Evaluation of reconstructed sea surface temperatures based on Uk 0 37 from sediment surface samples of the North Paci fi c

The alkenone unsaturation index (Uk 0 37) as proxy for sea surface temperature (SST) is an important tool in paleoclimatology for reconstructing past ocean temperature variability. Typically, Uk 0 37 recorded in marine surface sediments shows a linear correlation with modern mean annual SST. However, in high-latitude oceanic regions, such as the subpolar Pacific, Uk 0 37-based SSTs do overestimate the mean annual temperature by up to 6 C, potentially leading to obscured paleoclimatic information drawn from stratigraphic Uk 0 37-records. The reason for this “warm bias” is still not well understood. Here, we present a compilation of 97 sediment surface samples from Multicores collected in the Bering Sea, the Okhotsk Sea and the North Pacific to evaluate the alkenone-temperature proxy against observational data from the North Pacific. Sediment surface samples were analysed for alkenones and the derived Uk 0 37-indices converted to water temperatures using different calibration equations established in the literature. Uk 0 37-based SSTs were then compared to instrumental SST data, as well as modern alkenone flux data from sediment traps in the North Pacific. Our results confirm that most Uk 0 37-based SSTs from the subpolar Pacific are 2e6 C too warm compared to instrumental mean annual SSTs for calibrations applied. However, with an uncertainty at the level of ±1.5 C or less reconstructed SSTs fit quite well to modern autumn temperatures north of the Subarctic Front (SAF), when maximum export flux of alkenones to the seafloor is indicated by sediment trap data. South of the SAF, reconstructed SSTs largely mimic the modern mean annual SST signal with an uncertainty of ±1.5 C or less, which is likely due to the attenuation of seasonality and longer growth season of coccolithophorids according to sediment trap data. Our study further demonstrates that Uk 0 37, when seasonality in alkenone production and export are known and considered, is able to provide reasonable estimates of SSTs in modern high-latitude ocean settings. We conduct a case study using available alkenone time-series derived from a sediment core collected from the south-western Okhotsk Sea to better understand the potential effect of seasonality in alkenone production on stratigraphic Uk 0 37-record in the subpolar Pacific. The case study from the Okhotsk Sea indicates that even a small shift in seasonality may lead to strongly biased SSTs with broader regional implications for paleoclimate reconstructions in high-latitude ocean settings. © 2020 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). Environmental Sciences and ener Str. 8, 28334, Bremen, ier Ltd. This is an open access artic

Several calibration studies between alkenone indices and SST using other algal culture, water-column particles and marine sediments further confirmed the relationship between the U k 0 37 index and SST (e.g. Sikes et al., 1991;Rosell-Mel e et al., 1995;Sikes et al., 1997;Sonzogni et al., 1997;Conte et al., 1998;Rosell-Mel e, 1998). Müller et al. (1998) were among the first to provide a global coretop calibration of U k 0 37 derived from 370 marine surface samples spanning the world ocean between 60 N and 60 S. The global coretop calibration shows a remarkably strong linear correlation with mean annual SST (U k 0 37 ¼ 0.033T þ 0.044; r 2 ¼ 0.958), but more importantly is virtually identical to the initial U k 0 37 calibration based on laboratory culture studies (Prahl and Wakeham, 1987;Prahl et al., 1988), attesting to the general applicability of alkenones to reconstruct past SST from marine sediments.
Although the strong statistical relationship between U k 0 37 and mean annual SST holds in many regions of the world ocean, several studies reported apparent discrepancies between reconstructed SST and mean annual SST, for instance in subpolar ocean regions (e.g. Sikes et al., 1997;Rosell-Mel e et al., 2000;Prahl et al., 2010;M eheust et al., 2013). For the subpolar Pacific, alkenone-derived SST estimates using the global core-top calibration of U k 0 37 notoriously overestimate (warm bias) modern mean annual SST (Prahl et al., 2010;M eheust et al., 2013). Prahl et al. (2010) reported a significant and systematic offset between alkenone SSTs and mean annual SST from Multicores of SE Alaska (55e61 N), where reconstructed SSTs were up to 4 C higher than the instrumental mean annual SSTs. One factor that may obscure U k 0 37 -based SSTs estimates is redox-dependent, compound-selective degradation of C 37 alkenones that can lead to significant alteration of the U k 0 37 -index after deposition (e.g. Hoefs et al., 1998;Gong and Hollander, 1999;Rontani et al., 2008;Rontani et al., 2013). Prahl et al. (2010) looked for epoxides indicative of selective alkenone degradation to further understand the conspicuous SSTs from the U k 0 37 -signal. In SE Alaskan sediments, selective alkenone degradation by aerobic bacterial processes were identified leading to an average "warming" effect of~1.4 C. However, it has been shown that even after adjustment of this potential diagenetic "warming" effect, U k 0 37 -based SSTs are still too warm and do not reflect the mean annual SST (Prahl et al., 2010). A similar systematic warm bias in U k 0 37 and mean annual SST has been reported from analysis of surface sediments in the Bering Sea and subpolar Northeast Pacific (M eheust et al., 2013).
In general, environmental factors such as non-thermal physiological stress to e.g. changes in nutrient concentrations may also affect to a variable amount the U k 0 37 -signal preserved in marine sediments (e.g. Epstein et al., 1998;Yamamoto et al., 2000). Based on culturing experiments of E. huxleyi it has been shown that U k 0 37 -values also vary with nutrient availability and growth stage (Epstein et al., 1998). However, Herbert (2003) argued that, in most cases, these factors produce errors at the level of 1.5 C or less that are not larger than the mean standard error of the entire regression for the global ocean calibration. Lateral transport of allochtonous alkenones is another factor that may obscure U k 0 37 values and has been proposed to explain anomalous SSTs in the high-latitude North Atlantic (Rosell-Mel e et al., 2000). Hence, all these different factors may randomly affect the preserved U k 0 37 -signal by a variable amount depending upon regional oceanographic conditions, however, do not fully explain the non-random and systematic departure of the U k 0 37 proxy from mean annual SST that has been reported from the subpolar Pacific.
The seasonal production and export of alkenones is another particularly important aspect in subpolar ocean settings (e.g. Sikes et al., 1997;Harada et al., 2006a;Seki et al., 2007). For the Southern Ocean, Sikes et al. (1997) combined water column data with sedimentary temperature estimates and concluded that alkenone distributions are seasonally biased towards summer, and reconstructed SSTs better fit to times when alkenone flux to the sea floor is high. How seasonality of coccolithophorid blooms (alkenone flux) in the subpolar Pacific is related (or not) to the U k 0 37 -temperature signal recorded in marine sediments is still not evaluated. Given the importance of U k 0 37 index in paleoclimatology further evaluation of the "alkenone thermometer" from regions where non-random, regional variations in the core-top calibration of U k 0 37 -temperature vs mean annual SST exists are essential to enhance the quality of paleoclimatic interpretations drawn from such records (Rosell-Mel e and Prahl, 2013).
Here, we report results from a compilation of 97 sediment surface samples from Multicores of the Bering Sea, the Okhotsk Sea and the North Pacific to further evaluate the U k 0 37 index as proxy for SST. This compilation consists of 42 new surface sediment samples from the subpolar Pacific analysed in this study. Additional data comes from published biomarker studies of Multicores recovered from coastal and marine sites throughout SE Alaska (16 samples, Prahl et al., 2010) and from R/V Sonne expedition SO202 to the eastern Bering Sea and North Pacific (39 samples, Gersonde et al., 2012;M eheust et al., 2013). Sediment surface samples were analysed for alkenones and the U k 0 37 -index converted to water temperatures using different calibration equations established in the literature. In a next step, instrumental SST data and published modern alkenone flux data from sediment traps are compared to further evaluate reconstructed alkenone temperatures against observational SST data of the North Pacific.

Surface currents
The surface circulation of the North Pacific follows a cyclonic circulation pattern that governs the rate of heat and nutrient exchange between different ocean regions of the study area (Fig. 1). The subarctic front (SAF) separates the study area into a northern (subpolar) and southern (transitional) zone of the North Pacific. Study sites south of the SAF are situated within the Kuroshio e North Pacific Current system. Intensity of the eastewest thermal fronts (the SAF and the Subarctic Boundary) are very pronounced, in particular on the western part of the North Pacific (Roden et al., 1982). North of the SAF in the subpolar Pacific, the main circulation regime consists of two prominent counter-clockwise circulating systems, the Alaska Gyre in the Northeast Pacific and the Western Subarctic Gyre (WSAG). The Alaskan Stream, a northern boundary current, carries relatively warm water masses from the Alaska Gyre along the Aleutian Island Arc into the WSAG, thereby flowing into the Bering Sea through several passes between the Aleutian Islands ( Fig. 1). In the Bering Sea, the surface currents of the Aleutian North Slope Current (ANSC) and Bering Slope Current (BSC) describe a large-scale counter-clockwise surface circulation. Surface waters leave the Bering Sea through the Kamchatka Strait back into the North Pacific via the East Kamchatka Current (EKC) and, to some extent, via Bering Strait into the Arctic Ocean (Stabeno et al., 1999). A part of the EKC enters the Okhotsk Sea through the Kurile Island Arc, thereby influencing surface waters of the Okhotsk Gyre (Fig. 1). The Kamchatka Current (KAC) describes the inflow into the Okhotsk Sea, the East Sakhalin Current (ESC) on the western flank of the Okhotsk Sea transports waters back to the North Pacific. Further to the south, the Oyashio Current (OC) develops and brings cold and nutrient-rich waters back to the North Pacific. The confluence of the Kuroshio Current (KC) that carries warm and saline waters to the north and cold and nutrient-rich OC happens next to the southern tip of Hokkaido.

Sea surface temperatures, sea-ice variability and marine productivity
The high latitudes of the North Pacific are characterized by pronounced seasonal variability in SST and sea-ice distribution, which is tightly coupled to the interplay between the Siberian High and Aleutian Low pressure systems (e.g. Niebauer et al., 1998). During winter, the Aleutian Low strengthens and moves southeastward, thereby leading to advection of cold air masses from the Arctic to the subpolar Pacific (Sekine, 1988) that causes strong sea surface cooling, intense winter mixing of nutrient-rich subsurface waters and the expansion of sea-ice in the Bering Sea and Okhotsk Sea. During this time interval sea-ice covers up to 75% of the Okhotsk Sea from winter to spring (Yang and Honjo, 1996). In contrast, winter sea-ice cover in the Bering Sea is rather moderate and confined to the shelf areas, thus most of the Bering Sea stays ice-free the entire year. During the summer months, increased insolation and weakening of the Aleutian Low lead to warmer SSTs, strong upper ocean stratification and ice-free conditions in the marginal seas (Ohtani et al., 1972). The seasonal temperature difference is quite pronounced in the study area, which typically ranges from 0 to 2 C in winter to 8e10 C in summer north of the SAF and from 10 to 15 C in winter to 20e25 C during summer months south of the SAF (Locarnini et al., 2006).
In the subpolar Pacific, marine productivity and export flux patterns to the ocean interior are intimately linked to seasonal variations in meteorological and physical oceanographic conditions. (Honda et al., 2002;Harada et al., 2003Harada et al., , 2006aNakatsuka et al., 2004;Seki et al., 2007;Tsutsui et al., 2016). The seasonal pattern of export production in the marginal seas follows the seasonal sea-ice pattern because phytoplankton production is heavily depressed during the sea-ice period (e.g. Honda et al., 2002). Results from sediment trap time-series of the subpolar Northwest Pacific and Okhotsk Sea show a first maximum in biogenic productivity that occurs during spring to early summer and is related to extensive diatom blooms (Honda et al., 2002;Seki et al., 2007). In the subpolar Northwest Pacific blooms of E. huxleyi typically occur from late summer to autumn with maximum flux of alkenones observed during OctobereNovember (Harada et al., 2006a). In the Bering Sea and Okhotsk Sea E. huxleyi blooms are restricted to the autumn season (SeptembereNovember), suggesting that a large proportion of alkenones are synthesized during limited periods (Harada et al., 2003;Seki et al., 2007;Tsutsui et al., 2016). Sediment trap data south of the SAF in the mid-latitude Northwest Pacific show extended blooms of E. huxleyi that rise from the beginning of spring (March) and remain high until autumn (November) (Yamamoto et al., 2007).
All sample material was freeze-dried and homogenized prior to biomarker analysis. For each lipid analysis, 1e5 g of sediment were extracted with an accelerated solvent extractor (ASE-200, Dionex) at 100 C and 1000 psi for 15 min by using dichloromethane (DCM) as solvent. Remaining extracts were separated by silica gel column chromatography into three sub-fractions with the following mixture of solvents: fraction 1, 5 ml hexane; fraction 2, a mixture of 5 ml DCM\hexane (1:1); fraction 3, 5 ml DCM. Alkenones were eluted in the third fraction and prepared in 100 ml hexane. In a few cases with lower alkenone abundance, sediment extracts were further concentrated in 20 ml hexane to avoid adsorption effects due to low alkenone concentrations (Grimalt et al., 2001). The third fraction was measured using a HP 6890 gas chromatograph, equipped with a cold injection system, a DB-1MS fused silica capillary column (60 Â 0.32 mm inner diameter, film thickness of 0.25 mm) and a flame ionization detector at the Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research (AWI). Individual alkenone identification (C 37:3 , C 37:2 ) was based on the retention time and comparison with an external standard, which were also used for controlling the instrument stability. Replicate analyses of a reference standard suggest an analytical uncertainty of~0.5 C for U k 0 37 . The alkenone unsaturation index as proxy for SST (Prahl and Wakeham, 1987) was determined for all core-top samples. SSTs were calculated using the relationship between U k 0 37 and mean annual temperature (U k 0 37 ¼ 0.033T þ 0.044) based on global core-top sediments between 60 S and 60 N in the Atlantic, Indian and Pacific oceans (Müller et al., 1998). The standard error of this calibration is reported as ± 0.050 U k 0 37 units or ca. ±1.5 C.

Results
Results of reconstructed SSTs from sediment surface samples were now compared to instrumental mean annual temperature data. At this point, the choice of the instrumental dataset is critical. For example, instrumental time-series show a strong warming trend in the Bering Sea leading to instrumental mean annual temperatures that are >0.5 C higher in 2009 compared to 1994 (see also Ren et al., 2014). Our dataset is composed of surface sediment samples collected from 1997 to 2013. To minimize the effect of recent warming in the subpolar Pacific to our dataset we extracted instrumental mean annual temperature data from World Ocean Atlas 2005 (WOA05; Locarnini et al., 2006) using the data analysis and visualization software Ocean Data View (Schlitzer, 2015). The reconstructed temperatures range between 2.7 C and 18.2 C in the study area and largely reflect the latitudinal trend in SSTs as expected from WOA05 data ( Fig. 2a and b; Table 1).
In a next step, residuals were calculated for each data point by subtracting mean annual SST (WOA05) from respective reconstructed U k 0 37 -based SSTs ( Fig. 2c and d; Table 1). Positive residuals of U k 0 37 -based SSTs reflecting warmer than mean annual SST and negative residuals colder than mean annual SST. South of the SAF U k 0 37 -based SSTs show a reasonable fit to mean annual temperatures, and with a few exceptions, residuals do not exceed the reported standard error of calibration of ±1.5 C ( Fig. 2c and d, Table 1). However, a closer look to study sites north of the SAF reveals large discrepancies between instrumental mean annual SST (WOA05) and alkenone temperatures. U k 0 37 -based SSTs overestimate instrumental mean annual temperatures by up to 5.3 C in the central subpolar Pacific and Gulf of Alaska, up to 3 C in the Bering Sea and up to even 6.1 C in the Okhotsk Sea ( Fig. 2c and d, Table 1). Warmer than expected reconstructed mean annual SSTs are apparent throughout the entire subpolar Pacific, including the Bering Sea and Okhotsk Sea, thus indicating that reconstructed temperatures north of the SAF generally overestimate instrumental mean annual SST.
Only at a few sites located off the coast of East Kamchatka and at northernmost sites of the Bering Sea and Okhotsk Sea, calculated residuals are within the error of the U k 0 37 calibration ( Fig. 2c and d).

Discussion
Based on our SST compilation we found significant regional discrepancies north and south of the SAF. Alkenone SSTs from core locations south of the SAF largely mimic the mean annual temperature signal (Fig. 2c), which are in accordance with available sediment trap data providing evidence for extended growth season of coccolithophorids from the beginning of spring (March) until autumn (November) in the mid-latitude Northwest Pacific (Yamamoto et al., 2007). Thus, we presume that the better fit of reconstructed SST to instrumental mean annual SST in the midlatitude Pacific reflects an extended, almost year-round growth season south of the SAF.
Considering the data north of the SAF, a systematic and nonrandom pattern in departure from core-top alkenone SST versus instrumental mean annual SST is apparent in all different regions of  the subpolar Pacific ( Fig. 2c and d). Our study corroborates the reported warm bias between U k 0 37 -based SSTs and mean annual SSTs in coastal marine sites throughout SE Alaska and the subpolar Northeast Pacific (Prahl et al., 2010;M eheust et al., 2013). Based on our extended North Pacific dataset (now including the subpolar Northwest Pacific, western Bering Sea and the Okhotsk Sea), alkenone-based SST estimates from the whole subpolar Pacific Ocean are generally in conflict with the premise from the global core-top calibration of U k 0 37 , where the best fit is obtained with mean annual SST (Müller et al., 1998) (Fig. 2c and d).
5.1. Seasonality and its imprint on sedimentary U k 0 37 -signal north of the SAF: the subpolar Pacific In the subpolar Pacific marine productivity and export flux patterns to the ocean interior are intimately linked to seasonal variations in meteorological and physical oceanographic conditions (e.g. Honda et al., 2002). From a biological perspective, sedimentary alkenone compositions should reflect the season when productivity of alkenones and flux of this biomarker signal to the sediments appear (e.g. Sikes et al., 1997). The best approach to evaluate the issue of seasonality in alkenone flux possibly encoded in sedimentary U k 0 37 signals is through combination of sedimentary alkenone temperature estimates with modern alkenones flux data derived from sediment trap time series.
In this regard, the North Pacific (including the Okhotsk and Bering Sea) is highly suitable because this region has been thoroughly studied with sediment trap time series during the past decades and seasonal production and export flux of alkenones are relatively well known. To further examine this issue we considered available information from published sediment trap time series of the subpolar Pacific, Bering Sea and Okhotsk Sea to identify times of alkenone flux to the sediment in the study area (Harada et al., 2003(Harada et al., , 2006aSeki et al., 2007) (Fig. 3a). In the subpolar Northwest Pacific blooms of E. huxleyi typically occur from late summer to autumn with maximum flux of alkenones observed during OctobereNovember (Harada et al., 2006a). In the Bering Sea and Okhotsk Sea E. huxleyi blooms are restricted to the autumn season (SeptembereNovember), suggesting that alkenones are synthesized during limited periods (Harada et al., 2003;Seki et al., 2007). A more recent study on nineteen-year time-series of coccolithophore fluxes further demonstrates that the main blooming season and export of E. huxleyi is heavily biased to OctobereNovember in the Bering Sea and subpolar Pacific (Tsutsui et al., 2016). From available sediment trap time series we identified October (autumn) as common midpoint for maximum alkenone flux north of the SAF. Accordingly, we extract respective instrumental October (autumn) SST data from WOA05 that were now compared to sedimentary U k 0 37 -based SSTs (Fig. 3).
Because of strong seasonality in the subpolar Pacific we first took into account the calibration of U k 0 37 (U k 0 37 ¼ 0.038T -0.082) proposed by Sikes et al. (1997) originally obtained from core-top sediments of Southern Ocean. It has been proposed that these calibration better capture seasonally biased SSTs also in the subpolar Pacific (e.g. M eheust et al., 2013). Consequently, U k 0 37 were converted to SSTs according to the Sikes et al. (1997) calibration and results compared to instrumental autumn SST data from WOA05, considering production and export of alkenones to the sediments reported from sediment trap data of the subpolar Pacific. Residuals were calculated for each data point by subtracting autumn SST (WOA05) from respective U k 0 37 -based SSTs applying the Sikes et al. (1997) calibration ( Fig. 3c and d; Table 1).
Using the Sikes et al. (1997) calibration, subpolar Pacific, as well as Bering Sea and Okhotsk Sea SST estimates are in general 2e4 C warmer than recorded from WOA05 autumn instrumental SST (Fig. 3). Anomalously colder than predicted SST estimates, however, are observed at a few coastal sites near Kamchatka. In general, calculated residuals suggest that the Sikes et al. (1997) calibration in the study area largely overestimates instrumental autumn SSTs (warm bias), as evidenced by very positive residuals of~2e4 C. (Fig. 3c and d). We note that results are different to a previous study from M eheust et al. (2013) that proposed a better fit of reconstructed SSTs of the subpolar Pacific using the Sikes et al. (1997) Southern Ocean calibration of U k 0 37 . However, the study of M eheust et al. (2013) did not consider observational data from sediment traps that clearly show the modern blooming period of E. huxleyi is heavily biased to the autumn season (Harada et al., 2003;Seki et al., 2007;Tsutsui et al., 2016).
To further evaluate the potential seasonal bias in sedimentary alkenone temperature estimates the commonly used U k 0 37 global core-top calibration equation (Müller et al., 1998) are compared to instrumental autumn SST (Table 1). We note that the U k 0 37 global core-top calibration equation is virtually identical to the initial U k 0 37 calibration based on laboratory culture studies based on North Pacific strain of E. huxleyi (Prahl and Wakeham, 1987;Prahl et al., 1988). However, the U k 0 37 global core-top calibration equation is based on sedimentary U k 0 37 , and does better represent natural oceanic conditions and variability. So we move on with the Müller et al. (1998) calibration to compare sedimentary alkenone temperatures to instrumental autumn SST (Fig. 4). Observations from re-examination of SST estimates are surprisingly straightforward. Using the Müller et al. (1998) calibration results in much improved fit between reconstructed SSTs and instrumental autumn SST (Fig. 4). Calculated residuals suggest no significant offset (no warm bias) within the different regions of the subpolar Pacific. Most study sites are within the standard deviation of ±1.5 C reported for the given calibration (Fig. 4 c and d). Our study thus further corroborates the observation that using the global core-top calibration, would produce errors at the level of ±1.5 C or less, which are not larger than the mean standard error of the entire regression for the global ocean calibration (Herbert, 2003). A bit surprising, further consideration of a seasonal (autumn) calibration of U k 0 37 from Müller et al. (1998) results in a worse fit to instrumental SST in our study area (Table 1, not shown). We conclude that the commonly used calibrations of U k 0 37 of Müller et al. (1998) and Prahl et al. (1988), when seasonality in alkenone production and export are known and considered, are the best estimators (with a few exceptions) of SSTs in the subpolar Pacific (Table 1; Fig. 4d).
Exceptions are again observed at a few sites near the coast off Kamchatka and at the northernmost site in the Bering Sea, where residuals clearly exceed the standard deviation of ±1.5 C (Fig. 4c  and d). The northernmost site in the Bering Sea shows anomalously low U k 0 37 -value of 0.135 or þ2.7 C. However, this value is at the lower limit of the temperature range of the global core-top calibration and should not be interpreted quantitatively (e.g. Herbert, 2003). We also observe colder than expected SSTs near the continental margin of Kamchatka that are situated within the outflow region of the EKC (Fig. 4d). We speculate that the anomalous colder than expected U k 0 37 -signal is probably affected by lateral transport of allochtonous alkenones from the Bering Sea to the North Pacific via the EKC. Lateral advection of allochtonous alkenones may obscure U k 0 37 values and has been reported to occur at analogue locations in the North Atlantic proximate to major fronts in SST and nutrients (Rosell-Mel e et al., 2000).  (Locarnini et al., 2006), with location of Multicore surface samples north of the SAF (symbols). Red triangles mark the sediment trap data location (Harada et al., 2003(Harada et al., , 2006aSeki et al., 2007;Tsutsui et al., 2016) considered in this study. (b): gridded SST data based on seasonal U k 0 37 calibration reported for cold regions (Sikes et al., 1997). (c): spatial distribution of calculated residuals (reconstructed SST Sikes et al., 1997 -WOA05 Autumn ). (d): detailed comparison of calculated residuals (reconstructed SST Sikes et al., 1997 -WOA05 Autumn ) north of the Subarctic Front (SAF). Dashed lines in gray denote uncertainty of ±1.5 C reported for the global calibration of U k 0 37 (Müller et al., 1998). This figure was generated with Ocean Data View (Schlitzer, 2015). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) 5.2. Potential implications for alkenone-derived SST reconstructions in the subpolar Pacific: a case study from a stratigraphic alkenone record of the Okhotsk Sea For the subpolar Pacific we demonstrate that the modern U k 0 37 -signal is tightly coupled to seasonality in alkenone production that needs to be considered to derive meaningful SST estimates from alkenone temperatures. In terms of paleoceanographic reconstructions, very little is known about e.g. seasonality in alkenone production in the geological past. Hence, the basic assumption is that seasonality in alkenone production (if considered at all) did not change e.g. through different climatic intervals. However, the growth season in high-latitude ocean settings might have changed through time, an issue that potentially leads to biased paleoclimatic information drawn from stratigraphic U k 0 37 -records. Here, we conduct a case study using an IMAGES sediment core collected from near Hokkaido, in the south-western Okhotsk Sea ( Fig. 5a; MD01-2412; lat 44 31.65 0 N, long 145 00.25 0 E; water depth, 1225 m) to assess a bit further the potential effect of changing growth seasons of coccolithophorides on stratigraphic U k 0 37 -records. We decided to consider MD01-2412 because it is the only alkenone temperature record from the subpolar Pacific covering the last glacial interval with millennial to centennial temporal resolution. Low alkenone content in MD01-2412 during the LGM has been attributed to enhanced seasonal sea-ice coverage resulting in light limitation on marine productivity and shortened coccolithophorid growth period (Ternois et al., 2000;Harada et al., 2004). Anomalously warm alkenone temperatures were reported from the stratigraphic U k 0 37 -record of MD01-2412 during the glacial period (Fig. 5). It has been speculated that a shift of the bloom season of the alkenone producer from autumn to summer may have caused the high alkenone temperatures recorded in the glacial sediments of the south-western Okhotsk Sea (Harada et al., 2006b(Harada et al., , 2008. First, we have a look into the modern seasonal variation in SST at Site MD01-2412 (Fig. 5b). The measured alkenone temperature at the core top is 9.1 C, which is very close to instrumental temperatures in the upper 20 m of the water column during JulyeAugust and OctobereNovember (Harada et al., 2006b). In the Okhotsk Sea, autumn is the main season of production for E. huxleyi and main export of coccospheres is from NovembereDecember (Broerse et al., 2000;Seki et al., 2007). Thus, modern alkenone temperatures at this study site reflect the autumn SST in the upper 20 m of the water column in autumn (Fig. 5b). Whether a shift of the bloom season to summer (July) may have caused the high alkenone temperatures recorded in the glacial sediments of the southwestern Okhotsk Sea (e.g. Harada et al., 2008) we need to estimate the seasonal SST gradient (D T). Because seasonal data is not available for the glacial interval we can only consider the modern monthly average data at 20 m water depth from WOA to assess D T. Based on instrumental SST data the modern D T is~4.5 C from summer (July) to autumn (October) season (Fig. 5b) near site MD01-2412. Hence, estimated D T is used to mimic a "shifted bloom season" of the alkenone producer from autumn to summer from alkenone time-series for the glacial interval (11e115 ka BP) of MD01-2412.
Comparison of the original alkenone time-series from MD01-2412 (Harada et al., 2006b) and the alkenone time-series with "shifted bloom season" are given in Fig. 5c. Looking at the "shifted bloom season" alkenone time-series the anomalously warm glacial alkenone temperatures have been vanished from the stratigraphic  (Locarnini et al., 2006) together with location of Multicore surface samples north of the SAF (symbols). Red triangles mark the location of sediment trap data (Harada et al., 2003(Harada et al., , 2006aSeki et al., 2007;Tsutsui et al., 2016) that has been considered in this study. (b): gridded SST data based on global U k 0 37 calibration (Müller et al., 1998). (c): spatial distribution of calculated residuals (reconstructed SST Müller et al., 1998 -WOA05 Autumn ). (d): detailed comparison of calculated residuals (reconstructed SST Müller et al., 1998 -WOA05 Autumn ) north of the Subarctic Front (SAF). Dashed lines in gray denote uncertainty of ±1.5 C reported for the global calibration of U k 0 37 (Müller et al., 1998). This figure was generated with Ocean Data View (Schlitzer, 2015). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) U k 0 37 -record. Based on this exercise, glacial alkenone temperatures at Site MD01-2412 are now about~2.5 C colder than during the Holocene. Hence, a past shift of the bloom season of the alkenone producer at Site MD01e2412 may explain a large part of anomalously high alkenone temperatures recorded in the glacial sediments of the south-western Okhotsk Sea (Harada et al., 2006b). Considering a glacial shift of the main production season of alkenones the time-series now better fits to available TEX 86 biomarker reconstructions and numerical simulations that also show colder glacial than Holocene SST of the central Okhotsk Sea during the glacial period (Lo et al., 2018). Lower glacial than Holocene SSTs have been also reported from TEX 86 biomarker reconstructions of the Bering Sea and the subpolar Northwest Pacific (Meyer et al., 2017). However, results of this case study are based on modern monthly average data from WOA. Hence, doubts are allowed how meaningful these results for paleoclimatic reconstructions really are. Nevertheless, our case study from the Okhotsk Sea underpins the potentially important role of seasonality on high-latitude settings with even small shifts in past bloom seasons may lead to heavily biased SST information drawn from such stratigraphic U k 0 37 -records.

Further implications
Based on combination of modern alkenone flux data from sediment trap time series with a comprehensive dataset of  Harada et al., 2006b) and the alkenone time-series with glacial "shifted bloom season" (in blue; this study). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) U k 0 37 -based SSTs from North Pacific core-tops revealed that U k 0 37 -based SST estimates largely correspond to instrumental SSTs during times of maximum export flux of alkenones in the subpolar Pacific. These results are in line with former observations from subantarctic waters where alkenone-based SST reproduces the SST values when alkenone flux to the sediment high (e.g. Sikes et al., 2005). It reinforces the idea that U k 0 37 , when seasonality in alkenone production and export are known and considered, is able to provide reasonable estimates of SSTs in high-latitude ocean settings (e.g. Prahl et al., 2010) and further attest, in principle, to the validity for wide use of the original global core-top calibration equation by Müller et al. (1998) to estimate past SSTs.
Given above considerations, our conclusions differ from studies that have tried to re-define the global core-top calibration to account for a seasonal bias in alkenones (Conte et al., 2006;Tierney et al., 2018). It has been shown that in case of subpolar oceans, the resultant nonlinear calibrations do not reduce the uncertainty in SST substantially (Tierney et al., 2018). According to our results, residuals in the lower temperature range of the linear global calibration equation for U k 0 37 are not entirely random, but hold valuable information about secondary controls encoded in the sedimentary U k 0 37 -signal, such as seasonality in production and export of alkenones in the subpolar Pacific. Finally, seasonality of maximum alkenone flux in sediment traps varies across the oceans and through time, a consistent, globally applicable, seasonal pattern is still not apparent (Rosell-Mel e and . Further studies combining sediment trap time series with sedimentary U k 0 37 -records are the key to identify regional differences in the U k 0 37 -signal in marine surface sediments and to further improve paleoceanographic interpretations based on sedimentary alkenone records.

Conclusions
In this study, we present a compilation of 97 sediment surface samples from Multicores collected in the Bering Sea, the Okhotsk Sea and the North Pacific that has been used to evaluate the alkenone-temperature proxy against observational data in the high-latitudes of the North Pacific. Surface samples were analysed for alkenones and the U k 0 37 -index converted to water temperatures using different calibration equations established in the literature. We found that south of the SAF alkenone derived SSTs show a reasonable fit to mean annual temperatures. However, study sites north of the SAF do not match instrumental mean annual SST and alkenone temperatures show a large, systematic offset (warm bias) throughout all different regions of the subpolar Pacific. To further investigate this issue we combined information from instrumental SST data as well as modern alkenone flux data from sediment traps to investigate the role of seasonality on the sedimentary alkenonesignal. Comparison of the sedimentary alkenone-signal to observational data revealed that U k 0 37 -based SSTs match modern autumn temperatures north of the Subarctic Front (SAF), when maximum export flux of alkenones to the seafloor is indicated by sediment trap data. Most sample sites are within the standard deviation of ±1.5 C reported for the global calibration provided by Müller et al. (1998). Our results demonstrate that U k 0 37 , when seasonality in alkenone production and export are known and considered, is a robust proxy for SSTs in high-latitude ocean settings. We conduct a case study using an IMAGES sediment core collected from the south-western Okhotsk Sea to assess the potential effect of seasonality on stratigraphic U k 0 37 -records in the subpolar Pacific. The case study from the Okhotsk Sea show that even a small shift in seasonality may lead to strongly biased SST reconstructions with broader implications for paleoclimatic information drawn from such stratigraphic U k 0 37 -records.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.