Division of the sites

Our limnological monitoring program at Cape Herschel has generated some of the longest (if not the longest) semi-continuous datasets from lakes and ponds in the High Arctic. These data are highly suitable to answer the questions we posed. Given that long-term lake water thermistor data and site-specific air temperatures were not available, we categorized the study sites based largely on long-term observational records from multiple mid-summer sampling campaigns and on our observations spanning nearly three decades. The microclimate differences we describe were obvious in every field season and were further supplemented by similar ice-cover observations collected by Dr. Wes Blake, Jr, and his colleagues, dating back to the mid-1970s, as described and documented below.

Gajewski is mistaken when he interpreted the dates listed in Table 1 of our paper as the only times we visited or observed ice cover on the ponds. Rather, what was presented in our article (as Table 1, [1]) are the notes taken only from days that detailed water chemistry sampling occurred (up to 58 water chemistry variables from each site as index values for these ponds and lakes). These were usually taken once a season, although more often for some lakes. We reiterate that the sites were visited repeatedly over each field season from 1983–2011 (usually lasting two weeks, and for several months in 1986, spanning from thaw to freeze). Furthermore, the lower Elison Pass area of Cape Herschel is visible from our base camp whereas the outlet from Moraine Pond was our drinking water supply, and so we visited that pond nearly every day during the field season. Snow/ice patterns were noted for all the study ponds during our many other excursions to various points on Cape Herschel and during regular fly-overs of many of the sites (including nearby Pim Island).

Peer-reviewed research has shown the power of observational evidence from field surveys (in combination with other lines of evidence) to document, for example, changes to ice cover [2]. The personal experiences of two of our co-authors should not be devalued. John Smol has visited Cape Herschel (and nearby Pim Island almost as often) regularly, nearly every 3–5 years for the past 28 years together with Marianne Douglas, who has likewise visited these sites for over two decades. Prior to that, Dr. Weston Blake Jr., who was emeritus scientist at the Geological Survey of Canada, performed decades of repeated observations and sampling of these lakes. Although not discussed in our paper, it was Blake (whose written comments on ice melt are discussed further below) who first pointed out that the ice covers between the sites were markedly different and directed the research team in 1983 to visit Col Pond first as, at that point in the summer, it would be the only pond that was ice-free. Our division of the sites at Cape Herschel and Pim Island is based on ground-truthing from decades of observations by the paper’s co-authors, and other members of the Cape Herschel Project.

In 1986, the sampling season at Cape Herschel spanned the entire growing season (June-August; thaw to start of re-freeze) such that for four of the 10 sites (Col Pond, Elison Lake, Moraine Pond, and Plateau Pond 2) we documented the timing of ice-off and -on. Cape Herschel and Pim Island were also visited early in the season in 1983 (providing a context for early season ice cover at West Lake) and Proteus Lake and Paradise Pond were added to the sampling regime in 1987, when the lakes were sampled in late summer. Consequently, for seven of our ten sites we have convincing observational data of the early and late season conditions with relative dates of ice-on and, to a slightly lesser extent, ice-off.

High Lake was not directly sampled until 2011 (but we, and Blake before us in the 1970s, had repeat observations of ice and snow cover from flyovers in helicopters) as it never had open water in all of the years when Pim Island was sampled. As such, High Lake is inferred to be one of our coldest sites, with one of the shortest ice-free seasons. Given the start of winter is mid/late August [3], it is reasonable to infer that the ice-free season at this lake is relatively short, as the ice does not start melting by mid-July while all the other nearby sites are partially or completely ice-free. By contrast, Elison Lake currently has an ice-free period of ~3 months. Sverdrup Pass, where the other two “oasis” ponds are located, was only visited by our team twice in mid-summer, in the 1980s and in 2011; however, these ponds are located in a warmer climatic sub-region (see below, studied by colleagues), and consequently experience a longer ice-free season.

Our detailed observations are supplemented by those of previous researchers who have worked extensively on Cape Herschel and nearby Pim Island, such as the late Dr. Weston Blake Jr. (1930–2021). In a written communication asking him to comment on ice cover changes in the area, he noted:

“During my field seasons at Cape Herschel and Pim Island, which date back to the 1970s, the Plateau area has always been colder and more ice covered. The Col always opened first. The lower Cape area was always warmer and less ice covered. The top of Pim Island was of course always colder and more ice covered given its location and elevation”.

Signed by Weston Blake Jr. on Sept 19, 2019.

The “oasis” sites

Gajewski criticizes the characterization of “oasis” sites in Sverdrup Pass (which we only had the opportunity to visit and sample twice, in 1984 and once in 2011), which he argues is not warmer than Cape Herschel. However, many other researchers have worked in Sverdrup Pass (e.g., [4]), specifically because it has consistently been recognized as an Arctic oasis with anomalously warm temperatures (e.g., [5]).

Sverdrup Pass is in a different ecozone than Cape Herschel and Pim Island (Northern Arctic and Arctic Cordillera respectively; [6]). Cape Herschel and Pim Island are situated in the cold, sea-ice-filled waters in Smith Sound of the Nares Strait and experience a more ice-dominated climate than the continental Eureka/Nansen Sound and its adjacent lowlands (in which Sverdrup Pass is included). The 1950s-1970s provided some of the greatest detailed climate data for the region, while currently there are only three meteorological stations for Ellesmere Island. Using weather station networks and measurements spanning the years from 1953 to 1972, the Canadian Arctic Islands are subdivided into different climate regions based upon the major climate controls (such as cyclonic activity, sea-ice water regime, broad-scale physiographic features, and net radiation) [3]. These climate regions are further sectioned based on major local variations in local topography, aviation weather, maritime influence, temperature, precipitation, snow cover, and wind. For example, compared to Nansen Sound and its adjacent lowlands (including Sverdrup Pass), the Nares Strait region (including Cape Herschel and Pim Island) is characterized as having greater cloud cover, less July irradiance, later stream break-up, cooler July temperatures, and an earlier start to winter (by 5–10 days) and a later end to winter (by 5–10 days) [3]. These seemingly small differences (up to 20 days/year) can appreciably affect the length of an already short “summer” season (i.e., when the melt season is 90 days or less, an extra 10–20 days can make a substantial difference).

Gajewski presents temperature data from field camp reports at Sverdrup Pass and Cape Herschel over the course of the years 1974–1993. He admits, “there is little overlap between dates where Sverdrup Pass camps reported and when Herschel did.” The variability in climate among years is simply too great for making direct comparisons, overlapping years are required. Despite these obvious shortcomings, the data Gajewski presents run counter to his own arguments. In fact, Gajewski states that “June tends to be warmer in Sverdrup Pass than on in [sic] the Herschel area” and that “Below zero temperatures seem to return in late August at approximately the same time” (and seemingly slightly later at Sverdrup Pass), thereby inadvertently confirming our classification scheme. Importantly, it is the early temperatures that are critical for ice melt. After presenting these data, Gajewski concludes that “there are no data in May and early June to compare the end of the ice-cover season”. The lack of data spanning this critical period from ice-on to ice-off is exactly why we had to rely on our three decades of observations. In summary, Gajewski presents temperature data, that we did not use, only to argue that these data cannot be used to prove our classification scheme. Moreover, even if accepted, his analysis suggests that the critical early melt period (May and early June) may be warmer at Sverdrup Pass, as it is already warmer by mid-June.

Additionally, Gajewski asserts that Sverdrup Pass is vegetated because of substrate type and glacial meltwater, and that the greater vegetation in Sverdrup Pass is not a reflection of the warmer climate. However, in a study of vegetation and climate of the Queen Elizabeth Islands the Sverdrup Pass region was consigned to a completely unique, warmer bioclimatic zone that supported a denser and more diverse vascular plant community, a division that could not be explained simply by accounting for substrate type or moisture regime [5]. It was observed that “The mountains of Axel Heiberg and Ellesmere islands create a barrier that effectively shelters an intermontane region from both the central Arctic Ocean climate and travelling cyclonic systems. In this large intermontane zone regional minimums of cloud cover and maximums of temperatures and melt season duration are found.” [5]. Using meteorological data from permanent weather stations and non-standard weather data (seasonal scientific and oil-industry camps), the Eureka Sound intermontane area, which extends into the Sverdrup Pass region, was identified as having a melt period that averages “15–20 days longer than in coastal regions adjacent to the polar pack-ice”, and as unlikely (as compared to the coastal regions) to “experience mean daily temperatures below freezing once the melt season is established” [5]. Also, this intermontane region, due to the rain-shadow effect, experiences the lowest annual precipitation of all regions of the Queen Elizabeth Island [5].

In summary, while long-term comparative historical temperature data do not exist for these remote regions (which is why we conduct paleoecological analyses), there are repeat observations and visits of our study sites by our co-authors and other Arctic researchers spanning decades that support our microclimate divisions. It is reasonable to assume that the geological factors that partially differentiate the microclimatic differences among our sites (e.g., altitudinal differences, proximity to the coast) at present day were similar to conditions ~200 years ago (our period of interest) as the region has not experienced any substantial tectonic movement, changes in air-mass movement, or isostatic uplift over such a geologically short timeframe. This is not the case when considering Holocene scales (see section 3).

Nutrients

Gawjeski speculates, without providing any evidence, that “another interpretation of the “oasis” sites is that they are highly productive due to continual water supply from nearby glaciers and combined with bedrock, this provides more available nutrients”. Gajewski’s speculation, however, is not true. We can confirm that the oasis sites are not connected to any continuous sources of water.

Additionally, Gajewski regularly refers to the oasis sites as “highly productive”. The Sverdrup ponds are far from being highly productive. Lake and pond water chlorophyll a analyses are the standard measures for estimating lake or pondwater biomass (standing crop) or productivity. Both Sverdrup ponds are so unproductive that they are below the detection limit of analyses for chlorophyll a (corrected for phaeophytin) at < 0.1 μg/L and uncorrected at 0.5 and 0.6 μg/L for Sverdrup ponds 5 and 8, respectively (i.e., amongst the lowest chlorophyll a values one can find in North America [7]). Therefore, Gajewski’s subsequent comments regarding the supposedly “highly productive” nature of Sverdrup ponds, are incorrect. Yes, overall annual production (the operative word being “annual”) would be relatively higher in the warmer Arctic oasis like Sverdrup Pass due to the longer growing season and reduced ice cover (consistent with our data, findings, and interpretations), but water-column production is low at these sites, opposite to Gajewski’s conjectures.

While it is true that the “oasis” sites generally have higher nutrient concentrations compared to most of the other ponds in this study (as warmer sites would be expected to, as nutrient release from the catchment is accelerated), there is one important exception—the “cool” site, Moraine Pond. Moraine Pond has very similar total phosphorus (TP) concentrations to the “oasis” sites (13.6 μg/L vs 15.8 μg/L and 13.9 μg/L for SV5 and SV8 respectively) due to its relatively large catchment for its size. However, the Moraine Pond sediment record still registered a climate-related diatom assemblage shift with decreasing relative abundances of commonly epipelic diatoms and increasing proportions of epiphytic taxa. In contrast, Sverdrup Pass’s SV5 and SV8 recorded different assemblage shifts, namely changes among epiphytic taxa. These observations are not consistent with Gajewski’s assertions that nutrients are the driving factor at these sites.

In addition to our study sites, there are other, high-nutrient lakes and ponds across the Arctic (although they are moderately rare) that have shown the same transition from largely epipelic and epilithic diatom assemblages to higher diversity assemblages with increasing proportions of complex growth forms including epiphytic taxa. For example, Self Pond (Ellesmere Island) in the High Arctic, shows an unambiguous transition from epipelic to epiphytic taxa and has TP concentrations that are very similar to the “oasis” ponds of 12.1 μg/L [8], or pond MV-AT on Melville Island, showing the same type of diatom compositional transition and again with a TP similar to the “oasis” ponds (12.7 μg/L) [9]. What is consistent amongst these diverse ponds is recent warming.

In contrast, though, there are other high-nutrient sites in the Arctic that have quite short ice-free seasons. Meretta Lake, on Cornwallis Island, for example, has had a relatively short ice-free season. When limnologists studied Meretta Lake during the IBP Ecosystem Study, the lake became ice-free for only a few weeks in 1969 and 1970, and never lost its central ice cover in 1971 (see Fig 4 in [10]). However, raw sewage was dumped into this lake for nearly 50 years, starting in the 1940s, leading to high nutrient influxes. Despite this, there was only a very muted response in the diatom record of the lake sediment core [11]. Meretta Lake’s assemblages are dominated by small fragilarioid taxa [12], epipelic generalist and opportunistic taxa that tend to thrive in a diversity of conditions where other diatom species struggle. Yet, despite an influx of nutrients to the site, the sedimentary diatom assemblage did not undergo a marked shift towards a greater diversity of epiphytic forms.

Another example is Annak Lake [13–17], a lake also used for sewage discharge since the 1960s by the community at Sanikiluaq situated on the Belcher Islands (Nunavut). It has a well dated ²¹⁰Pb profile, that is further constrained by stable geochemical metals and other proxies linked to human sewage in the core profile, anchoring and corroborating the ²¹⁰Pb dates (as described in the cited papers). Modelled TP concentrations show hypereutrophic conditions (TP > 600 μg/L) and elevated nutrient levels in Annak Lake, yet the diatom record shows that, while the lake remained ice-covered largely in the 1960s, the diatom assemblage was overwhelmingly dominated by small fragilarioid taxa. It was only later (during the two decades following the sewage inputs), but coincident with warming and declining ice covers, that the assemblage changed, reflecting the establishment of new aquatic habitats and longer growing seasons.

Nutrients, due to the greening of the Arctic and increased soil development and erosion, can play a role in some of the diatom assemblage changes we recorded (and we address this in our paper–see page 17, bottom paragraph [1])–and this is most likely why we saw an increase in Nitzschia species (a nutrient indicator in Arctic lakes and ponds) in many of our sites. But this is different than the shift in diatom life strategy from commonly epipelic/epilithic taxa to commonly epiphytic species and more complex growth forms requiring longer open-water periods. This particular compositional change was detailed in our publication as a shift from “condition 1” to “condition 2” and is characterized by relative decreases in commonly epilithic and epipelic species to increasing relative abundances of commonly epiphytic species (a major habitat and life strategy shift). We went into detail on how that mechanism might work (i.e., how changes in the duration of ice cover might have an effect on the establishment and growth of aquatic mosses and how those mosses support a diverse array of diatoms) throughout our paper, but especially in the section “Mechanisms of change”. The Gajewski comment fails to consider the biology of the diatom taxa. Rather, the taxonomic information of our diatom records was reduced to principal component axes, without understanding the autecology of the species considered. For example, our cores from West Lake do register diatom compositional changes (which we discuss in our publication, and which is addressed further below), but these changes do not reflect the establishment of novel epiphytic habitats, which is the ecological change we are tracking. Such an interpretation requires an understanding diatom autecology. A shift among diatom taxa with similar affinities/tolerances to prolonged ice conditions does indeed signal a limnological change, but not a change in ice cover resulting in the opening of novel habitats. An understanding of the autecology of the diatom taxa is missing from Gajewski’s analysis. This understanding is fundamental in interpreting the diatom data.

Additionally, subfossil chironomids, an independent bioindicator group that respond to temperature and habitat changes, but that do not respond directly to nutrients [18–20], were analyzed from our “warm” sites [21], using the same sediments as a previous study of diatoms on the “warm” Cape Herschel ponds [22]. The chironomids showed striking changes indicative of increased diversity and increases in macrophyte-associated taxa, matching the diatom data and the inference of aquatic moss establishment in these ponds. Two independent lines of evidence based on organisms from different phyla support the same conclusion.

Depth

Gajewski notes that “the four groups of lakes differ in depth and nutrient concentration which would affect diatom abundance and diversity”. Ideally, we would have liked to have sampled the shallow, elevated, exposed ponds on Pim Island as our “cold” sites. Unfortunately, due to more extensive ice conditions on Pim Island, very few sites exist, and none of these sites possessed depositional areas that would allow us to obtain suitable sediment cores because of their low productivity, small size, shallow nature, and extremely small catchments (much smaller than any of the Cape Herschel ponds). However, our study sites are still comparable. In particular, High “Lake” (a “cold” site), although originally called a “lake” by Blake in his early field notes, is ecologically and physically a “pond” (i.e., defined here as freezing to the bottom in winter), and comparable in size and depth to Elison Lake (a “warm” site), which is also ecologically a pond. High Lake is only 4 m deep at its centre and Blake attempted to obtain a sediment core through the ice in June 1981. Based on his experiences with the site, Blake suggested that we should consider it a “cold” pond when sampling, given the ice thicknesses that can develop on nearby lakes. For example, a nearby lake on Ellesmere Island had 5.45 m of ice in June of 1986 and in August 1987, still had 3–4 m of ice [23]. Field notes from June 1981 documented ice thicknesses or both West Lake and 260 m Lake on Pim Island of 2.4 m and 2.6 m, respectively. Ice thicknesses from Ward Hunt Lake off the northern coast of Ellesmere Island recorded ice thicknesses that “ranged from 4.0–4.1 m when measured in May 1998 and August 2003, with a moat of open water that develops over the littoral zone of the lake in late summer” [24]. Although we are not able to sample in winter, it is highly likely that similar ice thicknesses can develop on Pim Island, and that likely all, or almost all, of High Lake freezes to the bottom and therefore can ecologically be classified as a “pond”, despite its informal name.

The other two “cold” sites in our study are ecologically classified as lakes and therefore tend to form several meters of ice over the winter (see above) but do not freeze completely to the bottom (although the nearshore littoral would do so). This thick layer of ice on lakes is retained even longer through the summer season than nearby ponds. The shift that we are tracking from a depauperate assemblage dominated by a few epipelic taxa to an assemblage with a greater diversity of benthic life strategies has been well documented and reported on in both lakes and ponds across the Arctic (e.g., [25, 26]).

All diatoms found in the lake sediment cores were benthic taxa with the one expected exception: the planktonic diatom Cyclotella radiosa in the core from West Lake, our deepest site. Therefore, almost all of the diatom production in our study is benthic and would respond to benthic processes. Aquatic mosses grow very slowly in the Arctic and can be up to 10 years old [27]. They can establish in Arctic lakes and ponds (e.g., [28, 29]), and their growth is limited primarily by snow and ice cover (i.e., they do not grow when it is completely dark and/or if they are frozen) but also by underwater irradiance, which is affected by the length of the day and the sun’s elevation. The length of ice-free conditions (i.e., conditions with sufficient light for photosynthesis) is critical, regardless of depth–open water, not snow-covered ice, are required for littoral mosses (and for diatom epiphytes) to grow. Consequently, the mechanism that we are investigating is as applicable in the lakes as it is in the shallower ponds.

Ice melt phenology

Gajewski argues that the melting candled ice for Arctic lakes (taken from observations of primarily Alaskan lakes [30]) would scour and disturb nearshore sediments and vegetation. This Alaskan lake setting that would not be reflective of the environment sampled on Pim Island or other High Arctic sites. In reality, many Arctic lakes support aquatic mosses and nearshore sediment (e.g., [29]). Our co-authors have witnessed ice melt at Cape Herschel and other High Arctic settings for decades and have repeatedly observed extensive, mature, and clearly intact moss banks emerge from the thawing ice year after year in our study ponds and in other lakes across the Arctic. The fronds of the mosses in many of these ponds and lakes are often well over 6 cm long, and frequently longer, with mosses of similar size in a High Arctic setting reaching ages of 6–10 years (e.g., [27]). Presumably Gajewski is assuming that, after scouring each year, new mosses would have time to grow and establish in the very short summer; however, given bryophyte life history and physiology, Gajewski’s hypothesis is untenable given the size of the mosses established in many of these lakes and ponds. It is possible that ice scour (but not due to candling) could indeed affect moss growth, but this has only been observed on larger water bodies whereby the wind can push a floating ice “pad” against the shore, resulting in moss disturbance limited to the immediate nearshore environment (for example at Char Lake, Cornwallis Island). However, that does not preclude the establishment of mosses at these sites (as in Char Lake; [25, 27]) as in many of these lakes one can see moss bands that start below the area of ice scour and continue to the limit of the photic zone.

It appears as if Gajewski is arguing, in a general sense, that ice melt processes can disturb nearshore sediments? Yes, of course, this could happen. We respectfully question the relevance of Gajewski’s comment to our study cores that: 1) show near-monotonic declines in excess ²¹⁰Pb activity to background levels; and 2) abrupt transitions in diatom assemblages that in some cases show almost 100% species turnover. Neither of these are possible if ice melt processes disturbed the sediments of our study ponds.

Gajewski’s comments on ice melting in lakes also lack the environmental context that influences ice phenology in these extreme climates, such as the high elevation lakes on Pim Island. As noted above, ice on some of these High Arctic lakes can be 3–5 m thick in August, with only shallow nearshore moats of open water at the height of summer thawing. Gajewski cites Hobbie [30] to support his contention that 80% of a High Arctic lake’s ice cover can thaw in 24 hours.

When asked about this contention, the chapter author John Hobbie (written communication, July 26, 2020) states that:

“The 1984 publication was based on detailed observations on a deep lake in the Brooks Range of northern Alaska. A 2 m-thick ice cover melted down to less than 0.5 m thick over 1 month. At this point the ice crystals separated and floated vertically. A slight breeze often set hundreds of square meters free; they immediately fell on their sides and melted within hours. Finally, a strong wind pushed the whole ice sheet towards the north end of the lake at which time all the ice crystals melted in one day. Thus, a sheet of very thin ice that had covered 80% of the lake disappeared in a single day. In this same publication a study of a lake in northern Greenland is described; a 3.4 m thickness of ice was reduced to 2.0 m in a single summer so there was no loss in area but a sizeable loss in volume. Lakes on Pim Island would be like the Greenland lake and would not thaw in 24 hours”.

Furthermore, the observations on Alaskan lakes in [30] were clearly important work but were not meant to be applied so broadly. Gajewski fails to acknowledge that sediment accumulation persists in the nearshore of these lakes and it is therefore not plausible that an entirely new population of nearshore vegetation is re-established every year. Also, if ice development were to preclude moss growth, why then are some of the shallowest ponds on Cape Herschel 100% moss covered (repeated personal observations). We have watched waterbodies lose their ice, repeatedly, over decades. If mosses were destroyed by ice each year, the massive moss beds (often several decades old) that carpet the bottom of these ponds and lakes could not exist.

Criticism 2. Gajewski challenges our results based on dating of the study cores

Specifically, he notes “Problems with their chronologies and lack of dating in some cores make it difficult to test hypotheses about the timing of changes in the diatom assemblages”

Dating is notoriously difficult in the Arctic due to very low rates and concentrations of ²¹⁰Pb deposition and is a well-known and documented challenge in Arctic paleolimnology [31]. Therefore, we were conservative in our interpretations based on dates. For our “warm” sites we stratigraphically matched our cores to previously dated sediment cores from the same ponds [22] using biostratigraphic boundaries clearly delineated by the diatoms (i.e., the occurrence of almost 100% species turnover), a commonly utilized and accepted method not only in paleolimnology but also in paleoceanography and geology, and which is the basis for the worldwide correlation of the geological timescale. The diatom boundaries in the previous study [22] occurred in the cores at approximately equilibrium depth (the point where supported fraction of ²¹⁰Pb met the unsupported fraction)–which was ~ 1850 AD [22]. Gajewski’s contention is that we cannot tell that the changes occurring at the “warm” sites occurred before some of our “cool” sites. However, at Moraine Pond and Paradise Pond (both located at Cape Herschel), the diatom stratigraphic changes occurred when there was still unsupported ²¹⁰Pb in the sediment (i.e., before equilibrium depth is reached). The depth of equilibrium is clear in our figures as we also include the activities for unsupported ²¹⁰Pb (which is inferred (as is standard) by the activity of ²¹⁴Bi). This shows, whether one disagrees with our chronologies or not, that the diatom changes in the “cool” sites likely occurred after the changes in the “warm” sites, i.e., after 1850 AD (as there was still unsupported ²¹⁰Pb activity in the core at the point of transition).

Additionally, since the publication of our 2017 study, anthropogenic Hg has been analyzed in some of the same sediment cores [32] under discussion here. Due to industrialization and atmospheric deposition of Hg, the start of a post-1850 enrichment profile of Hg can be used as an independent stratigraphic marker of ~1850 AD [33]. Col Pond, Moraine Pond, SV8, and West Lake all showed marked Hg enrichment commencing with our chronological estimate for 1850 AD, while Proteus Lake also recorded a strong signal of Hg enrichment, the beginning of which (representing ~1850 AD) was within the error of the basal date of our ²¹⁰Pb age-depth model [32]. Elison Lake also showed a post-1850 enrichment that matched the diatom stratigraphic changes (i.e., matching our 1850 AD date [22]), however the profile was muted because the magnitude is smaller, such that the timing of the trend is less clear [32]. This is unsurprising as Elison Lake is a large but shallow pond with a small catchment for its size. Paradise Pond and SV5 also had low magnitude Hg enrichment [32], which matches the low levels of ²¹⁰Pb deposition at these sites–such that a lack of a strong trend due to low Hg deposition was foreseeable. The timing of the marked post-1850 Hg enrichment in Col Pond matched our diatom changes. An enrichment in Elison Lake, although less pronounced, matched the published chronology [22]. The same signal matching the dating in multiple cores provides considerable support for our ²¹⁰Pb chronologies and that the diatom shifts recorded at the “cool” sites Moraine Pond and Paradise Pond occurred after the “warm” sites.

Gajewski takes issue with our basal dates for the Sverdrup Pass cores. However, the objective of our study was to examine diatom responses to recent regional warming during the past ~200 years. Thus, the sole purpose of obtaining ¹⁴C dates for the “oasis” cores was to ensure that the cores were at least 200 years old, but not to generate an age-depth model through extrapolation of the ²¹⁰Pb dates to the one ¹⁴C date. We had ²¹⁰Pb dated the SV5 core and knew that the activities were too low to obtain a reliable chronology, and so basal dates were needed. The SV8 core had not been dated yet, so as a precaution, we also sent basal sediments from that core for ¹⁴C dating to ensure the core was long enough (i.e., that the bottom was at least 200 years old) to answer our question. Fortunately, SV8 yielded a workable ²¹⁰Pb chronology. This is the reason why the “oasis” cores have ¹⁴C basal dates, while the rest of the cores had none; the ²¹⁰Pb dates were sufficient for the rest of the sites to demonstrate that the cores spanned our period of interest. The difference in the sedimentation rates inferred by the ²¹⁰Pb chronologies and from the base of the last ²¹⁰Pb dated interval to the ¹⁴C date are not a sufficient reason to distrust the ²¹⁰Pb chronology or suspect these cores are less than 200 years old.

Gajewski suggests four possibilities for the discrepancy between the single ¹⁴C date and end of the ²¹⁰Pb dates for the SV8 core (that there is little accumulation in the core, that the dates are too old, that there is a hiatus in the core, and that the sediment is “removed from the system” by mixing). Sediment cores are known to dewater with depth, such that intervals at the top of the core often represent shorter time intervals than the same-sized interval at depth. Therefore, many age-depth models are often curved, as is the SV8 model (e.g., [34]). A combination of dewatering and lower sedimentation rates in deeper sediments (as these sites have still experienced increases in the length of the ice-free season and increased production as a result (see chl a Fig)) is a plausible explanation of the discrepancy in the ²¹⁰Pb and ¹⁴C dates. There is also a possibility that the basal dates are too old. The “oasis” ponds are located in dolomite landscape, and so could incorporate old carbon. To obtain the best possible dates, we sent the sediments to an expert in macrofossil retrieval and identification for ¹⁴C dating who isolated woody herbaceous stem fragments (SV Pond 5 and SV Pond 8), sedge Carex sp. achenes (SV Pond 8) and partial leaves of Dryas integrifolia and Cassiope sp. (SV Pond 5), which were analyzed for age dating. The choice of dated material should have mitigated potential of old carbon contamination. However, despite our best efforts, there is a small possibility of old carbon being incorporated into the dated material (which, of course, is true of all radiocarbon dating).

Gajewski’s third and fourth speculative possibilities for the discrepancies in the ²¹⁰Pb and ¹⁴C basal dates, which were presented without supporting evidence or having observed the cores, are not likely. If there was a hiatus in the core, due to the ponds being continuously ice-covered, we would expect changes in the core lithology and in the diatoms in the lead up, which show little change through that period, especially at the high resolution at which the diatom cores were counted. The last possibility deemed by Gajewski as “likely” was that the sediments were “constantly disturbed, re-suspended and removed”. There is no evidence to support his assertions. The preservation of sedimentary profiles with curves of declining unsupported ²¹⁰Pb indicates that sediments are, in fact, being deposited. This is further supported by anthropogenic Hg profiles for the cores [32], which would be impossible if this period of the sediment record was somehow “missing”. If all shallow ponds and lakes in the Arctic were mixed, then the hundreds of papers on ponds and shallow lakes would be invalid. Thankfully, this is not the case, as mixed cores can easily be identified by the variety of sediment proxies studied.

There are two main ways of checking whether a core has been mixed: 1) by the decay curves for ²¹⁰Pb activities and, 2) by the stratigraphic changes (i.e., diatom or other proxy changes) in the core. Note that all our cores are sectioned and counted at very fine (0.25 cm– 0.5 cm) intervals due to the commonly low sedimentation rates of these Arctic systems. In the lakes and ponds where initial ²¹⁰Pb concentrations were not too low, we record declines in unsupported ²¹⁰Pb activities with depth until reaching supported ²¹⁰Pb (background) levels (i.e., the typical ²¹⁰Pb profile). If mixed, these profiles would be more homogeneous throughout, with very little directional trend. That is not the case with SV8. We also record similar marked anthropogenic Hg enrichment profiles [32] in these cores. Again, if these cores were mixed, we would not see evidence of Hg enrichment, but rather a more homogeneous profile. Further support for sediment mixing not being an issue comes from the striking changes we record in the fossil diatoms themselves, with marked shifts over just one or two centimeters (e.g., Col Pond, Elison Lake, Moraine Pond, Plateau Pond 2, SV5, West Lake). Some of these changes (such as at Col Pond) are nearly complete taxonomic turnovers. Even the sedimentary records with more subtle assemblage shifts (High Lake, Proteus Lake, and SV8) show some taxonomic changes (e.g., Hygropetra balfouriana coming in at the top 0.5 cm at High Lake that was hitherto absent), which, combined with the more marked radio-isotope profiles indicate that these sedimentary sequences are not mixed. Again, if the sediments of all shallow ponds/lakes were mixed, we would not record such striking taxonomic changes over just a few centimetres. Finally, to support that these cores have not been mixed by freeze-thaw cycles, we refer to our own observations of the sites where the cores from [22] were collected (by Blake through the ice in the 1970s). Even now, decades after the cores were taken, the imprints of the original coring holes remain in the middle of each pond, evident during flyovers in a helicopter. The impressions from the deep, meters-long cores taken by Blake through the ice remain because there is no considerable disturbance of the sediment.

Criticism 3. Gajewski argues our interpretation of diatom changes is not consistent with long-term studies and he argues that diagenesis and mixing further complicate matters

Specifically, he notes “The interpretation of changes in the diatoms in these short cores is contradicted by a longer sediment sequence from the same area. In addition, diagenesis or mixing of the uppermost sediments makes it difficult to draw conclusions from such extremely short sediment sequences.”

The Gajewski criticism asserts the need for a Holocene-scale context. However, as is clear throughout our paper, our study was focused on investigating a potential driver of a commonly reported diatom response to recent regional warming (i.e., over the past ~200 years), not to climate fluctuations over the Holocene. Some of our profiles extend beyond the pre-industrial period, due to the generally slow sedimentation rate in the Arctic, but our period of interest is well-defined throughout our paper as the past ~200 years. Importantly, a Holocene context would not be appropriate for investigating the effect of microclimate differences between our sites, which would likely have experienced minimal change over the short (~200 years) timeframes; however, at millennial timescales, microclimates would likely have changed. For example, the Cape Herschel foreland only emerged from the sea ~8,500 BP, with rapid isostatic uplift occurring in the first 2,000 years, with evidence suggesting that the Cape was actually an island between 6,000–5,500 BP [35]. Some of the ponds closest to the ocean are much younger–for example “Horseshoe Pond” has a basal date of 1,740 +/- 70 (GSC-2841;[35]). The isostatic uplift was also asymmetrical [35, 36]. Additionally, regional glaciers have expanded and contracted throughout the Holocene [37–39], and ocean circulation in the Nares Strait may have been quite different, with the Nares Strait being blocked with glacial ice until 10,000 years ago [40], which would affect any interpretations one might make to infer the microclimate at these sites in the early-mid Holocene. Overall, it would be extremely tenuous to infer microclimatic differences between the sites over the entire Holocene, as the area has been so dynamic over these geologic timescales. The focus of our paper is clearly on the past ~200 years. Confusingly, Gajewski does not believe we can characterize the ponds and lakes into micro-habitats based on several decades of detailed observations yet suggests we should extend interpretations of these microclimatic differences to thousands of years ago.

1) Gajewski claims the categorization of our study sites into four grouping is based on single or limited visits

The dates provided in the table were simply the dates of our detailed water chemistry analyses and were not the only times that the sites were visited or observed. Rather, many of the study ponds on Cape Herschel are visible from our base camp and even some that are not, such as Moraine Pond, were visited almost daily. Frequent aerial surveys and on-the-ground sampling, dating back to mid-1970s, support our ice cover phenologies. Our observations on ice melt, as well as those of others, extend back nearly forty years.

2) Gajewski argues that High Arctic ponds do not accumulate organic matter

Gajewski mis-read a progress report on the first coring attempts at Cape Herschel, when the Cape was still under snow cover and the ponds were difficult to locate [59]. All the ponds that Gajewski assumed had no organic sediment are some of the same study ponds discussed in our study [1]. He cites Blake [59] to support his claim; we provide a written statement from Blake that directly contradicts Gajewski’s interpretation. The collection of sediment cores (several decimeters in length) from the study ponds underscores the obvious–High Arctic ponds at Cape Herschel do accumulate organic matter.

3) Gajewski argues that meteorological data indicate no differences in temperature between Cape Herschel and Sverdrup pass

Gajewski compares temperature records from the two regions over years where there is little overlapping data, making direct comparisons impossible. Still, even with his own analysis, he shows Sverdrup Pass is likely warmer, especially at the critical spring period that determines ice melt. Gajewski, who presented these data, then concludes that these temperature data do not help resolve the issue he raised.

4) Gajewski argues that the chronologies in Griffiths et al. [1] are not robust enough to distinguish a difference in the timing of the diatom changes between the “warm” sites and the “cool” sites

We acknowledged in our study [1] the challenges of sediment dating in the High Arctic, which is why we were very conservative in our interpretations based on dates. The changes at the “warm” sites occurred where the equilibrium depth had been reached (i.e., where there was no longer unsupported ²¹⁰Pb in the sediment) based on matching cores from the same ponds in a previously published study (i.e., [22]). In contrast, the diatom stratigraphic changes for the “cool” sites occurred when there was still unsupported ²¹⁰Pb in the sediment (i.e., before equilibrium depth was reached). This suggests that the diatom changes in the “cool” sites most likely occurred after the changes in the “warm” sites, i.e., after 1850 AD (as there was still activity in the core at the point of transition). After publication, our chronologies have been independently verified with post-1850 Hg enrichment profiles [32] from the same sediment cores, which can be used as an independent stratigraphic marker of ~1850 AD.

5) Gajewski argues that diatom concentration, biogenic silica and loss-on-ignition data are necessary for interpretations of past change

The proxies Gajewski favors are weaker metrics of change than what we report and are often fraught with difficulties. Sedimentary chl a includes all algal groups (unlike biogenic silica) and the spectral method captures the main diagenetic products, which provides a consistent chlorophyll signal (and by extension a signal of whole-lake production) over long timescales much better than LOI. Concentration data are best presented as fluxes and the Arctic dating models are not robust enough for such analyses. Importantly, though, the proxies Gajewski champions would not change our interpretations. In fact, diatom concentration data [50], biogenic silica and LOI [51] (as well as % carbon; [32]) exist for several of the study cores, and these data directly contradict Gajewski’s claims. For example, diatom concentration data typically show highest concentrations in the surface sediments, which refutes his claim that warmer temperatures lead to diatom dissolution. Loss-on-ignition data similarly contradict his claim that the study ponds do not accumulate organic matter, as does the proxy of organic carbon in the sediments [32].

6) Gajewski argues that the trends in diatom profiles can be explained by dissolution of the valves associated with periods of warmer temperatures

If Gajewski’s assertion was correct, it would refute hundreds of papers published every year on diatom paleolimnology from subarctic, temperate, and tropical regions. His own co-authored publication [56] on diatoms from Arctic regions warmer than our study sites directly contradict his argument.

7) Gajewski claims that rapidly melting candled ice each spring scours High Arctic lakes and ponds of their sediment and mosses

Gajewski’s argument about ice melt is based on a study from an Alaskan lake ~2,800 km to the southwest of our study site. His incorrect application of that study to High Arctic lakes was corrected by the original author. Gajewski provides personal anecdotes on ice melt in Arctic lakes to make the claim that sediment disturbance in nearshore environments is possible. This claim is most easily refuted by the presence of multi-year moss growth along the shores and bottoms of the ponds and the fact that we collected sediment cores with marked stratigraphic profiles and / or ²¹⁰Pb profiles with near-monotonic declines in excess ²¹⁰Pb activities from all of the study sites. We are discussing our data, which show this is not an issue. Also, we note that we do not core at the edges of lakes and ponds.