The oceanographic configuration around Iceland is predominantly characterized by the interplay of two water masses, i.e., warm and saltier Atlantic water versus cold Arctic or subpolar waters. From the south flows the Irminger Current (IC) – a branch of the warm and salty Atlantic current, which moves northwards along the west Iceland coast and continues along the north Iceland coast, extending down to several hundred meters (Hopkins, 1991) (Fig. 1). Flowing southward from north to west Iceland, the East Greenland Current (EGC) transports cold and low-salinity polar waters. A branch of the EGC, the East Iceland Current (EIC), turns eastward and flows southward along the east coast of Iceland (Hopkins, 1991). The EGC carries icebergs and sea ice formed in the Arctic Ocean and in East Greenland fjords (Sigtryggsson, 1972). The wide transitional zone between the polar waters and the Atlantic waters in the Denmark Strait is defined as the polar front. The position of this front is known to vary on annual, interannual and longer timescales (Malmberg and Jónsson 1997; Sigtryggsson, 1972). During episodes of extensive sea ice, the contribution of polar waters from the EGC to EIC is relatively large and responsible for carrying sea ice, icebergs and cold, low-salinity waters to the northwest coast of Iceland (Sigtryggsson, 1972).

The polar front is also expressed in the phytoplankton blooms around Iceland. In the Arctic or subpolar waters, the early onset of stratification in spring gives rise to a rapid shallowing of the mixed layer and triggers the early spring bloom (in early April) north of Iceland. In the south of Iceland, the weakly stratified water column in the Atlantic water and an associated deep mixed layer delays the spring bloom (e.g., Zhai et al., 2012, and references therein). The spring bloom initiation varies by up to a month between different regions around Iceland. Within the southern Iceland shelf region, the spring bloom generally starts in near-shore waters (mid-May) and is delayed with increasing distance from the coast, where it is affected by the interaction between runoff and wind regime (Thordardottir, 1986).

Sample material was collected around Iceland during long-chain diol cruises (Cruise Report 64PE341, de Haas, 2011; and Cruise Report 64PE357, Baas and Koning, 2012) in the summer of 2011 and 2012 on board the R/V Pelagia (Fig. 1, Table 1). During July 2011, suspended particulate matter (SPM) from a water depth of ca. 5 m and surface sediments were collected at different stations (St) around Iceland (Table 1), and a sediment trap was deployed at 1850 m water depth at St 1 (water depth at this station is 2255 m), located in the northern part of the Iceland Basin, to recover sinking particulate matter (Fig. 1). During July 2012, SPM was collected at 20 and 50 m water depth in a transect from the site of the sediment trap deployment to Reykjavik (Fig. 1, Table 1) after the sediment trap was recovered.

SPM was obtained by filtering through 142 mm diameter glass-fiber filters (GFFs) with a pore diameter of 0.7 µm using a McLane Research Laboratories WTS 6-1-142LV in situ pump installed on a CTD (conductivity–temperature–depth) rosette frame. The CTD measured the vertical distribution of temperature, salinity, turbidity, oxygen and fluorescence. SPM filters were frozen directly after filtration and stored at -20 ∘C until analysis.

A McLane Parflux 78H-21 sediment trap (aperture area: 0.5 m2) was set to collect sinking material every 17.5 days in a 21-cup automated sampling carousel covering one complete annual cycle (Table 2). Prior to mooring the sediment trap, the sample cups were filled with a mercuric-chloride-poisoned and borax-buffered solution of seawater collected from the deployment depth (1 g L-1 of HgCl2; pH ∼ 8.5). After recovery of the sediment trap, collecting cups were stored in the dark at 4 ∘C.

Sediment cores were taken using a multicorer, sliced into 1 cm wide sections and frozen onboard. The upper 1 cm was analyzed for biomarkers.

In the laboratory, any larger “swimmers” were removed from the sediment trap collecting cups prior to subdividing into two volumetrically split aliquots using a Folsom wet splitter (Sell and Evans, 1982) with a precision of > 95 %. One half was stored in the dark at 4 ∘C, and the second half was used for lipid analysis. The trap material was centrifuged at 3000 rpm for 15 min, followed by pipetting the water layer and washing with bidistilled H2O (3×) to remove the HgCl2 and borax solution. The sediment trap samples (n=21) and surface sediment samples (n=10) were freeze-dried, homogenized in agate mortar and extracted after the addition of extracted diatomaceous earth in an accelerated solvent extractor 350 (ASE 350, DIONEX) using a solvent mixture of 9:1 (v:v) dichloromethane (DCM) to methanol (MeOH) at 100 ∘C and 7.6 × 106 Pa. The solvent from all the extracts was reduced by TurboVap LV Caliper, dried over Na2SO4 and concentrated under a stream of N2, yielding a total lipid extract (TLE). Three internal standards were added to the TLE from the sediment trap samples, i.e., 10-nonadecanone (C19 ketone) for alkenones, C22 7,16-diol for long-chain diols and the C46 glycerol trialkyl glycerol tetraether (GTGT) for GDGTs (Huguet et al., 2006). Activated copper and DCM were added to the TLEs of the sediment trap samples that were found to contain elemental sulfur. After being stirring overnight with a small stirring bar, the TLEs were filtered over a pipette column containing Na2SO4 and dried under a stream of N2.

SPM filters (n=14) were freeze-dried and half of each filter was saponified according to de Leeuw et al. (1983) by refluxing for 1 h with 1 M KOH in MeOH (96 %). After cooling, the solvent was acidified with 2 N HCl in MeOH (1:1, v:v) to a pH of 2, and transferred to a separatory funnel containing bidistilled H2O. The residual filters were further extracted using H2O : MeOH (1:1, v:v, ×1), MeOH (×1) and DCM (×3), and all solvents were combined in the separatory funnel. The DCM layer in the separatory funnel was separated from the H2O : MeOH layer and the remaining H2O : MeOH layer was extracted three times with DCM. DCM layers were combined and rotary evaporated to near dryness. Thereafter, the obtained extracts were acid hydrolyzed (3 h reflux with 2 N HCl : MeOH, 1:1, v:v) and neutralized with 1 M KOH in MeOH (96 %). Then 3 mL bidistilled H2O was added to the acid hydrolyzed extracts and the lipids were extracted using DCM (4×). The filter material remaining after base hydrolysis was also acid hydrolyzed (3 h reflux with 2 N HCl : MeOH, 1:1, v:v). After cooling, the solvent was neutralized with 1 M KOH in MeOH (96 %) and transferred to a separatory funnel containing bidistilled H2O. Subsequently, the acid-hydrolyzed residual filters were extracted using H2O : MeOH (1:1, v:v, ×1), MeOH (×1) and DCM (×3), and the extracts were combined in the separatory funnel. The DCM layer was collected and the remaining H2O : MeOH layer was extracted three times with DCM. All extracts, obtained by saponification and acid hydrolysis, were combined, dried under N2, eluted in DCM over a pipette column containing Na2SO4 and dried under a stream of N2.

Extracts of SPM, descending particles and surface sediments were separated into apolar, ketone (containing alkenones) and polar fractions (containing GDGTs and long-chain diols) by column chromatography using a Pasteur pipette filled with Al2O3 (activated for 2 h at 150 ∘C) using 9:1 (v:v) hexane : DCM, 1:1 (v:v) hexane : DCM, and 1:1 (v:v) DCM : MeOH as the eluents, respectively.

SPM was collected during two cruises: in July 2011 at six stations around Iceland at ca. 5 m water depth, and in July 2012 in a transect (St A–G) from the northern Iceland Basin to Reykjavik at 50 m and, at some locations, also at 20 m water depth (Fig. 1).

Alkenones were detected in all samples except from St 13. Values for the U37K′ index varied between 0.26 and 0.53 (or 6.4 and 14.7 ∘C when translated into temperature) in the SPM around Iceland during summer 2011 (Fig. 2a, open green diamonds) and between 0.26 and 0.45 (corresponding to 6.5 to 12.5 ∘C) during summer 2012 (Fig. 2b, open green diamonds). GDGTs were detected in all samples and the TEX86 values ranged between 0.49 and 0.55 (corresponding to 13.1 to 17.9 ∘C) in SPM around Iceland (Fig. 2a, dark blue circles) and between 0.34 and 0.50 (corresponding to 1.0 to 14.4 ∘C) along the transect of 2012 (Fig. 2b, dark blue circles). TEX86L varied between 0.28 and 0.32 (corresponding to 9.4 to 13.7 ∘C) around Iceland (Fig. 2a, open blue circles) and between 0.22 and 0.29 (corresponding to 2.5 to 10.6 ∘C) along the transect (Fig. 2b, open blue circles). Long-chain alkyl diols were not detected in SPM around Iceland collected during summer 2011 (Fig. 2a), while C28 and C30 1,13- and 1,14-diols and C30 and C32 1,15-diols were only detected in SPM collected at St A, B, F and G during the transect in the summer of 2012. LDI values varied between 0.08 and 0.49, corresponding to -0.4 to 12 ∘C when converted into temperature (Fig. 2b, open brown squares). Values for U37K′, TEX86 and LDI and corresponding temperatures are reported in Supplement Table S1.

Sinking particulate matter was collected between 15 July 2011 and 16 July 2012 at St 1, using a sediment trap deployed at 1850 m water depth. Bulk sediment fluxes varied between 4 and 165 mg m-2 day-1 (Fig. 3b), with high fluxes occurring in July, September and October 2011 and from May to July 2012. C37 alkenone fluxes varied between 0.02 and 130 µg m-2 day-1, peaking during spring and summer, i.e., in July 2011 and in May and June 2012 (Table 2; Fig. 3c). Fluxes of the GDGTs used for the calculation of the TEX86 index ranged between 53 and 2300 µg m-2 day-1 with highest values recorded from May to June 2012 (Table 2; Fig. 3d). GDGT-0 and crenarchaeol fluxes followed the same pattern as those of the GDGTs used in the TEX86, ranging from minimum values of ca. 170 µg m-2 day-1 to a maximum of 15 and 11 mg m-2 day-1, respectively (Fig. 3d). Values for the BIT index were always below < 0.01. Fluxes of long-chain 1,13- and 1,15-diols used in the LDI were low compared to alkenones and GDGT fluxes and varied between 1.5 and 170 ng m-2 day-1, with highest values recorded during July, September and October 2011 and from May to June 2012 (Table 2; Fig. 3e). Generally, the flux of the C30 1,15-diol was always low (up to 2 ng m-2 day-1) or even below the detection limit for most of the intervals. The fluxes of saturated and monounsaturated C28 and C30 1,14-diols were substantially higher than those of the 1,13- and 1,15-diols (Fig. 3e). The highest summed mass flux of the C28 and C30 monounsaturated 1,14-diols was recorded in September 2011 with a flux of 3.7 µg m-2 day-1 (Fig. 3e). C28 and C30 saturated 1,14-diol fluxes varied between minimum values of 3.8 ng m-2 day-1 during the second half of April and maximum values of 770 ng m-2 day-1, with high fluxes recorded during July, September and October 2011 and May and June 2012 (Table 2; Fig. 3e).

U37K′-based temperatures derived from settling particles ranged from 5.3 to 11.4 ∘C (Fig. 4), with maximum values recorded at the end of summer (September 2011) and late winter (end of February 2012) and minimum values in spring (from May to June 2012) (Fig. 4, green line), when the highest flux is observed. Temperature estimates based on TEX86 varied between 6.8 and 9.6 ∘C (Fig. 4, dark blue line) and those based on TEX86L varied between 12.6 and 17.5 ∘C (Fig. 4, light blue line), with the highest values recorded from November 2011 to April 2012. TEX86L-temperature estimates based on the 0–200 m calibration showed absolute values ranging from 10.3 to 13.9 ∘C (Fig. 4, red line). For sampling periods when 1,13- and 1,15-diols were detected, the LDI-based temperatures vary between -2.7 and 0.2 ∘C (Fig. 4, brown line and open squares). Values for U37K′, TEX86 and LDI and corresponding temperatures are reported in Table S2.

Surface sediments were collected in July 2011 from 10 stations around Iceland (Fig. 1; Table 1). The U37K′ index varied between 0.26 and 0.53 in the surface sediments, yielding SST estimates between 7 and 11 ∘C for St 13 and St 7, respectively (Fig. 5, open green diamonds). TEX86 ranged between 0.36 and 0.44 with SST estimates between 2.4 and 9.2 ∘C (Fig. 5, dark blue circles). TEX86L ranged between 0.19 and 0.33, resulting in TEX86L-derived temperatures between -1.2 ∘C at St 13 and 14 ∘C at S. 1 (Fig. 5, open light blue circles) or between -0.1 and 11.4 ∘C using the 0–200 m calibration (Fig. 5, open red circles). LDI values varied between 0.02 and 0.27, with SST estimates between -2.1 and 5.2 ∘C, reaching high values in the coastal stations, St 3 and St 8 (Fig. 5, open brown squares). Values for U37K′, TEX86 and LDI and corresponding temperatures are reported in Table S3.

Long-chain alkenones are produced by several haptophyte algal species thriving in the photic zone (Volkman et al., 1980, 1995; Marlowe et al., 1984) and are, therefore, thought to reflect SST. Although previous studies of cold polar waters (< 4 ∘C) of the North Atlantic have shown relatively high abundances of C37:4 (e.g., Sicre et al., 2002), the C37 alkenones in SPM, descending particles and surface sediments around Iceland comprised only C37:3 and C37:2, and no C37:4 was detected. Comparison of U37K′-derived temperatures with in situ temperatures showed generally lower U37K′-derived temperatures, differing by up to 3.4 ∘C for SPM around Iceland (Fig. 2a) and by up to 6.6 ∘C for the SPM transect (Fig. 2b). Reduced temperature differences (up to 2.6 ∘C) were observed when we compared U37K′-derived SSTs with summer temperatures at 50 m water depth (Fig. 2b, purple crosses; derived from the World Ocean Atlas (WOA) 09 database; Locarnini et al., 2010), at which most of the SPM was recovered from the transect. Possibly, the alkenones collected in the SPM did not represent recently produced material but alkenones synthesized over several months. Since SPM was collected in July, the warmest month of the year, the U37K′ would reflect lower temperatures if the signal also reflected material synthesized in the preceding colder months.

Interestingly, U37K′-derived SSTs of sedimenting particles also show major discrepancies compared to satellite SSTs, i.e., somewhat higher U37K′-derived SSTs were observed from January to mid-May (differing by around 2–3 ∘C) and lower temperatures from mid-May to July (differing by up to 4.9 ∘C) (Fig. 4, green line and open diamonds) at the time of the highest alkenone flux (Fig. 3c). The underestimation of temperatures by U37K′ in sedimenting particles in July is consistent with the discrepancy between U37K′-derived and in situ temperature observed for SPM for the same time period. The application of the U37K′ SPM calibration proposed by Conte et al. (2006) also results in a general overestimation (by up to 2.7 ∘C) of U37K′-derived SSTs in both the SPM and sedimenting particles (data not shown), suggesting that the difference between U37K′-derived and in situ temperature is not due to calibration issues.

Higher U37K′-derived SST in cold periods could perhaps be attributed to the gradual sinking of alkenones that were produced in preceding warmer time periods. Similar discrepancies, i.e., U37K′-SSTs overestimating in situ winter temperatures and underestimating in situ summer temperatures in the surface mixed layer, have been previously described for alkenones in sediment traps from other subpolar and midlatitude regions (e.g., Sikes et al., 2005; Harada et al., 2006; Seki et al., 2007; Yamamoto et al., 2007; Lee et al., 2011). In the Mediterranean, Arabian Sea and the Pacific, U37K′-SSTs lower than in situ SST during a high alkenone flux have been attributed to either alkenone production at the thermocline depth or to nutrient deficiency (e.g., Ternois et al., 1997; Prahl et al., 2000; Harada et al., 2006; Popp et al., 2006).

A compilation of previous sediment trap studies in the North Atlantic has shown that the U37K′ export signal produced in surface waters is not equivalent to the vertically transported U37K′ signal collected in the underlying sediment traps or accumulating in surface sediments (Rosell-Melé and Prahl, 2013). In the current study, the U37K′-SST value obtained for the surface sediment at St 1, ca. 10.7 ∘C (Fig. 5), corresponds well with annual mean SST from WOA09 of 9.4 ∘C (Locarnini et al., 2010) but is higher than the U37K′-derived temperature of the flux-weighted average of our sediment trap data (7.1 ∘C; Table 3). This difference seems mainly due to the anomalously low U37K′-derived temperatures at the time of high alkenone fluxes. These discrepancies could result from (1) a bias from advected or resuspended alkenones by oceanic currents masking the local pattern of export production from overlying surface waters (e.g., Prahl et al., 2001), as has been previously noted in the NE Atlantic (Rosell-Melé et al., 2000); (2) an unusual subsurface production of alkenones in the summer of 2011 leading to a cold bias for that year only; or (3) selective degradation of alkenones when they are sedimenting on the sea floor (e.g., Hoefs et al., 1997; Rontani et al., 2013).

The calculation of sinking velocities for alkenones can be done following the approach of Fischer and Karakas (2009) and Mollenhauer et al. (2015) and uses the offsets between minimum and maximum temperatures recorded in the calculated U37K′ temperatures and SST. Maximum derived temperatures are obtained for sampling periods 4 (5–23 September 2011) and 14 (26 February–15 March 2012), while the maximum satellite SSTs are in the months June and July (Fig. 4 and Table S2). The minimum alkenone-derived temperature is during sampling period 19 (24 May–10 June 2012), and minimum satellite SSTs are in February–April 2012 (Fig. 4 and Table S2). Based on this, average sinking velocities (0–1850 m) for alkenones are ca. 25–30 m day-1, which is substantially lower compared to the sinking rates of alkenones in the filamentous upwelling region off Cape Blanc (Müller and Fischer, 2001). However, our estimates are quite uncertain as alkenone-derived temperatures do not show a clear seasonal trend compared to the seasonal SST trend (Fig. 4). Furthermore, an offset of more than 2 months between alkenone production and their fluxes recorded at 1850 m water depth seems in contrast with the similar timing of alkenone flux patterns and net primary production (derived from OceanColor Web; Behrenfeld and Falkowski, 1997) over this period (Fig. 3a, c).

In surface waters in the Norwegian–Iceland seas, the abundances of coccolithophore communities are usually higher during the high-bloom period (summer) than during the low-bloom period (late summer and fall) (Baumann et al., 2000). Furthermore, high alkenone fluxes were also previously observed from April to June, with a rapid decline until August, in 1989 in the NE Atlantic (Rosell-Melé et al., 2000). Consequently, the increased flux of alkenones in spring likely reflects the spring bloom. U37K′-derived SSTs at the peak flux of alkenones thus likely reflect spring and early summer temperatures, in agreement with previous studies from high-latitude sites (Sikes et al., 1997; Ternois et al., 1998; Rosell-Melé et al., 2000; Sicre et al., 2005, 2006; Conte et al., 2006; Hanna et al., 2006). Degradation during transport in the water column or in the oxic sediment layer is expected to result in higher U37K′-derived SSTs, since the C37:3 has a higher degradation rate than C37:2 (Prahl et al., 1988, 2003; Hoefs et al., 1998; Gong and Hollander, 1999), and this is indeed what we observe in the surface sediment compared to the flux-weighted mean (Table 3). An alteration of alkenones in the sea floor may thus explain the mismatch between surface sediment signal and flux-weighted mean signal. Finally, it is important to keep in mind that the data obtained with the sediment trap only provides a snapshot in time, while the surface sediment stores information collected over decades to centuries. Thus, the offset of the weighted average U37K′-derived SSTs may be just a particular feature for the year 2011 and not representative of what happened over the last few centuries.

To test the effect of seasonality and diagenetic alkenone alteration around Iceland, we compared U37K′-SST signals from all surface sediments with annual mean and seasonal SSTs (derived from the WOA09 database; Locarnini et al., 2010) (Fig. 6a; color code as in Fig. 1). U37K′-derived SSTs show a good linear correlation with annual mean SSTs, although absolute temperature values are higher at each station, with temperature differences ranging from 1 to 4 ∘C (Fig. 6a), generally higher than the calibration error, i.e., 1.5 ∘C (Müller et al., 1998) and with the highest deviations for the northernmost stations. When we compared U37K′-derived temperature values with SST from different seasons, the best fit is obtained with the summer mean SSTs (Fig. 6b). This is in agreement with peak alkenone fluxes recorded in our sediment trap during late spring and early summer. Thus, the sedimentary signal of U37K′-SST around Iceland seems in general to reflect the maximum production season of alkenones.

GDGTs in the marine environment are likely biosynthesized by Thaumarchaeota (Sinninghe Damsté et al., 2002), which are omnipresent in the global ocean, including the polar regions (e.g., Hoefs et al., 1997; DeLong et al., 1998; Schouten et al., 2000). TEX86L was developed for polar oceans in order to improve the correlation between TEX86 and SST (Kim et al., 2010), but a recent study has shown that TEX86 is still suitable as well (Ho et al., 2014). The TEX86L-SST estimates in the SPM around Iceland showed a highly variable relationship with in situ temperatures, showing temperatures up to 6 ∘C higher around Iceland during 2011 (Fig. 2a), while for the SPM transect, the TEX86L-SST temperatures were up to 7 ∘C lower compared to summer temperatures at 50 m water depth obtained from the WOA09 database (Fig. 2b). Even higher offsets, both positive and negative, up to 11 ∘C, were obtained with TEX86-SST estimates (Fig. 2a, b). A reason for the poor correspondence of TEX86-derived temperatures with in situ and satellite temperatures could be a depth habitat effect, since Thaumarchaeota can thrive deep in the marine water column (Karner et al., 2001; Herndl et al., 2005), although they tend to have their highest cell numbers at depths < 200 m (e.g., Karner et al., 2001). We also applied a specific TEX86 SPM calibration proposed by Schouten et al. (2013), but temperatures did still show significant offsets with in situ temperatures (data not shown).

Part of the mismatch may be due to the fact that Thaumarchaeota are smaller in cell size than the 0.7 µm pore diameter of the SPM filters (Könneke et al., 2005) and thus may not be quantitatively captured on the GFFs (Ingalls et al., 2012), possibly affecting the TEX86 values. However, other studies have shown comparable TEX86 values obtained with both 0.7 and 0.2 µm pore diameter filters (e.g., Herfort et al., 2007) or good correspondence with depth (Schouten et al., 2012) or seasonal (Pitcher et al., 2011) profiles of isoprenoidal GDGT concentrations from 0.7 µm pore diameter filters and thaumarchaeotal rRNA gene abundances from 0.2 µm pore diameter filters. Thus, the filter size is unlikely to have affected TEX86 values. Another issue may be that we analyzed saponified SPM filters, combining both intact and core (non-intact) GDGT-based lipids. In the natural environment, core lipid GDGTs are derived to a substantial portion from dead cells and may thus represent a fossil signal from other areas or represent an integrated annual temperature signal. Lipp and Hinrichs (2009) showed notable differences in derived temperatures using core vs. intact polar lipid (IPL) GDGTs from marine sediments. Usually, TEX86-derived temperatures are higher for IPL GDGTs than for core GDGTs, as was observed in SPM from the Arabian Sea (Schouten et al., 2012); thus, reduced TEX86-derived SSTs from the SPM transect may related to core-GDGT contributions. However, this does not fully explain the dissimilarities observed in the SPM around Iceland, and thus we lack a clear explanation.

Regarding sedimenting particles, TEX86L-derived SSTs (Fig. 4, blue line and open circles) were all much higher (up to ca. 9 ∘C) than satellite SSTs. Reduced differences (up to ca. 5 ∘C higher than satellite SSTs) were obtained when temperature values were estimated using the TEX86L 0–200 m calibration (Kim et al., 2012b) (Fig. 4, red line and open circles and dashed purple line). Interestingly, differences in temperature decreased significantly when TEX86-derived temperature and satellite-derived SST were compared (Fig. 4; dark blue line and filled circles), particularly during times of low GDGT fluxes (Fig. 3d). During times of high GDGT fluxes, TEX86-SST were lower than satellite SST by up to 4.6 ∘C, which may suggest that the temperature signal is not derived just from surface waters. This is supported by TEX86 temperature values from SPM collected at St 1 during both cruises; TEX86-derived temperatures from the surface waters are slightly higher than the satellite SSTs, and significantly higher than the TEX86 values derived from the material collected in the same months in the sediment trap at 1850 m water depth. We observed notable differences in estimated temperatures when we used different calibrations, obtaining better results with the TEX86 calibration (Kim et al., 2010). Similar findings were made by Ho et al. (2014), who applied TEX86L and TEX86 in different polar and subpolar regions, such as the Pacific sector of the Southern Ocean and the Subarctic Front in the North Pacific. Lateral transport of GDGTs is not likely to have an effect on TEX86 temperatures since isoprenoid GDGTs are less susceptible to long-distance advection than alkenones (Mollenhauer et al., 2008; Shah et al., 2008; Kim et al., 2009a). Short-term degradation has also been shown to have no significant impact on the TEX86 (Schouten et al., 2004; Kim et al., 2009b). Terrigenous GDGTs are also unlikely to be the reason for the offset in estimated temperatures around Iceland, based on the low values of the BIT index (< 0.01) (Weijers et al., 2006, 2009) and the low correlation (R2 < 0.08) between the BIT index and TEX86 values (cf. Schouten et al., 2013). Comparison of the flux-weighted mean TEX86 value with surface sediment at St 1 shows similar values, i.e., 8.5 and 9.2 ∘C using TEX86-SST, 14.5 and 14.0 ∘C using TEX86L-SST, and 11.7 and 11.4 ∘C for TEX86L 0–200 m, suggesting no alteration of the GDGT signal during transport to the sea floor (Table 3).

As with alkenones, the sinking velocities for GDGTs can be estimated by comparing TEX86 values in the sediment trap with observed SST. Maximum TEX86-derived temperatures were observed for sampling periods 11–17 (5 January–6 May 2012) and minimum temperatures for sampling periods 4 and 18 (5–23 September 2011 and 6–24 May 2012, respectively) (Fig. 4 and Table S2). Following the method of Fischer and Karakas (2009) and Mollenhauer et al. (2015), an offset between the production of GDGTs and their fluxes at 1850 m water depth of 4–6 months would result in estimated average sinking velocities (0–1850 m) of ca. 10–15 m d-1, slower than that for the alkenones. However, it should be noted that this estimate is highly uncertain, as TEX86 temperatures do not show a clear seasonal trend in comparison to SST (Fig. 4).

GDGT fluxes were high during July, September and October 2011, followed by May and June 2012 (Fig. 3d), and showed similar patterns to bulk sediment flux and primary production (Fig. 3a, b). Similar GDGT flux patterns, i.e., high fluxes at times of high primary productivity, have been observed in the Arabian Sea (Wuchter et al., 2006), the Santa Barbara Basin, off the coast of southern California (Huguet et al., 2007), and the upwelling region off Cape Blanc in Mauritania (Mollenhauer et al., 2015). This was explained by a more efficient transport of thaumarchaeotal cells, and thus GDGTs, to deeper waters by the packaging activity of zooplankton thriving after a phytoplankton bloom. This could be also the case in the northern Iceland Basin, where the bulk sediment flux may act as an important mechanism for transporting these lipids to the seafloor. However, the fact that highest fluxes are observed in late summer does not mean that the annual TEX86 signal here also represents summer temperatures, as previously suggested by Ho et al. (2014) for surface sediments from the Arctic, northern Pacific and Southern Ocean, since the flux-weighted mean TEX86 value from the sediment trap, and the TEX86 value from the underlying sediment are both similar to annual mean SST (Table 3).

TEX86- and TEX86L-derived SST in the surface sediments distributed around Iceland correlate with WOA09 annual mean SSTs (Fig. 6c, e), although they do show a substantial offset, varying from 1 to 4.7 ∘C. A similar fit is obtained with winter mean temperatures (Fig. 6d, f), whereas the correlation with summer mean temperatures is substantially poorer (data not shown). This latter observation agrees with the idea that the highest GDGT flux during summer does not automatically result in a TEX86 signal, which records late summer temperatures (Fig. 3d). This is in contrast with observations by Ho et al. (2014), who showed anomalously high SST estimates in surface sediments from the Arctic, northern Pacific and Southern Ocean, and obtained the best correlation of TEX86-SSTs with summer SSTs.

To test if the sedimentary TEX86 signal around Iceland mainly reflects subsurface temperature waters as suggested for some other regions (e.g., Huguet et al., 2007; Lopes dos Santos et al., 2010; Kim et al., 2012a, b), we compared TEX86L 0–200 m temperature estimates (Kim et al., 2012b) with the temperature of the upper 200 m of the water column based on WOA09. We obtained a better correspondence with both annual and winter mean temperatures (Fig. 6g, h), with differences ranging from 0.5 to 3.5 ∘C. This suggests that TEX86-derived signals in the surface sediments around Iceland may reflect subsurface (i.e., 0–200 m) temperatures.

Long-chain alkyl diols, either the 1,13- and 1,15-diols involved in the LDI or the 1,14-diols produced by Proboscia diatoms (Sinninghe Damsté et al., 2003) and Apedinella radians (Rampen et al., 2011), were below the detection limit in SPM sampled around Iceland in summer 2011. In the SPM sampled during the 2012 transect, small amounts of long-chain 1,15-, 1,14- and 1,13-diols were detected. This suggests that July is not a period of high productivity for long-chain diol producers around Iceland. However, long-chain 1,13- and 1,15-diol mass fluxes were relatively high during July, September and October 2011 and from May to July 2012 (Fig. 3e), suggesting that diol producers were present, although not at the time, depths or exact locations as where the SPM was sampled. This suggests a patchy distribution of diol producers. Indeed, it has been observed that diatom distribution and composition around Iceland is highly variable and strongly influenced by different environmental variables and particularly by summer sea surface temperature (Jiang et al., 2001).

Interestingly, the fluxes of 1,13- and 1,15-diols were always much lower than C28 and C30 saturated and monounsaturated 1,14-diol fluxes, as well as 3 orders of magnitude lower than those of alkenones and GDGTs. This indicates that 1,13- and 1,15-diols and their producers are not abundant in this environment. The presence of C28 and C30 saturated and monounsaturated 1,14-diol fluxes suggests that Proboscia diatoms have bloomed in late summer and autumn and late spring and summer. This is further supported by the identification of C27 and C29 mid-chain hydroxyl methyl alkanoates in the sedimenting particles (data not shown), which are also biomarker lipids from Proboscia diatoms (Sinninghe Damsté et al., 2003). Interestingly, trace amounts of both C28 and C30 1,13-diols have been identified in Proboscia species (Rampen et al., 2007), and the relatively high abundance of saturated and monosaturated 1,14-diols, combined with a similar flux pattern as that of the 1,13 and 1,15-diols, suggests that Proboscia may also be a source for the 1,13- and 1,15-diols in this area.

Where LDI values could be calculated for the SPM, temperatures were substantially lower than satellite SSTs (Fig. 2b), exceeding the calibration error of 2 ∘C (Rampen et al., 2012). For the sediment trap, LDI values were not always calculated due to the nonquantifiable amount of C30 1,15-diol. For cases where it was possible, the temperature values were, like the SPM, much lower than satellite temperatures (Fig. 4, brown line and open squares; Table 3). The eukaryotic phytoplankton generally responsible for the production of 1,13- and 1,15-diols are likely eustigmatophyte algae, autotrophs living in the upper photic zone (Volkman et al., 1992; Rampen et al., 2012), and on a global scale, the LDI correlates best with late summer and early autumn SST (cf. Rampen et al., 2012). Thus, a contribution from colder deep water is unlikely to explain the low temperatures observed with the LDI. A similar observation is made for the LDI-derived SSTs in the surface sediments around Iceland which are always lower than annual mean SST, even when compared with the coldest season, i.e., winter mean SST (Fig. 6i, j). Furthermore, there is no correlation of LDI values with SST. Also in the surface sediments, relatively low abundances of 1,13- and 1,15-diols compared to 1,14-diols were observed. The mismatch of LDI values with temperature as well as the low abundances of long-chain 1,13- and 1,15-alkyl diols reinforce the hypothesis that Proboscia diatoms seem to be at least a partial source of 1,13- and 1,15-diols in the Iceland region. This suggests that the LDI may not be applicable in this region. Therefore, we advise that, if 1,14-diols dominate the distributions of long-chain alkyl diols, the LDI should be applied with great caution.