDunton et al., 2006, 2012 Bell et al., 2016
  con rm that terrestrial/freshwater OM is currently an important sub-
    sidy to multiple trophic levels in coastal Arctic waters.
  We provide compelling evidence for the transfer of terrestrialand/or
  freshwater OM from benthic food webs to upper level consumer spe-
    cies, such as beluga whales. Omnivorous shes, in particular Arctic
    cod, which is one of the most abundant Arctic shes and is known to
  be a key link between benthic and pelagic organisms (Craig et al.,
   1985 Hop and Gjøsæter, 2013; ), may be vital in the transfer of terres-
  trial/freshwater OM subsidies to upper trophic levels. More research is
 needed that examines multiple terrestrial biomarkers to determine
 this link conclusively.
  Climate change is predicted to alter carbon sources in the Alaskan
    Beaufort in severalways, which mayin turn affect the relative amounts
  of energy sources available to coastal food webs. Net marine primary
  productivity, which is tightly coupled to sea ice cover, will likely in-
     crease (Arrigo et al., 2008), but the balance between benthic (MPB)
 and pelagic (phytoplankton) production may change (Glud et al.,
    2009). Terrestrial/freshwater inputs will be affected by changing river
   inputs ( ), decreased glacial runoff (Peterson et al., 2006 Nolan et al.,
  2011), and increased coastal erosion due a longer open water period,
  larger areasof open water, increased storminess, and decreased perma-
   frost ( ; ). It is not yet clear ifOvereem et al., 2011 Barnhart et al., 2014
   these combined phenomena will ultimately result in increased or de-
 creased delivery of C
T
  to the coastal Beaufort Sea or how this may affect
    in situ lagoon and nearshore production. Increased terrestrial/freshwa-
 ter inputs may lower production because erosional processes are often
  associated with higher water turbidity ( ), or this terres-Glud et al., 2009
    trial/freshwater OM may be readilyassimilated by lagooninvertebrates.
 We show that some terrestrial/freshwater OM, which is incorporated
  into benthic invertebrates, is also transferred to upper trophic levels
  such as omnivorous shes and beluga. As such, these larger animals
  will likely be affected by aspects of climate change that affect land-sea
  coupling as well those that cause sea-ice loss.
   Supplementarydata to this article can be found online at https://doi.
org/10.1016/j.fooweb.2018.e00081.
Acknowledgements
  The authors gratefully thank the anonymous reviewer whose con-
  structive criticisms greatly improved the quality of this manuscript.
    We thank the Captainsand crews of the R/V Proteus and the R/V Norse-
    man II for their invaluable navigation skillsand support of our research.
   We are very grateful to eld assistants R. Thompson, S. Linn, S. Mohan,
  and J. Smith for their long hours and enthusiastic support of this work.
  T. Connelly contributed greatly to the eldwork (sample collection
  and processing) and provided substantive comments that signicantly
    improved the quality of early versionsof this manuscript. This eld in-
   tensive campaign would not have been possible without the support
    of D. Payer and others at the Arctic National Wildlife Refuge and at
  U.S. Fish and Wildlife Service headquarters in Fairbanks. We would also
  like to thank K. Jackson(Texas based logistic support), S. Schonberg (as-
  sistance with ArcGIS and invertebrate identi cation), P. Garlough (iso-
    tope sample processing), M. Khosh (river sampling), M. Nolan (river
  sampling) and R. Churchwell (river sampling). Funding for this work
  was provided by the National Science Foundation under grant ARC-
  1023582 and LTER grant OPP-1656026. This research was performed
   while N. McTigue held an NRC Research Associateship award at the
  NOAA National Centers For Coastal Ocean Science.
References