Today’s Arctic climate is warming faster than most other regions and losing summer sea-ice cover at historically unprecedented rates [27, 186]. This pattern of “Arctic amplification” is due to the changes in albedo [145], heat exchange between the atmosphere and ocean and other processes [146, 172] that are consistent with paleoclimate evidence for elevated polar temperatures during past warm periods [21, 126]. In addition to sea-ice decline, concerns exist about other climate-related processes that affect Arctic Ocean environments, such as submarine methane release [166], glacier melting [70], greater riverine discharge [147], marine ecosystem shifts [75], changes in biological productivity [9, 198], habitat loss and extinction [163], and carbon cycling [5, 180].

Instrumental and observational records are too short to fully evaluate the long-term effects of climate change on Arctic ecosystems, but two disparate fields—paleoclimatology and molecular genetics—now provide a unique context for assessment of climate change in the Arctic. In contrast to model simulations of future climatic and ecosystem change, paleoclimatology and genetics look back in time, using geochronology, physical, geochemical and paleoecological proxy methods, and DNA-based molecular clock analyses. Here we assess marine ecosystem response to past climate changes using an integrated approach based on Arctic sediment records of past intervals of warmth, orbital-scale glacial-interglacial cycles, and abrupt climate transitions coupled with DNA-based phylogenetic reconstructions and fossil records of polar vertebrate lineages. Although all parts of Arctic marine ecosystems cannot be studied, our study involves a wide variety of taxonomic groups and several key biological metrics of Arctic ecosystems including biodiversity, primary productivity, biogeography (range expansion and contraction) and hybridization. We address the fundamental question: does climate change cause large-scale loss of biodiversity through species’ extinctions (α diversity) or rearrangement of species abundances within local communities, geographic range shifts (β diversity) [52, 65], or ecosystem restructuring [28, 76].

Advances in Arctic paleoclimatology

Most early studies of the Arctic Ocean sedimentary record were based on cores taken from research stations floating on sea ice in the 1960s and 1970s, which provided important discoveries but were geographically limited and lacked sufficient stratigraphic and age control [183]. Since the early 1990s, cruises led by German, Swedish, Canadian, Russian, and US researchers expanded the spatial and temporal coverage of Arctic sediment cores used for paleoceanography [135] (Fig. 1, Supplementary Table 1). In addition, greatly improved age control now allows a more complete reconstruction of Arctic Cenozoic climate history that, with exceptions, allows correlations with paleoclimate records from extra-Arctic regions.

Fig. 1
figure 1

Map of selected sediment core sites used for paleoceanographic reconstruction. The key symbols designate the age of the record. The black triangle is the Cenozoic record from IODP ACEX Project [10, 129]. “Orbital” cores record multiple glacial interglacial cycles. “Orbital, MIS 11” cores are orbital records that include warm interglacial Marine Isotope Stage 11 ~400 ka. “Holocene, Late Holocene” cores contain the last 1000–2000 years. Core sites keyed as “Productivity” were used in Arctic productivity studies. Red lines show the approximate margins of ice sheets. The Laurentide-Innuitian, Greenland and Eurasian ice sheet margins are maximum extent during the Quaternary [97, 188]. The Iceland ice sheet extent is the LGM [89]. Following [97], red crosshatched areas may or may not have been covered by ice sheets. NWR Northwind Ridge. Supplementary Table 1 provides information about core sites. Basemap is International Bathymetric Chart of the Arctic Ocean (IBCAO) [96]. See O’Regan [135] and Stein et al. [182] for additional core records

Rapid advances have also come from the development of sediment proxy methods used to reconstruct environmental conditions and biological, chemical and physical processes influenced by climate (Supplementary Table 2). Examples used in the following discussion of Arctic climate and ecosystem evolution include micropaleontological records of benthic and pelagic communities, proxies of sea-ice cover, sediment transport, marine biological productivity, ocean temperature, salinity, dissolved oxygen and circulation, and ice sheet and ice shelf activity.

Cenozoic climate in the Arctic

In 2004, the Arctic Coring Expedition (ACEX), part of the Integrated Ocean Drilling Program (IODP Expedition 302), recovered 428 m of sediment from the central Arctic Lomonosov Ridge dating back to 56 million years (Ma) [10, 129, 182]. For the first time, a unique, though incomplete record of Arctic climatic and faunal evolution can be compared to the Cenozoic greenhouse-to-icehouse climate transition established on the basis of deep-sea foraminiferal δ18O records of sea level and temperature and ice core records of atmospheric CO2 concentrations and temperature (Fig. 2). Initial study of ACEX Paleocene-Eocene micropaleontological records Expedition 302 [59] identified numerous diatoms (~40 taxa), silicoflagellates and ebridians (~40), palynomorphs (~58), agglutinated benthic foraminifera (~40) and, due to poor pre-Miocene preservation of calcareous shells, lesser numbers of calcareous nannoplankton, calcareous benthic and planktic foraminifers, and ostracode taxa.

Fig. 2
figure 2

Cenozoic climate history from central Arctic ACEX core (ae) compared with global ice volume and temperature (f, [30] and atmospheric CO2 concentrations (g compilation from [17] (blue curve) and alkenone-based pCO2 (red) from [208]. Major steps in Cenozoic climate events are labeled (Paleocene-Eocene Thermal Maximum (PETM), Eocene Climate Optimum, Eocene Thermal Maximum 2 (ETM2), Eocene/Oligocene (E/O) cooling, mid-Miocene Climate Optimum). a Cenozoic IRD, a sea ice proxy, based on terrigenous coarse sand fraction [179]. b Iron-oxide grain record of first perennial sea ice ~44 Ma (260 m core depth) and subsequent sea-ice variability [42]. c Paleoproductivity reconstruction from nitrogen fraction showing low productivity (<20 g C m−2 a−1) during the ice-covered Miocene and high productivity (~50–100 g C m−2 a−1) during warm, ice-free, and biologically productive Paleocene-early Eocene [106]. d Sea-ice diatom Synedropsis spp. abundance [184]. e TEX86- derived SST showing the PETM [174]

ACEX researchers also investigated key climatic and ecosystem events including the Paleocene-Eocene Thermal Maximum (PETM), an ~170,000-year long warm period about 56 Ma when sea-surface temperatures in the Arctic (SST) reached 22 °C [174]. In addition, ACEX recovered sediment from two younger hyperthermal periods—the Eocene Thermal Maximum 2 (ETM2) at 53.5 Ma [175] and the Azolla horizon ~48.5 Ma [22]. During ETM2 TEX86-derived SST estimates indicate Arctic temperatures reached 25 °C, dinoflagellate cysts document freshwater influx and eutrophication, and palm pollen suggests winter temperatures on adjacent continents exceeded 8 °C. The dominance of the genus Azolla, a free-floating, freshwater fern, and associated microfossils, characterized an ~800,000-year long interval of episodic fresh surface water, a stratified ocean, endemism in silicoflagellates and ebridians [134], SSTs of 10–14 °C [22], and intermittent oxygen depletion [181] (Fig. 2d, f). During Paleocene-Eocene hyperthermal events, marine primary productivity in the central Arctic varied greatly with maximum values reaching 50–100 C g m−2 year−1 [106, 181]. These values are comparable to those from today’s highly productive Arctic marginal ice zones [138] and higher than estimates for the central Arctic Ocean over the last 18 Ma, including today (Fig. 2c).

During the interval 48–45 Ma, Arctic SSTs fell by as much as 5–10 °C depending on which proxy method is used [182, 200]. This cooling is coincident with the inception of a winter sea-ice regime seen in ice-rafted debris (IRD) [179] and sea-ice diatom records [184] (Fig. 2a). There is also lithological evidence for ephemeral perennial sea ice at times between 47 and 44 Ma [42] (Fig. 2b). Climate history of the late Eocene, Oligocene and early Miocene is poorly known because one age model calls for a major sedimentary unconformity from 44 to 18 Ma [11], and another for a condensed zone representing the interval from 36 to 12 Ma [149]. This introduces uncertainty in identifying key Cenozoic cooling events, such as the Eocene/Oligocene transition ~34 Ma, and their biological impacts. There is, nonetheless, evidence for stepwise cooling during the last 18 Ma of the Cenozoic greenhouse-icehouse transition. For example, IRD, mineral, and radiogenic proxies record a shift from a mid-Miocene climatic optimum (~15 Ma) toward a colder climate since about 13 Ma [41, 66, 79, 179].

Early- to mid-Pliocene global climate (5–3 Ma) serves as an important benchmark for understanding modern climate because Pliocene atmospheric CO2 concentrations were near today’s level (400 ppmv, [139, 171]), but global mean annual temperature (MAT) was about 2.5–3 °C higher [53] and peak sea level ~22 m higher [127]. Pliocene Arctic Ocean summer SSTs were appreciably warmer than modern and seasonally sea-ice-free conditions existed in some regions [108, 121]. Non-marine proxy records from continental sections also point to a warm Pliocene climate in the high latitudes of the northern hemisphere. At Lake El’gygytgyn (Lake “E”) in Siberia summer temperatures were 8 °C warmer than modern [21] and at Ellesmere Island, Canada, summer and MAT were 11.8 and 18.3 °C higher than today [13]. In addition to periods of warmth, the Pliocene saw continued intensification of Northern Hemisphere glaciations and crossing of climate thresholds at 4 and 2.75 Ma as ice sheets reached Arctic coastlines [107]. Such warm Pliocene conditions allowed a major trans-Arctic migration of mollusks [58, 195], ostracodes [37], and other groups ~4.5–3.8 Ma when the Bering Strait opened [71, 194]. The direction of this migration was mainly from Pacific-to-Atlantic and probably led to the evolution of some of today’s endemic Arctic species.

Quaternary glacial-interglacial cycles

Climatic cycles driven by changes in earth’s orbital geometry (eccentricity, tilt and precession) are known throughout the geological record. Orbital cycles have been recognized in early to mid Eocene Arctic sediments from the ACEX core site [141, 167], but they are much better known from Quaternary sediments deposited during the last 600 ka across the entire Arctic Ocean. Quaternary glacial-interglacial cycles (here we use marine oxygen isotopic stage (MIS) terminology, [115]) signify changes in global ice volume and ocean temperature inferred from deep-sea foraminiferal oxygen isotopes (Fig. 3a). These are accompanied by changes in atmospheric temperature and CO2 concentrations known from Antarctic ice cores and other proxy records (Fig. 3b, c). In the Arctic Ocean, glacial-interglacial cycles are seen in a variety of proxies: manganese concentrations [92, 118, 148, 155, 156], sediment physical properties (grain size, bulk density) [136], mineral assemblages and trace elements [61], organic biomarkers [204], and stable isotopes [1, 153, 176] (Fig. 3d–f). Variability in these proxies reflects massive changes in ice cover, river runoff, and ocean circulation during opposing extremes of interglacial warmth with summer sea-ice-free conditions and glacial-age ice cover.

Fig. 3
figure 3

Quaternary Climate. Global (top panels ac) and Arctic (bottom panels df) climate proxies for the last 600,000 years. a Benthic oxygen isotope curve reflects global ice volume and temperature, marine isotope stages are numbered [115], b EPICA dome C deuterium, a temperature proxy [99], c EPICA dome C CO2 curves [measured at Bern (green), Grenoble (blue), Taylor Dome (gray) and Vostok (red)] [173]. Deuterium and CO2 values are on EDC3 gas age scale, d Arctic ostracode density composite from five western Arctic cores (Cronin et al. 2013), e Manganese content in sediments from Oden 96-12 core [Lomonosov Ridge [92], f Sediment density (Lomonosov Ridge IODP 302 ACEX core [136]). Green bars show periods that, based on Arctic proxies, were likely seasonally sea-ice free. These correspond to particularly warm marine isotope stages 5e and 11. Gray bars denote glacial stages with thick Arctic sea ice inferred from proxies corresponding to marine isotope stages 2, 6, 10, and 12

Although not gaining as much attention as past warm periods, glacial periods in the Arctic and adjacent subarctic deserve special attention because they provide a stark environmental contrast with interglacials and concrete evidence for the resiliency of marine ecosystems in the face of large-scale climate oscillations. Records of the Last Glacial Maximum (LGM, MIS 2, ~24–19 ka) and the penultimate glacial MIS 6 (~150 ka) have excellent age control [150], broad spatial sediment core coverage, extensive submarine geophysical surveys, and onshore glacial geological mapping. At these times, the Arctic Ocean was reduced to ~50 % of its current area due to the combined effects of a 125 m fall in global sea level (increased ice-sheet volume), which exposed the vast Arctic continental shelves, and the expansion of ice sheets and ice shelves bordering the Arctic Ocean [43, 90, 91, 93, 94]. During glacial maxima, the cryosphere, including ice sheets, ice shelves, glaciers and sea ice, was substantially more extensive than what we see today (Fig. 1). The Laurentide, Innuitian, Eurasian, Barents Sea-Svalbard, and Icelandic Ice Sheets covered large parts of continental regions adjacent to the Arctic Ocean [56, 188], but perhaps as important, extensive ice shelves as thick as 1 km have been identified from submarine glacial landforms mapped using geophysical methods on the Chukchi margin and Yermak Plateau [97, 98, 131, 196], the Lomonosov Ridge and Chukchi Plateau [151], the Lomonosov Ridge [110], and the Morris Jesup Rise [95]. During peak glacial conditions, sea ice was so thick during glacial maxima that little or no IRD could be transported to the central basin from continental margins leaving a sediment-starved central Arctic Ocean [150, 153]. Although the thickness of glacial-age sea ice is not known precisely, multiyear sea ice, called paleocrystic ice, thicker than today’s >40 m-thick ice shelves off Ellesmere Island, may have dominated the glacial Arctic Ocean before the main phase of deglaciation began at 15 ka [20].

At the same time that the LGM Arctic Ocean proper was dominated by thick sea ice and ice shelves, sea ice extended far southward into subarctic regions of the Nordic Seas, the northern North Atlantic and the Bering Sea. Using dinoflagellate cyst assemblages from more than 50 core sites, de Vernal et al. [48] reconstructed spatial patterns of LGM sea ice, SST and sea-surface salinity from mid-to-high latitudes across the Northern Hemisphere. Among their findings were the presence of mid-latitude sea ice, stronger seasonality, nearshore to offshore SST gradients, and reduced surface salinities. Planktic foraminiferal assemblages and stable isotope values [132], and epipelagic ostracodes [38] also indicate southward migration of sea ice into the Nordic Seas and North Atlantic during glacial periods MIS 2, 4, and 6. Similarly, in the southern Bering Sea, LGM sea ice is evident from diatom assemblages [24].

Biological response to glacial-interglacial cycles

Perhaps the most striking biological manifestations of orbital cycles in the Arctic Ocean and surrounding seas are patterns of microfossil density, species diversity, and assemblage composition, which, when combined with physical and geochemical proxies, provide compelling evidence for ecosystem response to climate change. In contrast to the pre-Miocene sediments in the Arctic, which lack calcareous microfossils [see above], commonly preserved microfossil groups in the Quaternary include benthic and planktic [153, 201]) foraminifera, calcareous nannofossils [12], and ostracodes [39]. Along Arctic margins and in the Nordic Seas, diatoms [16], tintinnids (planktic ciliates, [169]), and dinoflagellates [120] also occur. In addition to faunal and floral remains, there are indirect proxies of oscillating biological activity, notably organic biomarkers of sea-ice diatoms and phytoplankton [60] and sediment manganese oxyhydroxide content related to terrestrial input, ice cover, and bioturbation [117, 118].

The density of calcareous microfossils in sediments from central Arctic ridges is directly linked to interglacial and glacial climate regimes and changes in sea-ice cover, surface productivity, sedimentation, and post-depositional processes [12, 119]. It is well established that foraminifera (benthic and planktic) and ostracodes are major components of the sand-sized fraction in interglacial sediments in contrast to near absence in glacial-age sediments [1, 80] (Fig. 3d). In addition to density, microfossil biodiversity is extremely variable in the eastern Arctic, where benthic foraminiferal diversity measured by the Fisher α and Shannon Wiener indices varied several-fold during the last glacial cycle [203], and in the western Arctic over the last few glacial-interglacial cycles [154]. Mechanisms driving species diversity patterns within the Arctic include the strength of inflowing warm Atlantic water, ice cover and surface productivity.

Microfossil assemblage composition (e.g. β diversity), measured by the relative abundances of environmentally sensitive species and genera, is also a useful measure of ecosystem dynamics. One striking example of climate-driven migration is the Pacific diatom Neodenticula seminae, a species that sediment records show disappeared in the North Atlantic Ocean ~800 ka. It has recently been discovered that this species has migrated back into the North Atlantic and Nordic Seas during the last 2 decades almost certainly in response to higher ocean temperatures allowing inter-oceanic migration [125, 161]. In Arctic cores, biogeographic range shifts occur frequently due to changes in climate and ocean circulation over various timescales. One widely used benthic foraminifera, Epistominella exigua, is a phytodetritus-eating, opportunistic species that dominates modern oceanic frontal zones [73]. Microfossil assemblages with dominant E. exigua indicate seasonally sea-ice-free and/or marginal ice zone conditions that characterized the early-mid Quaternary (~1.5 Ma–300 ka) prior to the development of perennial sea ice. This species is common during warm interglacials MIS 5 (125 ka, the Eemian Interglacial) and MIS 11 (400 ka) but absent during glacial periods [154].

As discussed above, ice shelves and thick sea ice covered the glacial Arctic Ocean, and, as a consequence, species were forced to migrate southward into extra-Arctic regions on a large scale. We can track the range expansion and contraction of sea-ice and marginal ice zone species because the ecology of several groups is well known from large, pan-Arctic surface sediment databases. In the case of dinoflagellates, fossil assemblages are used to estimate months of sea ice cover in subarctic seas [48, 49]. The epipelagic ostracode species Acetabulastoma arcticum, which today lives as a parasite on sea-ice dwelling species of the amphipod Gammarus, is also a useful sea-ice proxy in the Arctic and adjacent seas [38]. As expected from its ecology, this species occurs only in glacial age sediments (MIS 2, 4 and 6) in cores from the Nordic Seas and North Atlantic.

There is also evidence in the Arctic for two well-known global climate transitions involving changes in the pattern of orbital glacial-interglacial cycles—the Mid-Pleistocene Transition between 1.2 Ma and 700 ka [26], and the mid-Brunhes Event ~450–400 ka [205]. Importantly, both climate transitions involved changes in Arctic sea-ice ecosystems. For example, the mid-Pleistocene transition, a shift from 41 to 100-kyr glacial-interglacial cycles, is characterized by faunal turnover (including regional extinctions) in Arctic foraminifera and ostracodes and reduced marine productivity. These signal a change from a seasonally ice-free to mostly perennial sea-ice cover during interglacial periods [154]. Globally, the mid-Brunhes Event coincides with the glacial termination between MIS 12 and MIS 11 (~450–400 ka) after which interglacial periods had smaller continental ice sheets, higher sea level, warmer temperatures, and higher atmospheric CO2 concentrations. MIS 11 was an exceptionally warm interglacial, notable because, whereas atmospheric CO2 concentrations (~280 ppmv) and orbital insolation were similar to those of the Holocene interglacial, global sea level was higher than today, perhaps due to the collapse of parts of the Antarctic Ice Sheet [84, 160, 165]. Arctic sediments from the Northwind, Mendeleev, and Lomonosov Ridges show that during MIS 11, there was no summer sea ice and SSTs reached 8–10 °C [39]. Warm Arctic Ocean summers during MIS 11 are also evident in the Nordic seas and the subpolar North Atlantic [15, 100], in Lake “E” sediments [123] and from terrestrial pollen in cores off southern Greenland [50]. Subsequent interglacial and interstadial periods (MIS 9, 7, 5 and 3) also experienced, at least at times, summer sea-ice-free conditions [133, 137].

In sum, the contrast between glacial and interglacial oceanic environmental conditions in the Arctic and subarctic reflects frequent biogeographic marine ecosystem shifts of several thousand kilometers supporting the view that climate change alters β diversity but does not cause the systematic loss of species.

Abrupt, suborbital climate transitions

One pressing question is whether climate has reached a “tipping point” such that we are witnessing an abrupt climate reversal (over a century or less) [25]. The last deglacial period (~19–11.7 ka) included several well-known millennial climate events whose onsets and terminations were abrupt transitions. These include stadial periods called Heinrich Event 1 (H1, 17–15 ka), the Younger Dryas climate reversal (YD, 13–11.7 ka) and interstadials called the Bølling-Allerød (B/A, 14.6–13 ka), and the Preboreal period (PB, 11.7–9 ka) (Fig. 4). Importantly, past abrupt climate reversals had major impacts on Arctic marine ecosystems over timescales much shorter than orbital cycles and they provide a unique context for today’s changing Arctic. The last glacial period from 60 to 15 ka included multiple Heinrich Events, identified by ice-rafted sediment and sea-surface cooling in the North Atlantic Ocean, and Dansgaard-Oeschger (DO) cycles identified in Greenland ice core oxygen isotopes and extra-Arctic proxy records.

Fig. 4
figure 4

Abrupt climate change in the Arctic during the last deglacial period including Bølling-Allerød and Preboreal interstadials and Heinrich 1 (H1, 17–15 ka) and Younger Dryas (YD, 13–11.7 ka) stadials. a Benthic foraminiferal record of marine productivity from core PS2837-5 (1023 m water depth), Yermak Plateau, showing high interstadial and low stadial (YD) productivity [202]. bc. Two species of benthic foraminifera from core PS51/154 (270 m water depth) highlight ecosystem changes during abrupt stadial-interstadial oscillations [190]. Absence of C. neoteretis (dark blue) and dominance of C. reniforme (light blue, y axis reversed) at 15 and 13 ka signify ocean circulation changes related to freshwater influx at the end of H1 and the YD. d Oxygen isotope values of planktic foraminifer Neogloboquadrina pachyderma (sin) in core PS2458 from Laptev Sea continental margin (983 m water depth) show abrupt decline at 13 ka due to fresh water influx during YD [177]. Higher δ18O values reflect ice sheet retreat during Preboreal and Bølling-Allerød

Changes in the dominant species in benthic foraminifer assemblages occurred on the Yermak Plateau and Barents Sea slope during stadial-interstadial events. These changes suggest a more than twofold change in marine productivity (from 30 to >60 g C m−2 year−1) (Fig. 4a) [202]. On the Laptev Sea margin, changes in dominant benthic foraminiferal species occur over a century or less at the onset and termination of H1 and the YD. Decreases in planktic foraminiferal stable isotope values during the YD up to 1 per mil are known from the Beaufort and Laptev Seas and the Mendeleev Ridge [6, 157, 177]. Faunal and isotopic proxies signify complex hydrological changes in the surface and subsurface Arctic Ocean caused by freshwater influx probably from multiple catastrophic glacial lake drainage episodes [192] and changes in the strength of inflowing Atlantic water. It is worth noting that other types of catastrophic events disrupted Arctic marine ecosystems, such as mega-iceberg discharges caused by Eurasian Ice Sheet surge and collapse, which scoured the seafloor in the Kara-Barents Seas [95, 131, 152] and central Arctic as far back as 500,000 ka [110]. Space limits our discussion to the Arctic Ocean proper, but suborbital millennial-scale events also caused frequent marine ecosystem reorganizations in the Nordic Seas during the last glacial-interglacial cycle [14, 78].

Holocene climate oscillations

Although smaller in scale than glacial-interglacial cycles, climate variability during the Holocene interglacial period had significant impacts on polar biological systems. There is extensive evidence for an Early Holocene Thermal Maximum (EHTM) ~11–7 ka with regionally variable seasonally sea-ice-free conditions based on circum-Arctic lake and ice core records [101, 187], glacial geology [122], ocean temperatures [62], IRD [44], dinoflagellate assemblages [112], and sea-ice biomarkers [130]. The EHTM was followed by Neoglacial cooling, which witnessed the development of what we know as the preindustrial, perennial sea-ice-covered Arctic, culminating in the Little Ice Age (LIA, 1400–1900 C.E.). Temporally and spatially variable sea-ice cover throughout the Holocene is among the most notable discoveries of the last decade [170, 193] because it reflects an Arctic Ocean highly sensitive to insolation and unforced climate variability.

Similarly, high-resolution late Holocene records covering the last 1000–2000 years are particularly important because they provide baseline variability to interpret recent trends in sea ice and temperature. Terrestrial [40, 102], marine SST [178], and sea ice [104] proxies show natural climate variability during the late Holocene, including the Medieval Climate Anomaly (600–1400 C.E.) and the LIA, as well as anomalous 20th century patterns.

Arctic Ocean marine mammals

Marine mammals are a major component of modern Arctic sea-ice ecosystems [74, 105] and their molecular genetics and paleontology provide insights about past climate changes in the Arctic. The use of molecular sequences of DNA and proteins to infer species’ phylogeny and divergence times (i.e., a molecular clock) is an important aspect of phylogenetics [191]. These analyses, combined with vertebrate fossil evidence, can provide information about the temporal distribution of species, which can be used with paleoclimate data to better understand the Arctic climate-biological relationships, especially for vertebrate lineages (Supplementary Table 3). As we see below, molecular methods are increasingly applied to integrated paleoclimatic-ecosystem studies in the Arctic, so it is important to briefly consider the strengths and limitations.

The molecular approach involves comparison of the amino acid sequences of proteins or nucleic acid sequences (DNA or RNA) in different species [158, 191, 197, 209]. Molecular sequences will diverge by mutation from a common ancestral sequence at some rate, which is the time component of the “clock”. If the rate of sequence divergence is constant, then its extent will be a function of time and the phylogenetic relationships and time of divergence of the sequences can be estimated. If the time of divergence of the sequences is assumed to be equal to the time of divergence of the species, then an estimate of species’ divergence time is obtained. The assumptions of a constant rate of sequence divergence (depending on mutation rate and population genetic factors of selection, population size, migration) and that a sequence divergence reflects the species divergence are key factors affecting the accuracy of molecular clocks. Single gene sequences often do not reflect the species phylogeny so multiple genes or entire genome sequences are needed for robust analyses (e.g., [142]). DNA from extant animals is typically used to quantify sequence divergence, but ancient DNA (aDNA) from fossil material as old as 0.7 ma can also be used and provide valuable insights [168].

The accuracy of molecular clocks also depends on the accuracy of a fossil calibration date to identify the divergence time for at least one node of the phylogenetic tree of the taxa considered [7, 87, 124, 143]. Divergence time estimates can be controversial because of potential discrepancies of molecular clocks depending on the genes, calibration points, and models of molecular evolution considered [69, 158, 191, 197].

Case studies of vertebrate phylogeny with fossils and DNA sequences

In the case of Arctic climate change, the divergence time of polar bears (Ursus maritimus) and its sister species, brown bears (U. arctos), is especially relevant because there is concern about reduced summer sea ice habitat, especially for some geographic populations [3, 4, 54, 185]. Polar bears and brown bears are thought to have evolved from a common ancestor during the Pleistocene [111], and a polar bear fossil from the last interglacial (Eemian) period ~125 ka established this age as their minimum time of divergence [2, 88, 114].

Molecular clock estimates of the divergence time of polar bears and brown bears vary widely depending on the genes used. These include divergence times of 2–3 Ma using proteins [72], 110–130 ka with mitochondrial DNA (mtDNA, [8, 19, 46, 57, 109, 114, 189, 206, 207], 0.43–1.12 Ma with Y-chromosome DNA sequences [18] and 0.34–2.0 Ma with nuclear DNA sequences [57, 77, 206]. The most recent analyses of genome sequences estimated the polar bear-brown bear divergence at 340–480 ka [116], 1.2 Ma [23, 36], and 4–5 Ma [128].

Due to the inherent uncertainty of molecular clocks, some authors have refrained from applying them to these species [32, 67, 140, 199]. Cahill et al. [23] note that the molecular divergence times for bear species are relative, not absolute dates because of the uncertainty of the fossil record regarding bear species’ divergences. However, it is reasonable to infer the minimum age of U. maritimus is about 125 ka and more likely somewhat older, between 300 ka and 2 Ma. As discussed above, major climate transitions including the mid-Pleistocene Transition and mid-Brunhes Event occurred during this time frame.

Given the dynamic nature of climate-driven habitat changes outlined above, it is important to note that speciation may be accompanied by interbreeding between populations until there is permanent reproductive isolation. Extant populations of polar bears and brown bears have separate gene pools with minimal interbreeding [3436, 77, 144], but future interbreeding (i.e., hybridization) is hypothesized if sea-ice declines and polar bears spend more time on land [103]. Past interbreeding in these species is suggested by paraphyletic mtDNA phylogeny in which polar bears and brown bears from Admiralty, Baranof, and Chichagof islands (ABC) in southeast Alaska have haplotypes in a clade separate from other brown bears [32, 35]. In addition, polar bears and ABC brown bears share nuclear alleles [77, 116, 128], including <1 % of the autosomal genome and 6.5 % of the X-chromosome loci [23], but none of the Y-chromosome [18]. The pattern of genes shared by polar bears and ABC brown bears—maternally inherited mtDNA > X chromosome > autosomes > Y-chromosomes—is consistent with introgressive hybridization of male brown bears mating with female polar bears. This is hypothesized to have occurred about 12 ka when brown bears replaced polar bears during post-glacial colonization of the ABC islands [23].

Pinniped phylogenies also shed light on the development of the Arctic marine ecosystem. The pinnipeds, which include seals (Phocidae), sea lions (Otariidae), and walruses (Odobenidae), live in Arctic and subarctic seas with seasonal or perennial ice. Seals of the subfamily Phocinae (tribe Phocini) include three closely related genera in the northern hemisphere whose divergence has been estimated with fossil and molecular data relevant to our discussion. This includes the ringed seal (Pusa hispida), a primary prey of polar bears. The genus Pusa has a circumpolar Arctic distribution that in addition to P. hispida in the central Arctic includes Caspian seals (P. caspica) in the Caspian Sea, and Baikal seals (P. sibirica) in (freshwater) Lake Baikal, Siberia. Phoca includes the harbor seal (P. vitulina) in the temperate and subarctic northern hemisphere, and the spotted seal (P. largha) in the subarctic North Pacific Ocean. The gray seal (Halichoerus grypus) occurs in the North Atlantic Ocean.

However, seal classification is not definitive because of close relationships among various groups. For example, harbor seals and spotted seals are sometimes considered conspecific, and some taxonomists suggest that Pusa and Halichoerus could be reclassified as Phoca [45, 86]. This is reflected in equivalent mtDNA divergence (mean sequence divergence 3.36 %) of ringed seals, harbor seals, and gray seals, which has been used as a standard to calibrate a molecular clock for other taxa [7].

The fossil record shows that ringed seals occurred in the Arctic region during Quaternary interglacial and interstadial periods, including the eastern Beaufort Sea (~42 ka), Greenland (130 ka), and the Chukchi Sea (130 ka, [81, 162]). Phoca (harbor seal or spotted seal) fossils also occur in the Chukchi Sea (115–130 ka, [162]). This indicates that the oldest fossils of ringed seals and spotted/harbor seals in the Arctic are the same age as the oldest polar bear fossil from the Eemian (MIS 5) interglacial. Even though molecular clock estimates suggest a much older origin of polar bears, the fossil data provide a minimum estimate of their origin and that of ringed and harbor/spotted seals. This confirms that the bears and seals co-existed in the Arctic during MIS 5 and persisted until the present.

Molecular genetic data indicate that the Phocini radiated during the last 1–2 Ma. Analysis of 8935 bp of 16 nuclear genes and mtDNA indicates that Pusa and Phoca split 1.58 Ma; and within Phoca harbor seals and spotted seals split 0.4–1.3 Ma, and within Pusa ringed, Caspian, and Baikal seals split 0.7–1.8 Ma [68]. Analysis of 26,818 bp of 52 nuclear and mtDNA genes indicate Pusa and Phoca split 2.1 Ma; and within Phoca harbor seals and spotted seals split 1.1 Ma, and within Pusa ringed, Baikal, and Caspian seals split 2.0 Ma [86]. The differences in these estimates reflect the different genes and models used, but they also indicate that seal species, including ringed seals, probably existed over much of the Pleistocene and Holocene along with polar bears.

The walrus (Odobenus rosmarus) also lives in Arctic and subarctic sea-ice-covered regions. Two subspecies are generally recognized, the Atlantic walrus (O. r. rosmarus) in the central Canadian Arctic east to the Kara Sea and the Pacific walrus (O. r. divergens) in the Bering and Chukchi Seas. A population in the Laptev Sea is related to the Pacific walrus [63, 113]. The fossil record shows that the Odobenidae evolved in the mid-Miocene ~16–21 Ma [47] and O. rosmarus is the only extant species, although up to 14 genera and 20 species lived in the past [47, 81]. Odobenus rosmarus is thought to have migrated from the Atlantic to the Pacific about 600 ka [81]; walrus fossils in the Bering and Chukchi Seas date to about 130 ka, on Vancouver Island, British Columbia 70 ka Ma, and as far south as California ~270 ka [82].

Molecular clock estimates suggest the walrus family diverged from the sea lion family (Otariidae) about 15.1–18 Ma [68, 86]. There are no extant taxa for molecular clock comparison of walruses with other Odobenidae, but an estimate of divergence of the Atlantic and Pacific walrus can be made considering their mtDNA divergence of 1–1.6 % [33] and a rate of pinniped mtDNA evolution of 1.2 %/Ma [7]. These data suggest the Atlantic and Pacific subspecies split sometime between 83 and 133 ka, although there may have been gene flow between the oceans over this time considering the changes in sea-ice conditions described above.

Vertebrate range expansion and contraction during climate changes

Vertebrate paleontology often combined with paleoclimatic and/or molecular genetics provides key information about Arctic mammalian response to climate change. For example, Cooper et al. [29] recently analyzed genetic (13 events) and paleontological (18 events) megafaunal “transition events” for terrestrial taxa within the context of abrupt climate transitions including Dansgaard-Oeschger events identified in Greenland ice cores and Cariaco Basin sediments. They defined faunal transitions as geographically widespread or global extinctions, or invasions, of species or major clades. The bulk of the evidence indicated terrestrial vertebrates are affected by abrupt climate transitions.

In addition, there have been several studies in which polar bear evolution has been assessed in the context of orbital paleoclimate cycles over the past few million years [23, 46, 57, 77, 128]. If, as DNA and fossil evidence suggests, polar bears and their primary prey, ringed seals and other prey such as walruses, have existed for at least 125 ka and likely hundreds of thousands of years, then they experienced extreme climate conditions of glacial periods as well as partially or completely summer sea-ice-free interglacial periods (MIS 11, MIS 5 and the early Holocene). Microfossil proxy evidence for southward expansion of sea ice during glacial periods implies that vertebrate species that are dependent on sea ice habitat might have also migrated southward into the Nordic and Bering Sea-North Pacific regions.

Several lines of evidence support this idea of frequent geographically extensive range shifts, not only in terrestrial vertebrates [29], but sea ice based marine mammals as well. First, the close genetic relationships among bear species and among seal species discussed above, including evidence for hybridization, suggests dynamic population shifts. Moreover, large-scale range expansion during glacial periods is evident in the fossil record of vertebrates in extra-Arctic regions [81]. For example, the post-glacial Champlain Sea (13–9 ka, [159]) of New York, Vermont, and Canada has well-studied Arctic vertebrate faunas that include whales, walruses, brown bears and seals [64, 83]. Likewise, in coastal regions around Alaska, fossil records [31, 85] support molecular genetic data [23] showing that during the LGM, polar bears and ringed seals ranged as far south as the Gulf of Alaska, considerably south of their current Arctic ranges. In the case of summer sea-ice-free interglacial periods, the presence of winter sea ice habitat, polar bears’ ability to fast during summer [164], seals ability to use land areas in the absence of sea ice, and the availability of new prey species shifting ranges into the Arctic may have allowed survival during warm periods. Walrus also have an extensive glacial and post-glacial fossil record [55] including specimens from the paleo-Hudson River Valley on the New York and New Jersey continental shelf dated at ~10.6–11.2 ka [51].


The Cenozoic ecosystem changes in the Arctic described above are summarized in Figs. 5 and 6 within the context of climate changes over different timescales. Several conclusions can be made. First, a seasonally ice-free marginal and central Arctic Ocean was common not only during Greenhouse worlds of PETM and Early Eocene, but also during the Pliocene, the early Quaternary before the Mid-Pleistocene Transition, during MIS 11, MIS 5 and regionally during the early Holocene. During orbital climatic cycles of the last few hundred thousand years, interglacial periods were characterized by perennial and at times seasonal sea ice cover and inhabited by marine ecosystems similar to those of the pre-industrial Holocene. Some species thought to be dependent on summer sea ice (e.g., polar bears) survived through these periods. In contrast, during glacial periods the much smaller Arctic Ocean and much of the adjacent continents were covered with massive ice sheets, thick ice shelves, and sea ice making large regions virtually uninhabitable to most species that inhabit today’s Arctic. Despite the scale, frequency and rapidity of Quaternary climate changes, Arctic marine ecosystems associated with sea-ice habitats were extremely resilient, adapting through geographic range expansion into the Arctic during warm periods, and south into extra-Arctic regions during glacial periods. The stratigraphic record of the last 1.5 Ma indicates that no marine species’ extinction events occurred despite major climate oscillations. The Cenozoic sedimentary record is too incomplete to conclude that large climate transitions caused extinction of Arctic species, but hopefully future IODP coring will recover more complete records [182]. More generally, future cross-discipline studies of Arctic species and ecosystems combining molecular methods and paleoclimate reconstructions will result in a better understanding of how biological systems respond to climate changes.

Fig. 5
figure 5

Summary of Arctic Ocean biological and climatic events during the Cenozoic. Blue letters are marine mammal events, red are climatic events, green are biological events. See text and Supplementary Tables 1–3 and Supplementary references for sources

Fig. 6
figure 6

Summary of Arctic Ocean biological and climatic events during mid-to-late Quaternary orbital glacial-interglacial cycles. Blue letters are marine mammal events, red are climatic events, green are biological events. See text and Supplementary Tables 1–3 and Supplementary references for sources