Growth Increments of Coralline Red Alga Clathromorphum Compactum Capture Sea‐Ice Variability Links to Arctic and Atlantic Multidecadal Oscillations (1805–2015)

Given sea ice's importance in global climate regulation, fully understanding the role of natural temperature and atmospheric patterns like the Arctic Oscillation (AO), North Atlantic Oscillation (NAO) and Atlantic Multidecadal Oscillation (AMO) in its variability is critical. While instrumental AMO and reliable AO records are available since the mid‐1800s and 1958, respectively, satellite sea‐ice concentration data sets start only in 1979, limiting the shared timespan to study their interplay. Growth increments of the coralline algae, Clathromorphum compactum, can provide sea‐ice proxy information for years prior to 1979. We present a seasonal 210‐year algal record from Lancaster Sound in the Canadian Arctic Archipelago capturing low frequency AMO/NAO variability and high frequency interannual AO/NAO prior to 2000. We suggest that sea‐ice variability here is strongly coupled to these large‐scale climate processes, and that sea‐ice cover was greater and the AO more negative in the early and late 19th century compared to the 20th.

• Annual algal growth increment widths (sea ice cover proxy) correlate most strongly with summer Arctic Oscillation (AO) trends • Abrupt width increases in the early 1800s and 1920s capture sea-ice cover decreases in phase with or slightly lagging behind +AMO phases • The algal proxy record from Lancaster Sound captures +AO-related sea-ice export into the Canadian Arctic Archipelago Supporting Information: Supporting Information may be found in the online version of this article.
appeared to drive AO/NAO variability, although with an opposing pattern than that seen in the Canadian Arctic (Labe et al., 2019;X. Y-Yang et al., 2016; for an opposing study, also see Kolstad & Screen, 2019).Evidently, many factors control arctic sea-ice variability, yet the relative roles of natural and anthropogenic forces are still uncertain (Delworth et al., 2016).Further, reliable satellite sea-ice records are only available since the late 1970s and AO records prior to 1958 have many associated inconsistencies, challenging the ability to resolve how longterm natural climate patterns drive sea-ice variability.
In the absence of long instrumental records, tree-ring-or coral-based proxy records (Gray et al., 2004;Saenger et al., 2009), multi-proxy (terrestrial, ice core, lacustrine or coral archives: Mann et al., 1995) and modeled (Delworth & Mann, 2000) AMO records have attempted to clarify the periodicity of the AMO.Other studies have used historical and proxy records to study the interplay between AMO and sea-ice (Divine & Dick, 2006;Frankcombe et al., 2010;Macias-Fauria et al., 2010;Miles et al., 2014).Similar work has been accomplished with the AO, which have also used the above listed archives (D'Arrigo et al., 2003;Rimbu et al., 2001Rimbu et al., , 2003;;Sicre et al., 2014;Young et al., 2012), and deep-sea sediment cores (Darby et al., 2012).However, sediment cores typically provide lower-temporal resolution records than tree-ring, coral, ice-core, and lake varve records, while the latter archives have been unable to directly capture oceanic or regional variability north of the tree line.

Algal Data Preparation and Analysis Methods
Individual Clathromorphum compactum crusts were collected via SCUBA in 2016 at 18-20 m depths near Beechey Island, northwestern Lancaster Sound, Nunavut, Canada (74°42′54.46″N,91°47′29.35″W;Figure 1).Crusts were prepared into thick sections with an Isomet Precision Saw, ground and polished with a Struers Labopol polishing disk in 9, 3, and 1 μm steps, with ultrasonic bath immersion between steps.Thick sections were imaged with an Olympus VS-BX reflected light microscope paired to an automated stage, then stitched together with Geo.TS software.The three highest quality specimens (IDs: 2, 15, and 41) were selected for geochemical analysis (Figure 2).Specifically, annual cycles of Mg/Ca ratios were used to validate growth increments and their widths.Increment widths from these three specimens were averaged to produce the site's master CASIP record discussed throughout.
Geochemical data were obtained at the University of Toronto's Earth Science Center with a NWR 193 UC laser ablation inductively coupled mass spectrometry (LA-ICP-MS) system linked to an Agilent 7900 quadrupole mass spectrometer.Line scans were ablated (5 μm/second) along the growth axis, using an aperture size of 10 × 70 μm, and a 10 Hz laser pulse rate (see details in Hetzinger et al., 2011).Age models were constructed with crossdating methods (Leclerc et al., 2022;Text S1 in Supporting Information S1).In addition, prior to 1880, only sample 41, providing the longest continuous chronology, was used to extend the record back to 1805.While an additional intra-specimen transect was used to validate the early sections of the age model, we acknowledge that the lack of multiple crossdated specimens for the period of 1805-1880 has associated uncertainties.

Instrumental Data and Statistics
Correlation analysis (linear regression) determined the relationship between the algal record and instrumental indices.Monthly AO index values are based on instrumental sea level pressure (SLP: Poleward of 20°N calculated  (Thompson & Wallace, 2000).
Monthly Hurrell North Atlantic Oscillation (NAO) index values are based on principal component analysis of SLP over the Atlantic.While the instrumental AO index goes as far back as 1899, early data issues include different SLP sources for different time periods, with discontinuities identified between data source transitions (Trenberth & Paolino, 1980).Therefore, only later instrumental AO index values (1958AO index values ( -2015) ) were used in this study due to confidence issues with early data points.Further, the NAO record (NCAR, 2003) was shortened to match the length of the AO record for even comparison to the algal record in Table 1.The correlation between the CASIP record and the full-length NAO record is reported and plotted in Figure 4. Monthly AMO index values are the 10-year running mean values smoothed from the Kaplan Extended SST V2 timeseries provided by NOAA PSL, Boulder, Colorado, USA (Kaplan et al., 1998).Seasonal means were calculated by averaging summer months (May-Oct).Spatial correlation analysis and linear regression to monthly NSIDC sea-ice concentration data set (see procedure in Leclerc et al., 2021) was computed using Matlab and m_map mapping toolbox (Pawlowicz, 2020).The software kSpectra (SpectraWorks, 2020; methods in Ghil et al., 2002) was used to run multi-taper and singular spectral analyses (SSA) on instrumental and proxy data sets to determine if the algal record shared AO, NAO and AMO frequency signatures.RStudio's Astrochron package (surrogatecor function: Meyers, 2014) was used to estimate the significance of serially correlated data such as decadally smoothed AMO and 10-year running averages, as these are prone to autocorrelation.

Results and Discussion
Since higher sea-ice cover, in typically colder years, limits growth, we expected a negative correlation between regional sea-ice cover and annual growth, and positive correlations with AO, NAO, and AMO.Accordingly, spatial correlation analysis shows strongly significant negative correlations (p < 0.001) between the Beechey Island growth increment chronology and regional satellite sea-ice concentrations (Figure 3).At a localized scale, the algal growth increment timeseries correlates significantly (R = −0.71;p < 0.001) with satellite sea-ice concentrations (Leclerc et al., 2022) (Figure 3b).The confirmation of the local sea-ice-algal growth relationship suggests that if AMO, AO or NAO and sea-ice are related in Lancaster Sound, the algal timeseries should record their signal.Indeed, the AO was significantly correlated at an annual resolution (p < 0.01), and the correlation to the NAO was markedly strong at decadal resolution (MCp = 0.02), however only until 2000 (Table 1, see Table S1 in Supporting Information S1 for other seasonal comparisons).There are also marked similarities between the Lancaster Sound, Labrador (Moore et al., 2017) and Spitsbergen (Hetzinger et al., 2019b) CASIP records, especially during the Early Twentieth Century Warming period (ETCW; Figure 4).
A lack of correlation between AO and sea-ice cover in recent decades has previously been documented (Feldstein, 2002;Overland & Wang, 2005, 2010;Stroeve et al., 2011) and is also manifested in the Beechey Island CASIP record after 2000.The lack of an AO to sea-ice correlation in the Canadian Arctic Archipelago (CAA) may also be related to recent shifts in the duration of ice bridges, landfast ice between landmasses which form in winter and block sea-ice export until they collapse in the summer.When ice bridges at M'Clure Strait or the Queen Elizabeth Islands (QEI) (Figure 1) collapse, sea-ice from the Arctic Ocean is free to import into the CAA, especially during positive AO phases (Howell et al., 2013).Contrary to the +AO-stimulated ice breakup/export acceleration, +AO-stimulated sea-ice import after ice bridge collapse may limit algal light access and mute the AO signal.In fact, since 2005 there has been an increase of ice inflow into the CAA through the Queen Elizabeth Islands, which tends to flow south toward Lancaster Sound (Howell et al., 2013).Other data from Nares Strait suggest that ice volume export through the Strait has increased recently in comparison to the 1997-2009 mean, linked to the trend of shorter duration of ice bridges (Moore et al., 2021).This may be responsible for the masked AO signal in the Beechey Island CASIP record since the turn of the millennium (Figure S1 in Supporting Information S1; Table 1).
Periods with larger growth increments coincide with a strongly positive AMO period in the mid-1900s, which followed the Early Twentieth Century Warming trend (ETCW: 1920s-1930s) (Figure 4).The ETCW has been shown to be associated with sea-ice retreat in the Barents Sea caused by stronger westerly winds between Spitsbergen and Norway (Bengtsson et al., 2004), and has also been recorded by C. compactum growth-Mg/Ca anomaly timeseries from Spitsbergen and Labrador (eastern Canada, subarctic) (Figure 4).Day et al. (2012) found that the recent +AMO phase could explain 5%-30% of satellite summer sea-ice loss and Miles et al. ( 2014) suggested AMO was a major driver of sea-ice variability from the past 800 years to the 1990s.Similarly, our data indicates that the AMO and ETCW affected ice decline in Lancaster Sound in the mid-20th century.Multi-taper spectral analysis results showed multidecadal variability in the algal chronology (significant at 99% level, 60-77-year signal, CASIP: 1805-2015; Figure S1 in Supporting Information S1), comparable to the posited periodicity of AMO (60-80 years) (Kerr, 2000;Schlesinger & Ramankutty, 1994).Significant (95% level) interannual signals (at 2 and 3 years) were also found, closely matching AO signatures (Figure S1 in Supporting Information S1) previously shown to affect sea-ice circulation in the Baltic Sea (Jevrejeva et al., 2003).
While the CASIP multi-taper results did not capture AO's decadal variability as reported elsewhere (Ramos da Silva & Avissar, 2005), the singular spectrum analysis (SSA) of the shortened CASIP record  identified significant variability at 7.6-10.3years responsible for more than 60% of variance (Figure S1 in Supporting Information S1).In the AO SUMMER record , most of the variability is interannual (2.5-5.1 years; details Text S2 in Supporting Information S1); a decadal signal (10.6-year) is explaining only 16.9% of total variance.In summary, multi-taper results did not exhibit the 8-10-year AO signals previously identified through wavelet power spectrum analysis (Ramos da Silva & Avissar, 2005).However, the shared variability at the approximately 2-3-year periodicity level seems to be what the sea-ice-AO and sea-ice-CASIP relationships are recording in the CAA.
The algal record prior to the 1900s suggests colder and heavier ice conditions in the 19 th century in comparison to the 20th century similar to the findings of indirect (temperature) sea-ice proxy tree ring records (D'Arrigo et al., 2003;Young et al., 2012).The algal chronology also suggests a period of less ice in the mid-1800s possibly due to more positive AO/NAO or AMO, or both (Figure 4).While, many have suggested that the Little Ice Age and colder conditions persisted until the late 1800s, this slightly warmer period in the mid-1800s is supported by multiple Arctic proxy records that find episodic warming at this time (Jennings & Weiner, 1996;Massé et al., 2008: records synthesized in Miles et al., 2020).This warming period is also corroborated by ice cap stratigraphy from nearby Devon Island, Greenland ice sheets and marine cores from the Labrador Sea, which suggested early warming in 1860s and a more intense warming trend beginning around 1890 (Keigwin et al., 2003;Koerner, 1977;Trusel et al., 2018).The mid-1800s mild warming period found in our record predates those found in other AMO proxy records from terrestrial archives (e.g., Gray et al., 2004), which shows a warming period later in the 1800s, and cooler 1830s-1840s (Figure 4).While there are justified reservations related to using terrestrial proxy records to understand past sea-ice variability, it is notable that sea-ice and NAO trends have been shown to lag behind AMO variability in some regions, and that the timing in AMO peaks and troughs are regionally variable (Alexander et al., 2014;Peterson et al., 2015).As the NAO and AO are highly correlated (Rigor et al., 2002), this could also apply to the AO-AMO relationship.
CASIP records are indicators of a combination of sea-ice variables affecting light penetration to the benthos: present/absent ice cover (related to melt/freeze up and wind and current dynamics), seasonal duration of ice cover, ice thickness and snow cover.Together, the AMO, AO, and NAO have the capacity of affecting all these variables.Samelson et al. (2006) suggested that the formation of land-fast ice in the CAA is controlled by both winds and air temperature, both are parameters influenced by these large atmospheric and ocean temperature patterns.Furthermore, Peterson et al. (2012) found that monthly longshore wind anomalies in the Beaufort Sea, which are heavily influenced by AO, stimulated 43% of Lancaster Sound's volume transport anomaly variance.This is supported by the significant relationship between the Beechey Island CASIP record and gridded sea-ice concentrations on the exterior CAA coast bordering the Beaufort Sea (Figure 3a).

Summary and Conclusion
The C. compactum growth increment chronology from Beechey Island recorded: The development of longer high-resolution proxy records such as CASIP timeseries is critical to understanding the role of cryospheric-atmospheric feedbacks in the many intertwined components of the global climate system (Gao et al., 2015).The Canadian Arctic Archipelago, which tends to trap multi-year ice (Howell et al., 2008;Kwok, 2015), makes up a significant part of the Last Ice Area, predicted to be the last arctic region to experience summer sea-ice cover (Moore et al., 2019).As this area will become increasingly crucial in the coming years, and potentially more hazardous for naval travel (Howell et al., 2022), C. compactum CASIP records can provide important historical and pre-industrial baselines.While it is reasonably well understood that atmospheric patterns have an effect on sea-ice extent, the interplay between coastal sea-ice cover and atmospheric patterns, especially in the CAA remains relatively unexplored.Here we find moderate links between internal variability and sea-ice trends.However, we note that these links are muted in recent decades (especially after 2000) possibly due to anthropogenic forcing and enhancement of ice flow through QEI gates in the Canadian Arctic Archipelago (Howell et al., 2023).

Data Availability Statement
The Beechey Island, NU, and Kingitok Islands, NL CASIP records (Leclerc et al., 2023;Moore et al., 2017) and proxy AMO record (Gray et al., 2004) are accessible through the NOAA National Centers for Environmental Information (NCEI) Paleoclimatology Data Repository.The Spitsbergen CASIP record is available through the GSA Data Repository (Hetzinger et al., 2019b) Environmental data sets include: The monthly AO Index values were extracted from KMNI Climate Explorer (Thompson and Wallace, 2000).Monthly Hurrell North Atlantic Oscillation (NAO) index values were extracted from NCAR Climate Data Guide (NCAR, 2003).Monthly AMO 10-year running mean values smoothed from the Kaplan SST V2 were extracted from NOAA PSL1 (Kaplan AO pattern on SLP anomalies) computed through the National Centers for Environmental Prediction-National Center for Atmospheric Research (NCEP/NCAR) reanalysis

Figure 1 .
Figure 1.Representation of an Arctic Oscillation (AO) negative phase in the Arctic.Beaufort High (BH; orange); Icelandic Low (IL; light blue); Queen Elizabeth Islands (QEI: green); M'Clure Strait (purple); Beechey Island collection site (yellow dot); Lancaster Sound (yellow region).Negative AO phases promote a clockwise circulation of the Beaufort Gyre and stronger BH sea level pressure promoting counter-clockwise gyre circulation and ice convergence.The opposite holds true for positive phases.Ocean circulation shown as red arrows (based on Figure 3.29 in AMAP, 1998) and length of the ice-on season as white to dark blue gradient (1979-2015 mean days with >15% SIC: sourced from NSIDC(Meier et al., 2017).

Figure 2 .
Figure 2. Overview (left) and magnified (right) images of C. compactum crusts from Beechey Island, Lancaster Sound.Laser ablation paths used along axis of growth indicated in red.Sample IDs shown in upper left corner.
(a) higher sea-ice cover during the 1800s in comparison to the 1900s; (b) slightly lighter sea-ice years in the mid-1800s; (c) the Early Twentieth Century Warming period; (d) significant sea-ice responses to AO/NAO from 1960 to 2000; (e) signal frequency links to AMO and; (f) lack of sea-ice response to AO/NAO from 2000 to 2015 possibly due to external factors such as the greenhouse gas (GHG) effect, ice-albedo feedbacks and shifting ice bridge behavior.

Figure 3 .
Figure 3. (a) Spatial correlation analysis between gridded Arctic SIC and Beechey Island growth increment chronology.Right plot shows Beechey Island region enlarged.(b) Plotted algal growth increment timeseries (black: anomalies = (annual value − average)/standard deviation) and NSIDC sea-ice concentrations (blue: 75 km 2 around Beechey Island site) (see Leclerc et al., 2022 for original figure of subplot B).Growth anomalies are plotted inversely.

Table 1
Pearson's Correlations (R-and p-Values) of Beechey Island Algal Growth Record to Climate Indices at Seasonal (Summer) and Decadal (10-Year Running Means of Summer Values) Resolutions