Draft Interactions between climate and landscape drive Holocene ecological change in a High Arctic lake on Somerset Island

Journal: Arctic Science Manuscript ID AS-2016-0013.R1 Manuscript Type: Article Date Submitted by the Author: 14-Sep-2016 Complete List of Authors: Paull, Tara; University of Ottawa, Geography, Environment and Geomatics Finkelstein, Sarah; University of Toronto, Office, 3129 Earth Sciences Centre Gajewski, Konrad; University of Ottawa, Geography, Environment and Geomatics Keyword: paleoclimate, lake sediments, diatom dissolution, pollen, paleolimnology


Introduction
Understanding long-term climate impacts on the aquatic communities of Arctic lakes requires separating the effects of climate variations from changes within the lake and in the watershed, as well as from the processes associated with lake ontogeny over the long time span of the Holocene (Fritz and Anderson 2013;Rühland et al. 2015).Numerous non-climate factors have been shown to influence diatom assemblages and production including limnological variables, bedrock geology and other watershed characteristics (e.g., Lim et al. 2001;Bouchard et al. 2004;Fortin and Gajewski 2009;Finkelstein et al. 2014).Changes through time of alkalinity, for example due to vegetation development or climate, alter the lake environment and aquatic communities (Anderson et al. 2008).However, comparative analyses have shown that diatom communities differ greatly between nearby sites due to factors of lake morphometry, dispersal and watershed characteristics (Smith 2002;Finkelstein andGajewski 2007, 2008;Rühland et al. 2015).
Diatom responses to pH are well documented, and in many cases, quantitatively reconstructed.Changes in pH through time in Arctic sediment cores may be inferred through transfer functions based on a modern calibration set and fossil diatom assemblages (Finkelstein et al. 2014).Changes in pH can also be reconstructed using indices of past production and lake chemistry (loss-on-ignition, LOI, and biogenic silia, BSi) (Fortin and Gajewski 2009).Bedrock lithology is the strongest predictor of lake water pH in Arctic systems; lakes with lower pH and poor buffering capacity are found on crystalline rocks whereas the large parts of the Canadian Arctic underlain by carbonate rocks support more alkaline lakes with significant buffering capacity (Fortin and Gajewski 2009).Superimposed on the primary control imposed by bedrock, variability in lake water pH over time may be in part influenced by climate, via fluctuations in D r a f t 4 dissolved inorganic carbon (DIC) mediated by lake ice phenology, in the more poorly buffered sites (Koinig et al. 1998;Wolfe 2002).Since a primary control of diatom communities in Arctic lakes is pH (Finkelstein et al. 2014), this mechanism likely affects diatom communities in poorly buffered lakes, particularly during cold periods when the ice cover persists over much or all of the summer.On millennial time scales, diatom-inferred pH and climate have been shown to be tightly coupled in poorly buffered High Arctic lakes (Wolfe 1996(Wolfe , 2002;;Michelutti et al. 2007).
Other long-term controls on the pH of Arctic lakes may relate to watershed processes, and landscape development (Law et al. 2015).
Autecological studies in the Canadian Arctic Archipelago have determined habitat preferences of individual diatom species based on their distribution and abundances in different habitats (e.g., Bouchard et al. 2004;Cremer et al. 2005;Karst-Riddoch et al. 2009).Planktonic diatoms, in Arctic lakes Cyclotella sensu lato, often increase in both relative and absolute abundances in response to availability of open water habitat during warm climate intervals (Douglas and Smol 1999;Rühland et al. 2008;Devlin and Finkelstein 2011;Rühland et al. 2015).Conversely, relative abundance of benthic and epiphytic diatoms often increase when open water habitats are less available during longer ice-covered seasons, or when suitable littoral and periphytic habitats develop (Smol et al. 2005).Further, diatom production and diversity are often highest during warm intervals when ice cover is reduced and the growing season lengthens (Douglas and Smol 1999), although the diatom-temperature relation is complex (Anderson 2000).Changes in habitat availability and growing season length may result in increased diatom diversity and production.Although most of these responses are observed in post-19 th century records, these patterns of community response are not always consistent in records that span several millennia (LeBlanc et al. 2004;Michelutti et al. 2007;Finkelstein and Gajewski 2007).

D r a f t 5
In order for diatom communities to provide a reliable record of environmental change, the assemblages must be well preserved in lake sediment records.Some studies have shown diatom-free zones in Holocene cores from the Canadian Arctic, which could either reflect dissolution of diatom frustules after deposition, dissolution between the death of the diatom and deposition in the core or absence of living diatoms during the time of sediment deposition (e.g., Smith 2002;Podritske and Gajewski 2007;Peros et al. 2010;Courtney Mustaphi and Gajewski 2013).Some species are more susceptible to dissolution than others (Ryves et al. 2001); diatom records in which differential dissolution has taken place therefore contain non-representative assemblages and will yield poor estimates of overall diversity or production.Although potential causes of dissolution are well known (Battarbee et al. 2001;Ryves et al. 2006), the importance of these factors is not well documented in Arctic freshwater ecosystems.
This study presents a diatom stratigraphy from Lake RS29 on Somerset Island in the Canadian Arctic.Diatom autecological information and community composition, as well as ecosystem parameters such as diatom production and diversity are used to infer Holocene environmental changes in the lake.The diatom record is compared to several independent proxy climate records from the Canadian Arctic, including a pollen record from the same core (Gajewski 1995) and a reconstruction of Holocene climate variability of the region (Gajewski 2015a).The potential causes of low diatom production, especially relating to possible dissolution, are examined.Finally, the merit of autecological information and diatom community indices in providing reliable inferences that are consistent with other independent proxies is assessed.

D r a f t
6 Lake RS29 (unofficial name) is located on western Somerset Island in the Canadian Arctic Archipelago at 73.140°N, 95.278°W, 160 m.a.s.l.(Fig. 1).Lake RS29 has a surface area of ~12 ha and the depth at the coring location was 14.5 m.The catchment area is small and the lake is closely surrounded by hills that are 30 to 40 m above the level of the basin.Water chemistry data are not available from the site, however, three lakes from the general area of RS29 on Somerset Island had average pH of 7.4 (+/-0.2),specific conductance of 40 (+/-7) µS cm -1 and total Phosphorus and total Nitrogen concentrations near or below detection limit (sites CI02-04; Bouchard et al. 2004).The regional vegetation is a polar desert, located in the High Arctic bioclimatic zone, where plant cover is sparse (~covering 2 to 40% of the landscape) (CAVM

2003).
Mean annual temperature at the nearest weather station in Resolute (176 km to the north of RS29) is -15.9°C and this area receives approximately 170 mm of precipitation annually (Environment Canada 2002).The bedrock is composed of relatively unweathered gneiss and granite with discontinuous overlying till veneer (Dyke 1983).The substrate has a pH between 5.5 and 7.2 (CAVM 2003).Dated whalebone from northwestern Somerset indicate that deglaciation began ca.10,900 cal yr BP and the island was fully deglaciated ca.10,000 cal yr BP (Dyke 1983).

Methods
Lake RS29 was cored in the summer of 1991 from a 2.4 m thick ice surface using a 5-cm diameter Livingstone corer.The coring hole was cased, and a driver was used to push the corer into the sediment.The uppermost sediments, including the sediment interface, were collected in a clear plastic tube fitted with a piston.The top 20 cm of the core were extruded at 1-cm intervals D r a f t 7 while the remaining sediment cores were wrapped in cellophane and aluminum foil and placed into PVC tubing.The core was stored at 4°C.
The core had been previously used for a pollen study (Gajewski, 1995).A total of eight samples were sent for radiocarbon analysis; see Gajewski (1995) for additional information (Table 1, Fig. 2).The R program BACON (Blaauw and Christen 2011) was used to derive a new chronology for this study, and this included calibrating the ages to calendar years (cal yr BP).
For this study, the core was sampled at contiguous 1-cm intervals to estimate organic matter, carbonate, and biogenic silica content.Organic and carbonate content were estimated using loss-on-ignition (Dean 1974;Heiri et al. 2001).Samples of 0.5 cm 3 were placed into crucibles and dried at 105°C in order to remove moisture and were subsequently ignited at 550°C for 4 hours and 950°C for 2 hours to estimate organic (LOI 550 ) and inorganic (LOI 950 ) carbon content.Biogenic Silica (BSi) concentration was determined using the wet alkali digestion method (DeMaster 1979;Conley and Schelske 2001).Aliquots were withdrawn at 2, 3, 4, and 5 hours from a sodium carbonate (Na 2 CO 3 ) solution heated at 80°C.The molybdosilicic acid spectrophotometric method (Parsons et al. 1984) was used to determine BSi concentration of the aliquots and the weight percent BSi calculated by regression analysis.
A total of 62 samples were processed for identification and enumeration of diatoms.
Samples were processed at 1-cm intervals for the top 40 cm of the core, at 2-cm intervals for depths 40-60 cm and at 5-10 cm intervals thereafter.Sediment was processed using standard acid-digestion methods (Battarbee 1986).Standard dilutions of the initial diatom slurry were prepared so that concentrations could be calculated.Diatom solutions were dried onto 18x18 mm coverslips and mounted onto slides with Naphrax®.Valves were enumerated at 1000x magnification using a Nikon Eclipse 90i light microscope with DIC optics.A minimum of 600 D r a f t 8 valves were identified in samples with high valve concentrations.In samples below 75 cm depth in the core, where valve concentrations were very low, slides were prepared from the most concentrated dilution and diatoms were enumerated in at least 30 transects to reach a minimum count of 100 valves.Diatoms were identified using image collections kept in the laboratory, taxon lists for Arctic lakes (Joynt and Wolfe 2001;Bouchard et al. 2004;Finkelstein et al. 2014) as well as standard taxonomic references (Krammer and Lange-Bertalot 1986-91;Patrick and Reimer 1996;Fallu et al. 2000;Moser et al. 2004).All diatom nomenclature was updated to reflect up-to-date conventions, and was verified using Algaebase (Guiry and Guiry 2016).A list of taxa recorded and taxonomic authorities are given in Table 2.
As discussed below, several studies of Arctic sediment cores have reported zones with no diatoms or very low concentrations, as was found in RS29.The literature was searched for studies reporting diatom assemblages in lake sediment cores from the Canadian Arctic and Greenland; records consisting of more than ~3000 years were considered (Table 3).For the subset of these cores where diatom-free zones were reported by the authors, the depths and ages of the diatom-free zones were recorded.
The statistical language R (R Core Development Team 2006) was used for data analyses.
Species richness was calculated using rarefaction analysis (Tipper 1979), based on a total count of 612 and 104 individuals to represent samples with both high and low valve concentrations.
The Shannon index was calculated as a measure of diversity, although we acknowledge the limitations of this metric in sediment cores (Peros and Gajewski 2008a).Changes in the diatom community were summarized using detrended correspondence analysis (DCA).The vegan package for R (Oksanen et al. 2006) was used to calculate the community indices and for diatom ordination; C2 was used to plot stratigraphic diagrams (Juggins 2008).Fossil diatom data were D r a f t 9 used in conjunction with the diatom calibration set and pH model of Finkelstein et al (2014) to produce a Holocene pH reconstruction for the RS29 record; model details are given in Finkelstein et al (2014).

Core chronology
The RS29 core measured 150 cm; its chronology is based on a total of 8 radiocarbon dates (2 on mosses, 2 on macrofossil fragments and 4 bulk sediment dates; Table 1), as well as the interface assigned as modern (1991 AD) (Gajewski 1995).Although the age of the sample from 53-55 cm was unusually old, it is essentially ignored in the model fitting process; in terms of model parameters, the memory was set to 0.7 and priors at 50 (Fig. 2).The model used in this paper was developed using Bayesian age-depth modeling (Bacon: Blaauw and Christen 2011), which uses an iterative curve fitting process which is more robust to outliers than the polynomial fit used previously.The new age model differs from that of Gajewski (1995) mostly in late Holocene sediments where the Bayesian model estimates the age of samples as younger than the cubic polynomial used in Gajewski (1995).Sedimentation rates are highest in the middle portion of the record (~ 7800 -5800 cal yr BP).A basal age of 10,020 cal yr BP confirms that lake sediments began to accumulate immediately following deglaciation of Somerset Island (Dyke 1983).

Diatom biostratigraphy
Of the 110 diatom taxa identified in samples from Lake RS29, 62 species were considered common, occurring at abundance of over 1% in at least 3 samples (Table 2; Fig. 3).The diatom D r a f t 10 record was divided into three zones on the basis of diatom preservation and valve concentration (Fig. 4).Zone 1 (6200-10,200 cal yr BP) was characterized by very low diatom concentrations and few taxa.Slides of undiluted diatom slurry had very low concentrations and reaching counts of 100 valves was difficult in most samples from this period.Many valves appeared poorly preserved; for example, highly silicified central areas and raphe ends of Naviculoid and Pinnularioid diatoms were abundant relative to intact diatom valves.This evidence for dissolution indicates that the diatom assemblages through Zone 1 are biased to the better preserved types, and assemblages must be interpreted with caution.Of the taxa that are present in Zone 1, the assemblage is co-dominated by a mixture of epiphytic or epilithic diatoms (Eunotia praerupta, Neidium affine and Pinnularia nodosa), and Tabellaria flocculosa, documented in benthic habitats in Arctic lakes (Antoniades et al. 2008).The small, benthic, often alkalophilous Fragilarioids Staurosira venter, S. lapponica, Staurosirella pinnata and in particular, Stauroforma exiguiformis, were present in this zone as well; these taxa especially predominate in the assemblages of the oldest two samples of the zone (10,000 -10,200 cal yr BP).These two basal samples were characterized by higher valve counts and concentrations.Overall, these assemblages yielded circum-neutral to slightly acidic values for reconstructed lake water pH in Zone 1 (mean = 6.5; standard deviation (std) = 0.1; Fig. 4).
At the onset of Zone 2 (6200 cal yr BP), diatom preservation improved markedly as diatom valve concentration and species richness increased, reaching values typical for High Arctic lakes with good diatom preservation through the Holocene (Finkelstein andGajewski 2007, 2008).Tychoplanktonic taxa Aulacoseira alpigena, A. Frustulia rhomboides (Fig. 3).Diatom-inferred pH declined in Zone 2 to 6.3 -6.5 (mean 6.4; std 0.1), and diatom to Chrysophyte cyst ratios reached maximum values in the middle of the zone.
Zone 3 began approximately 1400 cal yr BP according to the age model, and is characterized by an abrupt decrease in Aulacoseira alpigena, increases in benthic taxa such as Brachysira brebissonii, several species of acidophilic Eunotia and Psammothidium, and in the planktonic diatoms Cyclotella rossii and Discostella pseudostelligera (Fig. 3).Diatom and Chrysophyte cyst concentrations increased at the onset of Zone 3, and diatom diversity measures also showed further increases.Reconstructed pH increased through Zone 3 (mean 6.5, std 0.1) (Fig. 4).
Biogenic silica (BSi) increased over the Holocene.In this core, diatom assemblage diversity and richness is significantly correlated with total concentration (p<.001), but not with pH or temperature.
Detrended Correspondence Analysis (DCA) was used to summarize the trends in diatom assemblages through the Holocene (Fig 4).DCA axis one is highly correlated with total diatom concentration (r=0.94,p<0.001), suggesting that a major source of variation in diatom assemblages is the differential preservation of valves through Zone 1.The second DCA axis reflects the high variability of some of the more abundant diatoms in Zone 2. The third and fourth axes again illustrate that variability of the diatom assemblage is associated with changes in the diatom concentrations.Diatom concentration is negatively correlated equally with pH and July temperature (-0.3; p=0.04), suggesting that at least through the zone of poor diatom preservation, reconstructed pH is biased.These results underline the importance of considering D r a f t 12 the impact of differential preservation on the results of diatom-inferred pH reconstructions (Ryves et al. 2009).

Comparisons with available pollen data
Transitions in the diatom assemblages, as illustrated by the DCA scores (Fig 4) occurred at approximately the same time as marked changes in pollen assemblages and concentrations (Gajewski 1995).Diatom Zone 1 (10,200 -6200 cal yr BP) corresponds to higher pollenreconstructed air temperatures, higher percentages of low Arctic pollen taxa, mainly Salix, and maximum values for pollen concentration (Fig. 4).However, this time period also corresponds to minimum abundances of siliceous microfossils, suggesting high terrestrial but low aquatic production, or dissolution of the diatom assemblages.At ~6000 cal yr BP when pollen concentrations decreased and high Arctic pollen taxa such as Oxyria, Ranunculaceae and Saxifraga increased, suggesting cooling, the concentration of diatom valves in the sediment record increased.There was a further decrease in pollen concentration at ~1500 cal yr BP concurrent with an increase in both diatom and cyst concentrations as well as BSi, demonstrating divergent responses in terms of production proxies (pollen and diatom concentrations) for watershed vegetation vs. aquatic communities.

Discussion
The sediments of Lake RS29 yielded a complex siliceous microfossil record.Analyses of diatom concentrations indicate two major transitions; according to the proposed age model, these take place at 6200 and 1400 cal yr BP.These transitions also coincide with changes in diatom assemblages and vegetation changes recorded in a previously published pollen diagram from the D r a f t 13 same core (Gajewski 1995).Other transitions in the diatom assemblages, such as at 8200 and ~5200 cal yr BP, coincide with major climate transitions in the Arctic (Gajewski, 2015a).The marked shifts in diatom communities recorded an interaction between taphonomy (diatom preservation), climate and ontogeny of the lake.

Early to mid-Holocene paleoenvironments
The basal two samples in this record (10,000 -10,200 cal yr BP) record higher diatom concentrations in an otherwise low-concentration zone.They are dominated by small, colonial Fragilarioid diatom taxa (Staurosira venter, Staurosirella pinnata and Stauroforma exiguiformis) which typically occupy benthic habitats and are generally associated with early post-glacial paleoenvironments where glacial run-off supplies some mineral nutrients to sustain diatom communities in recently established ultra-oligotrophic lakes.These taxa have been widely reported in those conditions in early post-glacial sediments of Arctic lakes for the Holocene (Smol 1983;Cremer et al. 2001;Finkelstein and Gajewski 2007;Rouillard et al. 2012) and for earlier interglacials (Wilson et al. 2012).At Lake RS29, this early post-glacial assemblage was quickly replaced by a more unusual assemblage dominated by Eunotia praerupta, Neidium affine, Pinnularia nodosa and Tabellaria flocculosa.In Arctic lakes, these taxa occupy benthic and periphytic habitats, and are associated with higher dissolved organic carbon (DOC) concentrations (Antoniades et al. 2005(Antoniades et al. , 2008;;Fallu et al. 2000).Taken as a whole, the diatom assemblages in Zone 1 suggest some supply of base cations from the watershed to the lake, and a Arctic (Bouchard et al. 2004;Antoniades et al. 2008) are an order of magnitude lower than the maximum abundances observed in Lake RS29.
Despite the low abundance and probable dissolution of diatoms in Zone 1, this time period (10,200-6200 cal yr BP) is associated with maximum Holocene temperatures for this region of the Canadian Arctic (Fig. 4).The maximum abundance of Salix in the pollen record from the RS29 core, as well higher pollen concentrations and influx (Gajewski 1995;2015b) suggest warmer conditions, and possibly a longer growing season prior to ~6000 cal yr BP.A diatom record from Russell Island adjacent to nearby Prince of Wales Island (Finkelstein and Gajewski 2008) was also interpreted as signaling warmer temperatures prior to 6500 cal yr BP.
The spatio-temporal pattern of postglacial climates in the Canadian Arctic and Greenland has been summarized in several recent studies (Kauffman et al. 2005;Briner et al. 2016;Gajewski 2015a).The general conclusion is that warmest temperatures occurred earlier in the Holocene, and this is based on multi-proxy records from dozens of sites and is consistent between proxies.Maximum temperatures were identified in the early to mid-Holocene prior to 5200 cal yr BP, with the timing depending on region and this is documented in lake sediment records (Bradley 1990;Gajewski and Atkinson 2003;Kaufman et al. 2004;Gajewski,  Thus, a variety of independent paleoclimate estimates suggest that diatom Zone 1 in the RS29 record corresponds to a period of maximal warmth, with a higher vegetation density and terrestrial production on the landscape (Gajewski 2015b).

Diatom-free zones in Arctic lake sediments
The presence of a zone of very low diatom concentrations and influx, at a time that independent paleoclimate records show was the warmest period has been identified in several sites from the Canadian Arctic and Greenland.Cores reporting Holocene diatom analyses were obtained from the literature (Table 3).Of the 49 sites located (Table 3), 21 reported diatom-free zones (Fig. 5).
Of course, this is not a complete dataset as many sites with no diatoms would simply not be reported, since the site would be abandoned for study (Smith 2002).Lakes where the author had reported diatom-free zones fall into two broad groups.
The first group includes records where diatoms are completely lacking or only present in the uppermost samples (e.g., DV09, WB02, Sawtooth, Ward Hunt, RS36; see  Gajewski 2008;Rühland et al. 2015) and the presence of diatoms in the uppermost sediments is due to recent climate warming.However, our results from RS29 show that at least at this site, diatoms were missing during both warm and cold periods prior to the 20 th century, and indeed, they were missing in the sediments deposited earlier in the Holocene, when conditions were warmer than the 20 th century (Gajewski 2015a).A monotonically decreasing concentration of diatoms with depth in the sediment suggests post-depositional dissolution (e.g., Florian et al.

2015)
, as does the presence of diatoms only in the uppermost few cm.However, in some cases, authors report lack of evidence of dissolution in the diatom valves (e.g., Perren et al. 2003;Antoniades et al. 2007), although Ryves et al (2013) suggest that dissolution may not always be noticed.Dissolution may be inferred by the presence of only the more heavily silicified taxa, morphological criteria (Ryves et al. 2009), or concentration data (Ryves et al. 2002;Florian et al. 2015).For example, at Lake DV09, diatoms were only present in sediments of the past 150 years, but concentrations decreased from the surface downwards in the uppermost 15 cm (Gajewski et al. 1997).The presence of varves for at least the past 1000 years at Lake DV09 demonstrates that the lake was ice-free seasonally through this period, making it unlikely that diatoms were completely absent.Thus, any diatoms deposited on the sediment surface were Several factors are known to enhance dissolution of diatom valves in the water column or in surface sediments, including elevated lake water pH, temperature and silica depletion (Flower 1993;Ryves et al. 2001), elevated conductivity, salinity or alkalinity (Ryves et al 2002;2006;2013), grazing and bioturbation (Battarbee et al. 2001;Gibson et al. 2000), bacterial activity D r a f t 17 (Bidle and Farooq 1999;Ryves et al. 2006), fragmentation (Ryves et al. 2006) or reduced sedimentation rate (Ryves et al. 2013).Further, post-depositional changes in pore water chemistry creating micro-environments of elevated pH can promote dissolution of diatom valves down-core (Florian et al. 2015).
The modern pH and specific conductance were reported in several of the studies (Table 3).The mean pH of sites with a diatom-free layer at some point in time (i.e., those shown in Figure 5) was 7.7 (standard deviation 0.8), whereas it was 6.7 (std 0.7) at sites without a diatomfree zone reported.Average specific conductance was 144 µS cm -1 (std 205) for sites with missing diatoms and 74 µS cm -1 (std 141) for sites not reporting such levels.The pH and specific conductance are higher in the sites with diatom-free zones, although these values should be too low to cause dissolution (Ryves et al. 2002).However Ryves et al. (2002) also found preservation decreased in fresh waters as conductivity increased and Ryves et al (2006) report "substantial dissolution" at the pH values found in the lakes in our study.Fragmentation of diatoms, presumably due to more turbulence in the lake, may increase destruction of diatoms, even in lakes with low alkalinity (Ryves et al. 2006).Sites on Baffin Island, which is underlain by Precambrian Shield, rarely reported diatom-free layers, whereas sites in the northern, western and central Arctic, which are underlain by carbonates or other sedimentary deposits, have a higher tendency to contain these zones.Of course, the values of temperature, pH and alkalinity have changed over the course of the Holocene.Given these values of the water chemistry, however, a simple biogeochemical explanation for dissolution cannot be sufficient to explain all of these records.Due to some particular characteristic of the lake, diatoms are not preserved in these sites.

D r a f t 18
A second group includes sites where diatoms are lacking for a portion of the core (Fig 5) and these tend to be found in the older sections.The RS29 record is an example; essentially no diatoms were found in the sediments dated to greater 6,200 cal yr BP, low values from 6,200 -1,400 cal yr BP, and high values in the past 1,400 years.The inverse correlation between aquatic and terrestrial production, where pollen concentrations were high when temperatures are warm, as expected, but diatom concentrations were not, suggests the possibility that the diatoms were being dissolved at some point between the living diatom assemblage and the burial at depth in the sediment.Similarly, at site BC01 from Melville Island, diatoms only appeared in the sediments at ~5000 cal yr BP, at precisely the time where a pollen-based July temperature reconstruction from the same core began to show a decrease from high values reconstructed for the previous 8000 years, and when pollen concentrations decreased (Peros et al. 2010).At site KR02, diatoms were missing in the sediments between 8500-8200, 7600-7000 and 5600-5000 cal yr BP.In all three periods, both a pollen-based and chironomid-based July temperature reconstruction were above the long-term mean, and reconstructed pH (using the method of Fortin and Gajewski 2009), was above the mean (Fortin and Gajewski 2010).Smith (2002) found very low concentrations of diatoms at all of his sites in the older sediments.In several sites in southern Greenland and the eastern Arctic, zones with low concentrations of diatoms were found in the older sections as well, with diatom concentrations subsequently increasing at various times (Adams and Finkelstein 2010;Perren et al. 2112a;Florian et al. 2015;Law et al. 2015).
Most of these sites had well-preserved diatoms in sediments of the past 3000-4000 years, which corresponds, across the entire North America Arctic except for South Greenland, to the Neoglacial cooling (Gajewski 2015a).Diatoms were absent during the time that coincides broadly with the warmest part the Holocene.However, Law et al (2015) attributed increased D r a f t 19 dissolution of diatoms at two sites in Greenland to increased alkalinity in the early stages of the lake ontogeny.They suggested that time since deglaciation determined the alkalinity levels of the lake more than climate.In the western and central Arctic, the early stages of the lake history coincide with the warmest period, so untangling these two potential causes may be difficult.
Potential factors causing the lack of diatoms in the sediments, such as pH or alkalinity, nutrient or Si limitation in the water, temperature and ice cover, may vary over the course of the Holocene, but the changes are relatively small, so it would seem surprising that the disappearance of diatoms was caused by dissolution due to temperature or alkalinity changes, for example.In several cases where multi-proxy records are available, the continuous deposition of other proxies, during a time when diatoms were not found in the sediments, demonstrates that a continuous ice cover cannot be the explanation for the absence of diatoms (Lake KR02: pollen In these oligotrophic, dilute and low alkalinity lakes, it is possible that when the climate warmed, the diatom production increased over the course of the lengthened growing season.As the diatoms died and were removed from the water column or surface sediments, the Si was recycled, leading to few remains in the sediment.Extensive dissolution and Si recycling has been observed in more productive temperate systems (e.g., Conley et al. 1988;Nriagu 1978;Parker et al. 1977a, b;Ryves et al. 2013) and in Alaska (Cornwall and Banahan 1992).The increased terrestrial vegetation production, along with the continual presence of permafrost may have D r a f t 20 reduced input of Si to the lake.Changes in precipitation over the Holocene, which could have affected runoff are not well known for the Arctic.If this is the case, the inverse relation between terrestrial production (ie the pollen concentration) and diatom concentration in the sediments may be the result of a more productive lake, with high diatom production, but that the silica was recycled before the sediment was buried.
While further study is needed to determine the relationship between absence of diatoms in the sediments and warmer temperatures in records from Arctic lakes, our literature survey and results from Lake RS29 suggests the possibility of warm temperatures exacerbating diatom dissolution, given the right chemical environment and lake production.Increases in diatom production, as inferred by increases in biogenic silica, in diatom valve concentrations or in fluxes of photosynthetic pigments, in Arctic lake sediment records are often interpreted as indicative of warmer temperatures (e.g., Cremer et al. 2001;LeBlanc et al. 2004;Michelutti et al 2005;Rühland et al. 2015 and references therein).Other studies indicate that in some lakes, diatom production and community composition is more closely related to nutrient availability and light penetration (Baier et al. 2004;Malik and Saros 2016) or lake catchment hydrochemistry and only indirectly controlled by climate (Anderson 2000, Anderson et al. 2008, 2012;Law et al. 2015).
Our results suggest that this relationship may be further modified by dissolution in warm periods, and out of phase relations would appear between independent temperature proxies and the diatom production proxies (e.g., Michelutti et al. 2007;Wagner et al. 2008).

Mid-to late Holocene paleoenvironments
Biogenic silica continued to increase and diatom valves appeared much better preserved at the onset of Zone 2 (6200 cal yr BP), as major taxonomic shifts took place.More alkalophilic taxa D r a f t 21 that are important in Zone 1, such as Neidium affine, Staurosirella pinnata and Tabellaria flocculosa decreased, as a variety of acidophiles in the genera Aulacoseira, Eunotia and Pinnularia increase.Aulacoseira alpigena and Pinnularia microstauron were most abundant in this zone (Fig. 3).The high abundances of A. alpigena indicate an adequate supply of nutrients, in particular silica, and open-water conditions with high enough energy to maintain the position of this heavier, more silicified (tycho)planktonic diatom in the upper part of the water column (Wolfe and Härtling 1996;Miller et al. 1999).Further, the assemblage suggests the development of a mossy littoral zone, providing habitat for epiphytic Pinnularia populations.Diatom-inferred pH (Fig. 4) indicates increasing acidity between 5500 and 2000 cal yr BP.A post-glacial succession from small benthic Fragilarioids, recorded in the initial two samples of the record, to assemblages dominated by Aulacoseira spp.and other more acidophilic diatoms has been widely recorded in Arctic lakes (Smol 1983;Miller et al. 1999;Wilson et al. 2012).
The increasing acidity evidenced by these changes may be driven by a variety of processes.Natural acidification has been documented for many temperate lakes in forested catchments due to gradual leaching of base cations through weathering processes and export of humic acids with progressive pedogenesis and plant succession.These processes have been invoked as an important mechanism in southern Greenland (Anderson et al. 2008), but they are less applicable to Lake RS29, located in the High Arctic polar desert.The short snow-free season, thin active layer, cold temperatures through the summer, prevailing aridity, and low plant biomass means there is little pedogenesis and slow rates of chemical weathering (Gajewski 2015b).In the Arctic, lack of extensive vegetation or soil development suggests climate may be a more important factor than catchment processes in determining ecological communities (Wilson et al. 2012).Despite the potential for acidic weathering products in areas of the Arctic underlain D r a f t 22 by granitic rocks, it is not well established that these processes promote acidification of fresh waters in the Arctic.The presence of moss remains in the RS29 core (Table 1) indicates the establishment of a shallow littoral zone suited to mosses and other aquatic plants.Other pollen taxa increased at that time, including Artemisia, Polypodiaceae, Rosaceae and Ranunculaceae (Gajewski 1995).While the taxonomic resolution of the pollen record does not allow species to be identified, the flora of Somerset Island contains a variety of species within these groups that are found in aquatic or littoral habitats (Porsild 1964).Bryophyte-rich wetland plant communities typically yield acidic runoff, potentially contributing to the development of the acidophilic diatom community recorded in Zone 2.
Alternate explanations for acidification in poorly-buffered Arctic lakes relate to climate, which determines the duration of ice cover, and in turn regulates dissolved inorganic carbon speciation through pCO 2 , and hence lake water pH (Koinig et al. 1998).The inferred acidification of Lake RS29 coincided temporally with cooling climates as inferred by pollen and a variety of other proxies across this region (Gajewski 1995(Gajewski , 2015a)), although the correlation of pH and July temperature is low (0.15, p=0.2).Thus, the declining pH through diatom Zone 2 in the RS29 record may also be an effect of climatic cooling and prolonged ice cover, as has been suggested for other Arctic lakes with dilute waters in the middle and late Holocene (Wolfe 2002;Michelutti et al. 2007).
The uppermost diatom Zone 3 was characterized by an abrupt decrease in Aulacoseira seasons, facilitating the increase in relative abundance of diatoms in the Cyclotella sensu lato complex.The decline in Aulacoseira spp.may relate to enhanced thermal stratification, and less of the turbulence required to keep these heavy taxa afloat in the upper part of the water column (Rühland et al. 2008).This suite of changes is often associated with the onset of anthropogenic climate warming in the late 19 th or early 20 th century (Douglas et al. 1994;Smol et al. 2005;Rühland et al. 2008), although the situation may be more complex (Saros and Anderson 2015).
According to the age-depth model developed for Lake RS29 these shifts are occurring considerably earlier, as has been found elsewhere (e.g., Finkelstein and Gajewski 2007;Perren et al. 2009Perren et al. , 2012a;;Saros and Anderson 2015).Given the uncertainties of radiocarbon (Gajewski et al. 1995) and 210 Pb (e.g., Hadley et al. 2010) chronologies in Arctic lakes, the timing of these changes should be interpreted with caution.

Summary
At Lake RS29, the major diatom assemblages in the early Holocene included mostly benthic taxa reflecting circumneutral to slightly acidic conditions, more mineral nutrients and a warmer climate but with significant dissolution taking place.A shift to more acidophilic taxa, tychoplanktonics and large benthics began around 6200 cal yr BP, when diatom concentrations increased significantly as dissolution rates declined.A further shift took place around 1400 cal yr BP; at this time, the acidophilic taxon Aulacoseira alpigena declined abruptly, as diverse benthic and periphytic taxa, as well as planktonic diatoms in the Cyclotella sensu lato group increased, indicating a return to somewhat less acidic conditions, further development of littoral macrophyte communities, and increased duration of ice-free seasons.This record shows the importance of both climate and local habitat controls in explaining the changes in Arctic diatom D r a f t 24 records, as well as the need to consider taphonomic processes which affect diatom abundance.
The availability of pollen data from the same site allowed for an independent paleoclimate record to better evaluate the relative importance of climatic and watershed-driven controls on diatom assemblages.Figure 3. Relative abundances of common diatoms identified in Lake RS29.Species authorities are given in Table 2. Grey lines in lowermost panels are 5x exaggerations.Figure 5: Sites where diatom-free zones or dissolution were reported.See Table 3 3 for site information.Grey bar means diatoms present and white means absent; if low concentrations were reported the bar is half-filled.Also shown are the July temperatures for the regions from Gajewski (2015a).

1
Interactions between climate and landscape drive Holocene ecological change in a High Arctic lake on Somerset Island, Nunavut, Canada Tara M Paull 1 , Sarah A. Finkelstein 2 , Konrad Gajewski 1*

(
nivalis and A. perglabra increased in abundance in Zone 2, along with epiphytic Pinnularia microstauron, a suite of benthic Naviculoid (e.g., Humidophila schmassmannii, Navicula digitulus) and Achnanthoid Achnanthidium kriegeri, Nupela impexiformis, Psammothidium levanderi) taxa, and acidophilic warm and wet enough climate to support the development of the littoral zone vegetation, providing periphytic diatom habitats and releasing DOC to the lake.Further, given the warmer climate at this time (see below), the thicker active layer may have increased particulate and DOC watershed to lakes, influencing diatom assemblages (Fritz and Anderson 2013); high DOC concentrations have been inferred spectroscopically from other High Arctic lakes during early Holocene thermal maxima(Rouillard et al. 2012).Indications of significant dissolution of diatom valves further characterized diatom Zone 1.These indications include low concentrations and valve counts, the preservation of more heavily silicified central areas without surrounding frustules, and the assemblage composition itself suggests important bias.For example, maximum abundance of Neidium species in ~150 lakes and ponds in the Canadian possibly being dissolved over time in the carbonate-rich sediment (Courtney Mustaphi and Gajewski 2013; Outridge et al. subm).
alpigena, and an increase in numerous epiphytic (e.g., Eunotia exigua and Psammothidium marginulatum), and plankontic taxa (Cylotella rossii and Discostella pseodostelligera).Overall, diatom diversity continued to increase through Zone 3. The assemblages in this zone indicate a well-developed littoral zone with diverse periphytic diatom habitats, as well as longer open water

Figure 1 .
Figure 1.Location of Lake RS29 and other records mentioned in text.

Figure 2 .
Figure 2. Age depth curve of RS29.See text for details.

Figure 4 .
Figure 4. Diatom community indices including diatom and chrysophyte cyst concentrations and rarefaction diversity measures, pH reconstructed from the diatom assemblages, sample scores from detrended correspondence analysis, sediment parameters and biogenic silica estimates, pollen concentrations, and July mean temperatures estimated from pollen assemblages for the central Arctic.See text for details.

Figure 1 .Figure 2 .Figure 3 .
Figure 1.Location of Lake RS29 and other records mentioned in text.

Figure 5 :
Figure5: Sites where diatom-free zones or dissolution were reported.See Table3for site information.Grey bar means diatoms present and white means absent; if low concentrations were reported the bar is half-filled.Also shown are the July temperatures for the regions from Gajewski (2015a).
Table 3 for citations).Several authors have proposed that the absence of diatoms from these kinds of records

Table 1 .
Gajewski (2015a)tion.Grey bar means diatoms present and white means absent; if low concentrations were reported the bar is half-filled.Also shown are the July temperatures for the regions fromGajewski (2015a).Conventional radiocarbon ages and calibrated age BP (where year zero is AD1950) from lake RS29.

Table 2 .
Maximum abundances and number of occurrences of the common diatom taxa identified from Lake RS29; taxa are listed in order of decreasing maximum abundance. https://mc06.manuscriptcentral.com/asopen-pubs

Table 3 .
Gajewski (2015a)vestigate diatom-free zones in the Canadian Arctic and Greenland.All lakes in the table were considered, but only those where the authors report diatom-free zones or dissolution are included in Fig.5.The climate region is fromGajewski (2015a).The proxy column indicates which were available for each site: % = diatom percentages, C = diatom concentrations or influx, BSi = biogenic silica, P = pigments.Records had to be greater than ~3000 years in length to be included.Sp cond is specific conductance. https://mc06.manuscriptcentral.com/asopen-pubs

Table 2 .
Grey lines in lowermost panels are 5x exaggerations.Figure 4. Diatom community indices including diatom and chrysophyte cyst concentrations and rarefaction diversity measures, pH reconstructed from the diatom assemblages, sample scores from detrended correspondence analysis, sediment parameters and biogenic silica estimates, pollen concentrations, and July mean temperatures estimated from pollen assemblages for the central Arctic.See text for details.