Arctic sea-ice decline archived by multicentury annual-resolution record from crustose coralline algal proxy

Sea ice plays an important role in the global climate system because it influences albedo, heat and gas exchange, freshwater budget, ocean stratification, and deep water mass formation, among other ocean characteristics (1, 2). Presently, there is great concern that with ongoing warming, Arctic sea-ice decline is accelerating at an unprecedented pace not seen in at least the past 1,450 y (3). Sea-ice cover (SIC) is a sensitive parameter characterized by high variability in space and time (1) and analysis of sea-ice extent on a regional basis reveals that positive and negative anomalies in different sectors of the Arctic often occur simultaneously (4). Early 20th century Northern Hemisphere warming and sea-ice decline were mainly concentrated on the Atlantic portion of the Arctic with focal points in the Labrador Sea and Greenland, whereas mid–late 20th century sea-ice loss was most evident in the Russian Arctic (5–7). Modeling studies indicate that this differential sea-ice decline has likely been caused by Atlantic inflow variability into the Arctic (5, 8) and at least in part driven by the North Atlantic Oscillation (NAO) (7).

However, the SIC–atmosphere system is especially difficult to model with only a short observational record available. Instrumental data from satellite observations have been recorded only since the late 1970s when anthropogenic effects were likely already overprinting the natural climate system. Longer-term data from historical observations are far less dependable, because coverage is patchy and reliability sometimes uncertain. For a better understanding of long-term sea-ice variability, both in space and time, a network of high-resolution multicentury or millennial-scale sea-ice data derived from proxy records are needed (3, 9). High-resolution terrestrial archives such as tree rings, varved lake sediments, and ice-core records have been used to document the atmospheric forcing of, or atmospheric response to, changes in sea-ice cover (3). Unfortunately, the relationship of terrestrial records with sea ice is not direct and marine SIC proxy records are therefore preferable. Existing marine SIC proxy records are derived from biological, chemical, or sedimentological tracers preserved in marine sediment cores. Most sediment cores, however, suffer from poor temporal resolution, generally not exceeding decadal or even centennial scales (1, 10). Hence, to date, marine proxy sea-ice records are not able to resolve high-frequency climate oscillations during the past millennium such as the NAO that are crucial for understanding recent dramatic changes in SIC.

The yearly increment-forming photoautotrophic crustose coralline alga Clathromorphum compactum, which is widely distributed across Subarctic and Arctic coastal areas of the Northern Hemisphere (11), provides much needed annually resolved in situ sea-ice records spanning more than half a millennium. Clathromorphum has been calibrated as a proxy archive (12) and was subsequently used to generate multicentury-scale climate records from extratropical oceans (13). Photosynthetic Clathromorphum buildups have a fixed mode of life on rocky substrate and can therefore monitor and record light reaching the shallow seafloor in their high-Mg calcite skeleton. Life spans of individual plants are only limited by physical disturbance events on the seafloor, such as extreme grazing, seafloor instability, or iceberg scouring (11). Here we report on recently discovered living specimens in the Arctic with life spans of up to 646 y—the oldest continuously growing crustose coralline algae known to date. Annual growth rates of Clathromorphum are dependent on both light and temperature (14, 15) and decline slowly with temperature from 400 μm/y near the southern limit of distribution of C. compactum at 43°N to 240 µm/y at 52°N, the beginning of significant sea ice (Fig. S1). From that point northwards, even though temperatures continue to decline slowly, growth decreases sharply to 61 μm/y at 73°N. Growth north of 52°N in the western Labrador Sea and Baffin Island region is therefore related to declining solar insolation on the seafloor due to snow-covered sea ice. During the ice-free period, photosynthesis in this species is highly efficient and excess photosynthate is stored (11). With time under sea ice, the stored photosynthates are exhausted, limiting annual growth rates. In addition, water temperatures, which are uniform near −1.8 °C below sea ice under normal marine conditions (16), rise only during the ice-free period. Hence, the length of the annual open water interval, allowing atmospheric and solar heating of surface waters, controls the number of days when sea surface temperatures (SSTs) are above the sea-ice minimum. At the same time, this also determines the number of days available for significant algal growth to take place. The annual SST pattern in turn is preserved as cycles of Mg/Ca ratios in C. compactum. In the fall, recording of Mg/Ca ratios stops when growth ceases after the onset of SIC (as reflected by the narrow V-shaped character of the annual Mg/Ca cycle; Fig. S2) and only resumes when algal calcification is reinitiated following sea-ice retreat. Hence, longer ice-free periods, which also allow for more extensive summer sea surface warming, are reflected by higher annually averaged Mg/Ca ratios. Thus, the two variables, growth and annual Mg/Ca ratios, are both proxies for the duration of SIC. Combining both proxies yields a powerful means for SIC reconstructions.

We generated individual algal proxy records by averaging normalized annual growth and Mg/Ca records from live-collected specimens from an Arctic locality characterized by an ice-free period lasting from approximately June to September and a Subarctic locality closer to the southern limit of winter sea-ice extent with an average ice-free season from May through November (Fig. 1). Algal records from both locations are significantly correlated to observational data of SIC from both satellite and ice-chart records from the 1970s onwards (Fig. 2A). At the Arctic (Subarctic) locality, up to 39% (38%) of the algal signal (1982–2007) is explained by National Snow and Ice Data Center (NSIDC)-derived SIC, and only 6% (23%) by satellite SST (21). Hence, ice cover is the dominant component driving the algal records. Much of the remaining variability can be attributed to local, as opposed to regional, temperature and particularly ice conditions distinguishing the algal records from the spatially averaged temperature and sea ice observational data (Fig. 1 shows regions used for proxy verification). An averaged record of three individual algal time series from both locations (Fig. S3) yields a long-term relationship to observational SIC data off eastern Canada (19) and the instrumental wintertime NAO for the past 150 y (20) (Fig. 2B). Modeling studies have shown that the NAO exerts an influence on the spatial distribution of winter sea ice via wind-driven anomalies of sea-ice velocity, surface vertical heat flux, and possibly horizontal oceanic heat flux (7). There is strong observational evidence connecting Arctic sea-ice distribution with the positive NAO trend from the 1960s to the early 1990s (4). At this time SIC over the Labrador Sea region and eastern Canadian Subarctic was anomalously high due to wind-driven equatorward advection of ice (7, 22), which is reflected by an interval of low growth rates and Mg/Ca ratios in the algal record (Fig. 2B). In contrast, the most dramatic increase in algal growth and Mg/Ca anomalies took place during a negative NAO phase starting in the 1920s (Fig. 2B) accompanied by strong Arctic warming and sea-ice retreat (5). An accelerated sea-ice decline starting in the 1990s is evident from both the algal proxy time series and observational data (19).

A 646-y long algal reconstruction of sea ice emphasizes the expression of the Little Ice Age (LIA) off eastern Subarctic and Arctic Canada, which is characterized by negative algal anomalies between 1530 and 1860 pointing to extensive SIC (Fig. 3). This is consistent with low Arctic temperatures (23), an increase in SIC as observed in historical accounts from Iceland and the Fram Strait (24, 25), and shown by ice core-based proxy reconstructions from Svalbard and Devon Island (18, 26). In addition, a southward migration of Thule/Labrador Inuit from Hamilton Inlet (54°N) to the northern Gulf of Saint Lawrence during the period 1500–1650 coincided with the LIA-related ice expansion that began in the early 16th century (Fig. 3). That expansion is evidenced by an increase in ice-sensitive harp seal remains found in Labrador middens and the dominance of migratory harp seals in pioneer Inuit settlement middens in the northeastern Gulf of Saint Lawrence (27). Subsistence hunting methods of the Inuit relied on ice access and may have therefore strongly benefited from the regional increase in SIC. In contrast, an overall decrease in Arctic-wide ice cover between the late 15th and early 17th century has been observed in another multiproxy-based study (3). The authors explain the declining SIC during the first half of the LIA by increased advection of warm and saline water into the eastern Arctic Ocean from the North Atlantic, a mechanism that is commonly regarded as a key element of preindustrial Arctic SIC (8). Differences between the Arctic-wide observations and the northeastern Canadian algal SIC record indicate that the regional variability in sea ice observed in recent decades has already existed during the LIA and highlight the asynchrony of the LIA in distinct regions globally (4, 28).

Temporal variability in the algal record can be divided into three time slices (i): the pre-LIA portion (1365–1530), exhibiting singular spectrum analysis (SSA) and multitaper method (MTM) spectra with dominant frequencies centered between 43 and 25 y (95% significant), which explain 33% of the variance and frequencies between 3.5 and 5 y (11% of the variance) (Fig. S4); (ii) the LIA (1530–1860) is characterized by a combined high-frequency signal at 8 and 5.6 (99% significance, explaining 20% variance) and 37–38 y (95% significance, explaining 15%); and (iii) post-1860 multidecadal variability returns to a dominant 45- to 83-y signal (99% significance, explaining 26% of the variance). Thus, whereas multidecadal frequencies dominate the intervals of low SIC, the LIA is characterized by high-frequency variability in sea ice. A more unstable decadal-scale climate and higher-frequency sea-ice variability appear to be a feature of a cold background climate, compared with pronounced multidecadal variability during warmer intervals (29, 30). Hence, with future warming we can expect a continuation of multidecadal sea-ice variability that has dominated the Arctic since 1860. During these past 150 y, the algae have witnessed a sea-ice decline unprecedented in at least the past 646 y.