Long-chain diols in settling particles in tropical oceans: insights into sources, seasonality and proxies

Abstract. In this study we analyzed sediment trap time series from five tropical sites
to assess seasonal variations in concentrations and fluxes of long-chain
diols (LCDs) and associated proxies with emphasis on the long-chain diol
index (LDI) temperature proxy. For the tropical Atlantic, we observe that
generally less than 2 % of LCDs settling from the water column are
preserved in the sediment. The Atlantic and Mozambique Channel traps reveal
minimal seasonal variations in the LDI, similar to the two other lipid-based
temperature proxies TEX86 and U37K′. In addition,
annual mean LDI-derived temperatures are in good agreement with the annual
mean satellite-derived sea surface temperatures (SSTs). In contrast, the LDI
in the Cariaco Basin shows larger seasonal variation, as do the TEX86
and U37K′. Here, the LDI underestimates SST during the
warmest months, which is possibly due to summer stratification and the
habitat depth of the diol producers deepening to around 20–30 m. Surface
sediment LDI temperatures in the Atlantic and Mozambique Channel compare well
with the average LDI-derived temperatures from the overlying sediment traps,
as well as with decadal annual mean SST. Lastly, we observed large seasonal
variations in the diol index, as an indicator of upwelling conditions, at
three sites: in the eastern Atlantic, potentially linked to Guinea Dome
upwelling; in the Cariaco Basin, likely caused by seasonal upwelling; and in
the Mozambique Channel, where diol index variations may be driven by
upwelling from favorable winds and/or eddy migration.


sured in marine surface sediments is correlated with SST, and since then its application in paleoenvironmental studies has increased (see, e.g., review by Tierney, 2014). However, research has shown that despite their highest abundance being recorded in the upper 100 m of the water column, Thaumarchaeota can be present down to a depth of 5000 m (Karner et al., 2001;Herndl et al., 2005). Accordingly, GDGTs may be found in high concentrations below a depth of 100 m (e.g., Sinninghe Wuchter et al., 2005), and several studies have indicated that TEX 86 might be more reflective of subsurface temperatures in some regions (e.g., Huguet et al., 2007;Lopes dos Santos et al., 2010;Kim et al., 2012Kim et al., , 2015Schouten et al., 2013;Chen et al., 2014;Tierney et al., 2017;see Zhang and Liu, 2018 for review).
Most recently a SST proxy based on the distribution of long-chain diols (LCDs), called the long-chain diol index, or LDI has been proposed . This index is a ratio of 1,13-and 1,15-diols (i.e., alcohol groups at position C-1 and C-13 or C-15), and the analysis of globally distributed surface sediments has revealed that this index strongly correlates with SST. Since then, the index has been applied in several paleoenvironmental studies (e.g., Naafs et al., 2012;Lopes dos Santos et al., 2013;Jonas et al., 2017;Warnock et al., 2018). However, large gaps still remain in the understanding of this proxy. The largest uncertainty is that the main marine producer of LCDs is unknown. Although these diols have been observed in cultures of certain marine eustigmatophyte algae (e.g., Volkman et al., 1992Volkman et al., , 1999Méjanelle et al., 2003;Rampen et al., 2014b), the LCD distributions in cultures are different from those observed in marine sediments. Furthermore, Balzano et al. (2018) combined lipid analyses with 18S rRNA gene amplicon sequencing on suspended particulate matter (SPM) and did not find a significant direct correlation between LCD concentrations and sequences of known LCD-producers. Rampen et al. (2012) observed the strongest empirical relation between surface sediment-derived LDI values and SSTs for autumn and summer, suggesting that these are the main growth seasons of the source organisms. Moreover, the strongest correlation was also observed for the upper 20 m of the water column, suggesting that the LCDs are likely produced by phototrophic algae which thrive in the euphotic zone. Nevertheless, LDI temperatures based on surface sediments reflect an integrated signal of many years, which complicates the interpretation of the LDI in terms of seasonal production and depth of export production.
One way of resolving seasonality in the LCD flux and the LDI is to analyze time series samples from sediment traps that continuously collect sinking particles in successive time intervals over periods of a year or more. Such studies have been carried out for the U K 37 as well as for the TEX 86 and associated lipids (e.g., Müller and Fischer, 2001;Wuchter et al., 2006;Huguet et al., 2007;Fallet et al., 2011;Yamamoto et al., 2012;Rosell-Melé and Prahl, 2013;Turich et al., 2013). However, very few studies have been undertaken for LCDs. Villanueva et al. (2014) carried out a sediment trap study in Lake Challa (eastern Africa) and Rampen et al. (2008) in the upwelling region off Somalia. The latter study showed that 1,14-diols, produced by Proboscia diatoms strongly increased early in the upwelling season in contrast to 1,13-and 1,15-diols; thus they can be used to trace upwelling. However, neither of these sediment trap studies evaluated the LDI.
In this study, we assess seasonal patterns of the LDI for sediment trap series at five sites: in the Cariaco Basin, in the Mozambique Channel and three sites in the tropical North Atlantic. During this assessment, we compare the LDI values to satellite-derived SST, as well as results obtained for other temperature proxies, i.e., the TEX H 86 and U K 37 . Moreover, for the Atlantic and Mozambique Channel, we compare the sediment trap proxy signals with those preserved in the underlying sediments, after settling and burial. Finally, we assess the applicability of the diol index, based on 1,14-diols produced by Proboscia diatoms , as a tracer of upwelling and/or productivity in these regions.

Tropical North Atlantic
The ocean current and wind patterns of the tropical Atlantic are mostly determined by the seasonal latitudinal shift of the intertropical convergence zone (ITCZ; Fig. 1). The ITCZ migrates southward during boreal winter, and northward during boreal summer. During summer, the southeast trade winds prevail, whereas during winter the northeast trade winds intensify. The northeast trade winds drive the North Equatorial Current (NEC) which flows westward. South of the NEC, the North Equatorial Countercurrent (NECC) flows towards the east (Stramma and Schott, 1999). The South Equatorial Current (SEC) flows westward and branches off in the North Brazil Current (NBC; Stramma and Schott, 1999). When the ITCZ is in the north, the NBC retroflects off the South American coast and is carried eastward into the NECC, and thus into the western tropical Atlantic (e.g., Richardson and Reverdin, 1987). North of the NBC, the Guiana Current (GC) disperses the outflow from the Amazon River towards the Caribbean Sea. (Müller-Karger et al., 1988. However, during boreal summer the NBC may retroflect, carrying the Amazon River plume far into the western Atlantic (e.g., Lefèvre et al., 1998;Coles et al., 2013). In fact, every late summer/autumn, the Amazon River outflow covers around 2 × 10 6 km 2 of the western North Atlantic, and the river delivers approximately half of all freshwater input into the tropical Atlantic (see Araujo et al., 2017, and references therein). Two of the moorings in the tropical North Atlantic (M2 and M4) contain an upper ("U") and a lower ("L") trap, shown in the bathymetric section below (b) with traps depicted as red triangles and surface sediments shown as black crosses. A similar section profile is shown for the Mozambique Channel (c), where the sediment trap and the surface sediments are also indicated. All maps/sections were generated using Ocean Data View (Schlitzer, 2015). The approximate seasonal positions of the ITCZ are indicated, in addition to the North Equatorial Current (NEC), the North Equatorial Countercurrent (NECC), the South Equatorial Current (SEC), the Mauritania Current (MC), the Guinea Dome (GD), the North Brazil Current (NBC) and the Guiana Current (GC).
The eastern tropical North Atlantic is characterized by upwelling caused by the interaction between the trade winds and the movement of the ITCZ. Cropper et al. (2014) measured upwelling intensity along the northwest African coastline between 1981 and 2012, in terms of wind speed, SST and other meteorological data. They recognized three latitudinal zones: weak permanent annual upwelling north of 26 • N, strong permanent upwelling between 21 and 26 • N, and seasonal upwelling between 12 and 19 • N related to the seasonal migration of the trade winds. Southeast of Cape Verde, largescale cyclonic circulation forms the Guinea Dome (GD; Fig. 1), which centers around 10 • N, 22 • W (Mazeika, 1967), i.e., close to mooring site M1. The GD is a thermal upwelling dome, formed by near-surface flow fields associated with the westward NEC, the eastward NECC and the westward North Equatorial Undercurrent (NEUC) (Siedler et al., 1992). It forms a cyclonic circulation as a result of the eastward flowing NECC and the westward flowing NEC (Rossignol and Meyrueis, 1964;Mazeika, 1967). The GD develops from late spring to late fall due to the northern ITCZ position and the resulting Ekman upwelling, but shows significant interannual variability (Siedler et al., 1992;Yamagata and Iizuka, 1995;Doi et al., 2009) judging from general ocean circulation models. According to Siedler et al. (1992), upwelling is most intense between July and October when the ITCZ is in the GD region and the NECC is strongest.
At three sites, we analyzed five sediment trap series along a longitudinal transect in the North Atlantic (∼ 12 • N) to determine seasonal variations in the LDI. This transect has been previously studied for Saharan dust deposition in terms of grain sizes (van der Does et al., 2016), as the tropical North Atlantic receives approximately one-third of the wind-blown Saharan dust (e.g., Duce et al., 1991;Stuut et al., 2005), which might potentially act as fertilizer because of the high iron levels (e.g., Martin and Fitzwater, 1988;Korte et al., 2017;Guerreiro et al., 2017;Goudie and Middleton, 2001, and references therein). Furthermore, Korte et al. (2017) assessed mass fluxes and mineralogical composition, Guerreiro et al. (2017) measured coccolith fluxes for two of the time series, and Schreuder et al. (2018a, b) measured long-chain nalkanes, long-chain n-alkanols and fatty acids, and levoglucosan for the same sediment trap samples and surface sediments as analyzed in this study.
At site M1 (12.00 • N, 23.00 • W), the sediment trap, referred to as M1U, was moored at a water depth of 1150 m (Fig. 1). This mooring is located in the proximity of the Guinea Dome; therefore, it might potentially be influenced by seasonal upwelling. At station M2 (13.81 • N, 37.82 • W), two sediment traps were recovered, i.e., an "upper" (M2U) trap at a water depth of 1235 m, and a "lower" (M2L) trap at a depth of 3490 m. Lastly, at mooring station M4 (12.06 • N, 49.19 • W), an upper and lower trap series were also recov-ered and analyzed (M4U and M4L), at depths of 1130 and 3370 m, respectively. This mooring site may be seasonally affected by Amazon River discharge (van der Does et al., 2016;Korte et al., 2017;Guerreiro et al., 2017;Schreuder et al., 2018a). All sediment traps were equipped with 24 sampling cups, which sampled synchronously over 16-day intervals from October 2012 to November 2013, using HgCl 2 as a biocide and borax as a pH buffer to prevent in situ decomposition of the collected material.

Mozambique Channel
The Mozambique Channel is located between Madagascar and Mozambique and is part of the Agulhas Current system hugging the coast of South Africa (Lutjeharms, 2006). The Agulhas Current system is an important conveyor in the transport of warm and salty waters from the Indian Ocean to the Atlantic Ocean (Gordon, 1986;Weijer et al., 1999;Peeters et al., 2004). The northern part of the channel is also influenced by the east African monsoon winds (Biastoch and Krauss, 1999;Saetre and da Silva, 1984;Malauene et al., 2014). Between September and March, these winds blow from the northeast, parallel to the Mozambique coastline, favoring coastal upwelling. Additionally, the Mozambique Channel is largely influenced by fast-rotating, mesoscale eddies that migrate southward towards the Agulhas region. Using satellite altimetry, Schouten et al. (2003) observed four to six eddies on average, ca. 300 km in diameter, propagating yearly from the central Mozambique Channel (15 • S) toward the Agulhas area (35 • S) between 1995 and 2000. Seasonal upwelling occurs off northern Mozambique (between ca. 15 and 18 • S) (Nehring et al., 1984;Malauene et al., 2014), from August to March with a dominant period of about 2 months although periods of 1-4 weeks have also been observed (Malauene et al., 2014).
The sediment trap was moored at 16.8 • S and 40.8 • E, at a water depth of 2250 m ( Fig. 1; Fallet et al., 2010Fallet et al., , 2011 and was the same type as that used for the North Atlantic transect. We analyzed the LCD proxies for two respective time intervals: the first interval covered ca. 3.5 years, from November 2003 to September 2007, with a sampling interval of 21 days. The second interval covered another year, between February 2008 and February 2009, with a sampling interval of 17 days. Previously, Fallet et al. (2011) published foraminiferal, U K 37 and TEX 86 records for the first time interval, and the organic carbon content for the follow-up time series. For further details on the deployments and sample treatments, we refer to Fallet et al. (2011Fallet et al. ( , 2012. The two surface sediments are located across the narrowest transect between Mozambique and Madagascar, and were analyzed for U K 37 and TEX 86 by Fallet et al. (2012) and for LCDs by Lattaud et al. (2017b).

Cariaco Basin
The Cariaco Basin is one of the largest marine anoxic basins (Richards, 1975), located on the continental shelf of Venezuela. The basin is characterized by permanent stratification and is strongly influenced by the migration of the intertropical convergence zone (ITCZ). During late autumn and winter, the ITCZ migrates to the south which results in decreased precipitation and trade wind intensification that in turn induces upwelling and surface water cooling. This seasonal upwelling is a major source of nutrients that leads to strong phytoplankton growth along the Venezuelan coast (e.g., Müller-Karger et al., 2001;Thunell et al., 2007). Between August and October, the ITCZ moves northward again, resulting in a rainy season and diminishing the trade winds which inhibits upwelling. During this wet season the contribution of terrestrially derived nutrients is higher. Due to the prevalent anoxic conditions in the basin, there is no bioturbation; this has resulted in the accumulation of laminated sediments that provide excellent annually to decadally resolved climate records (e.g., Peterson et al., 1991;Hughen et al., 1996Hughen et al., , 1998. Moreover, in November 1995, a time series experiment started to facilitate research on the link between biogeochemistry and the downward flux of particulate material under anoxic and upwelling conditions (Thunell et al., 2000). This project (CARIACO; http://imars.marine.usf.edu/ cariaco, last access: November 2018) involved hydrographic cruises (monthly), water column chemistry measurements and sediment trap sampling (every 14 days). One mooring containing four automated sediment traps (Honjo and Doherty, 1988) was deployed at 10.50 • N and 64.67 • W, at a bottom depth of around 1400 m. These traps were moored at a depth of 275 m, just above the oxic/anoxic interface (Trap A), at 455 m (Trap B), at 930 m (Trap C) and at 1255 m (Trap D). All traps contained a 13-cup carousel which collected sinking particles over 2 weeks, and were serviced every 6 months. For further details on trap deployment and recovery, and sample collection, storage and processing we refer the reader to Thunell et al. (2000) and Goñi et al. (2004). In addition to the sediment trap sampling, the primary productivity of the surface waters was measured every month using 14 C incubations . For this study, we investigated two periods, i.e., May 1999-May 2000 and July 2002-July 2003 for traps A and B. These years include upwelling and non-upwelling periods, as well as a disastrous flooding event in December 1999(Turich et al., 2013. Turich et al. (2013) identified the upwelling periods, linked to the migration of the ITCZ, as indicated by decreasing SST in the CTD (temperature at −1 m water depth) and satellite-based measurements (indicated by grey boxes in Figs. 8 and 10), and shoaling of the average depths of primary production and increased primary production. Moreover, Turich et al. (2013) evaluated the U K 37 and TEX 86 proxies for the same two time series for which we analyzed the LCD proxies.

Instrumental data
Satellite SST, precipitation and wind speed time series of the M1, M2 and M4 moorings in the Atlantic derive from Guerreiro et al. (2017Guerreiro et al. ( , 2019, who retrieved these data from the Ocean Biology Processing Group (OBPG, 2014;Frouin et al., 2003), the Goddard Earth Sciences Data and Information Services Center (2016) (Huffman et al., 2007;Xie and Arkin, 1997) and the NASA Aquarius project (2015a, b) (Lee et al., 2012) (see Supplement of Guerreiro et al., 2017 for detailed references). The SST and chlorophyll a time series data for the Mozambique Channel were adapted from Fallet et al. (2011), who retrieved these data from the Giovanni database (for details see Fallet et al., 2011). Surface sediment proxy temperatures were compared to annual mean SST estimates derived from the World Ocean Atlas (2013) (decadal averages from 1955 to 2012; Locarnini et al., 2013). Sea surface temperature data for the Cariaco Basin were adopted from Turich et al. (2013) and combined with additional CTD temperatures from the CARIACO time series database for depths of 2, 5, 10, 15 and 20 m (http://www.imars.usf.edu/ CAR/index.html, last access: November 2018; CARIACO time series composite CTD profiles; lead principal investigator: Frank Müller-Karger).

Mozambique Channel
Aliquots of the sediment trap samples from the Mozambique Channel were previously extracted and analyzed by Fallet et al. (2011) andFallet et al. (2012), respectively. The sediment trap material was extracted by ultrasonication using a mixture of DCM/MeOH (2 : 1; v/v), dried over Na 2 SO 4 , and separated into apolar, ketone and polar fractions via alumina pipette column chromatography, by eluting with hexane/DCM (9 : 1; v/v), hexane/DCM (1 : 1; v/v) and DCM/MeOH (1 : 1; v/v), respectively. These existing polar fractions of the sediment trap material were silylated (as described above), dissolved in ethyl acetate and reanalyzed for LCDs by GC-MS. As no record was kept of the division of the extracts and polar fractions into aliquots, we report the results in relative abundance rather than concentrations and fluxes of diols.

Cariaco Basin
Sediment trap material was extracted as described by Turich et al. (2013). Briefly, 1/16 aliquots of the trap samples were extracted by means of Bligh-Dyer extraction with sonication using a phosphate buffer and a trichloroacetic acid (TCA) buffer. The extracts were then separated by adding 5 % NaCl in solvent-extracted distilled deionized water, the organic phase was collected, and the aqueous phase was extracted twice more. The extracts were pooled and dried over Na 2 SO 4 and separated by means of Al 2 O 3 column chromatography, eluting with hexane/DCM (9 : 1; v/v), DCM/MeOH (1 : 1; v/v) and MeOH. For this study, the DCM/MeOH (1 : 1; v/v) fraction was silylated (as described above), dissolved in ethyl acetate, and analyzed for LCDs using GC-MS. Similar to the Mozambique Channel samples, no record was kept of the division of extracts and polar fractions into aliquots; thus, we report the results in relative abundance.

GDGTs
The GDGT fractions of the surface sediments and sediment traps SPM samples of the tropical North Atlantic were analyzed for GDGTs using ultra-high-performance liquid chromatography mass spectrometry (UHPLC-MS). We used an Agilent 1260 HPLC, which was equipped with an automatic injector, interfaced with a 6130 Agilent MSD and HP Chem-Station software according to Hopmans et al. (2016). Compound separation was achieved by two silica BEH HILIC columns in tandem (150 mm × 2.1 mm; 1.7 µm; Waters AC-QUITY) in normal phase, at 25 • C. GDGTs were eluted isocratically for 25 min with 82 % A and 18 % B, followed by a linear gradient to 35 % B in 25 min and finally a linear gradient to 100 % B in the last 30 min. "A" denotes hexane; "B" denotes hexane/isopropanol (9 : 1; v/v). The flow rate was constant at 0.2 mL min −1 , and the injection volume was 10 µL. The APCI-MS (atmospheric pressure chemical ionization -mass spectrometry) conditions are described by Hopmans et al. (2016). Detection and quantification of GDGTs was achieved in selected ion monitoring mode (SIM) mode of the protonated molecules ([M + H] + ) of the GDGTs. We used a mixture of crenarchaeol and C 46 GDGT (internal standard) to assess the relative response factor, which was used for quantification of the GDGTs in the samples (cf. Huguet et al., 2006).
Sea surface temperatures were calculated by means of the TEX H 86 as defined by Kim et al. (2010), which is a logarithmic function of the original TEX 86 index : where the numbers indicate the number of cyclopentane moieties of the isoprenoid GDGTs, and Cren reflects an isomer of crenarchaeol, i.e., containing a cyclopentane moiety with a cis stereochemistry . The TEX H 86 values were translated to SSTs using the core-top calibration of Kim et al. (2010): The branched isoprenoid tetraether (BIT) index is a proxy for the relative contribution of terrestrially derived organic carbon (Hopmans et al., 2004). We calculated the modified version as reported by de Jonge et al. (2014de Jonge et al. ( , 2015 which is based on the original index as proposed by Hopmans et al. (2004), but includes the 6-methyl brGDGTs: where the numbers reflect different branched GDGTs (see Hopmans et al., 2004) and Cren reflects crenarchaeol. The branched GDGTs were always around the detection limit in the Atlantic samples, implying a BIT index of around zero and thus minimal influence of soil organic carbon (Hopmans et al., 2004); therefore, the BIT index is not discussed any further.

LCAs
The ketone fractions of the surface sediments and sediment traps samples of the tropical North Atlantic were analyzed for LCAs on an Agilent 6890N gas chromatograph (GC) with flame ionization detection (FID) after being dissolved in ethyl acetate. The GC was equipped with a fused silica column with a length of 50 m, a diameter of 0.32 mm and a coating of CP Sil-5 (film thickness = 0.12 µm). Helium was used as the carrier gas, and the flow mode was a constant pressure of 100 kPa. The ketone fractions were injected on-column at a starting temperature of 70 • C, which was increased by 20 • C min −1 to 200 • C followed by 3 • C min −1 until the final temperature of 320 • C was reached. This end temperature was held for 25 min. The U K 37 index was calculated according to Prahl and Wakeham (1987): The U K 37 values were translated to SST following the calibration of Müller et al. (1998): We also applied the recently proposed BAYSPLINE Bayesian calibration of Tierney and Tingley (2018). They and others have shown that the U K 37 estimates substantially attenuate above temperatures of 24 • C (e.g., Conte et al., 2001;Goñi et al., 2001;Sicre et al., 2002). The Bayesian calibration moves the upper limit of the U K 37 calibration from approximately 28 to 29.6 • C at unity. As our traps are located in tropical regions with SSTs > 24 • C, we applied this calibration as well.

LCDs
The silylated polar fractions were injected on-column on an Agilent 7890B GC coupled to an Agilent 5977A MS. The starting temperature was 70 • C, and was increased to 130 • C by 20 • C min −1 , followed by a linear gradient of 4 • C min −1 to an end temperature of 320 • C, which was held for 25 min. A total of 1 µL was injected, and separation was achieved on a fused silica column (25 × 0.32 mm) coated with CP Sil-5 (film thickness 0.12 µm). Helium was used as the carrier gas with a constant flow of 2 mL min −1 . The MS operated with an ionization energy of 70 eV. Identification of LCDs was carried out in full scan mode, scanning between m/z 50 and 850, based on characteristic fragmentation patterns (Volkman et al., 1992;Versteegh et al., 1997). Proxy calculations and LCD quantifications were performed via analysis (in SIM mode) of the characteristic fragments (m/z 299, 313, 327 and 341; Rampen et al., 2012; m/z 187 for internal diol standard). For quantification of LCDs in the sediment traps and seafloor sediments of the tropical Atlantic, the peak areas of the LCDs were corrected for the average relative contribution of the selected SIM fragments to the total ion counts, i.e., 16 % for the saturated LCDs, 9 % for unsaturated LCDs and 25 % for the C 22 7,16-diol internal standard.
Sea surface temperatures were calculated using the LDI, according to Rampen et al. (2012): These LDI values were converted into SSTs using the following equation : Upwelling conditions were reconstructed using the diol index as proposed by Rampen et al. (2008): In 2010, Willmott et al. introduced an alternative diol index, which is defined as the ratio of 1,14-diols over 1,13diols. As the index of Rampen et al. (2008) includes the C 30 1,15-diol, it can be affected by temperature variation; therefore, we would normally prefer to use the index of Willmott et al. (2010). However, we often did not detect the C 28 1,13-diol, or it co-eluted with cholest-5-en-7-one-3β-ol, compromising the calculation of the diol index of . Moreover, the temperature variations in all three sediment traps are minimal as recorded by the LDI. Accordingly, we chose to apply the diol index according to Rampen et al. (2008). Potential fluvial input of organic carbon was determined by the fractional abundance of the C 32 1,15-diol (de Bar et al., 2016;Lattaud et al., 2017a): The fractional abundance of the C 32 1,15-diol was always lower than 0.23, suggesting low input of river-derived organic carbon (Lattaud et al., 2017a).

Tropical North Atlantic
We analyzed sediment trap samples from a longitudinal transect (∼ 12 • N) in the tropical North Atlantic (two upper traps at a depth of ca. 1200 m, and three lower traps at ca. 3500 m; Fig. 1), covering November 2012-November 2013, as well as seven underlying surface sediments, for LCDs, LCAs and GDGTs. Below we present the results for these lipid biomarkers and associated proxies.

LCDs
The LCDs detected in the sediment trap samples and surface sediments from the tropical North Atlantic (Fig. 2) are the  (Fig. 1). When no bar is shown the LCD was not detected in the surface sediments.
The LDI ranged between 0.95 and 0.99 in all traps, corresponding to temperatures of 26.0 to 27.3 • C with no particular trends (Fig. 4). For most M2 and M4 samples the C 28 1,13-diol was below the quantification limit; hence, LDI was always around unity, corresponding to 26.9 to 27.3 • C (Fig. 4), whereas in other samples the C 28 1,13-diol co-eluted with cholest-5-en-7-one-3β-ol, prohibiting the calculation of the LDI and the diol index (Figs. 4, 5). The flux-weighted annual average LDI-derived SSTs were 26.6 • C for M1U, and 27.1 • C for M2U, M2L, M4U and M4L. The underlying sediment was very similar, with LDI values of between 0.95 and 0.98 corresponding to 26.0 and 26.9 • C (Fig. 6). The diol index varied from 0.03 to 0.30 in M1U, showing a pronounced maximum during spring (Fig. 5a). The diol index at M2 ranged between 0.01 and 0.03 without an evident pattern, whereas the diol index at M4 ranged from 0.01 to 0.10 and showed the same pattern in the lower and upper trap, with the highest values during spring (ca. 0.1), followed by a gradual decrease during summer (Fig. 5e, f).

LCAs
We detected C 37 , C 38 and C 39 long-chain alkenones in the sediment trap and surface sediments. The C 37 : 3 alkenone was generally around the limit of quantification for the M2L and M4L traps, and below the limit of quantification for four out of the seven surface sediment samples, whereas the C 37 : 2 alkenone was always sufficiently abundant. The annual mean fluxes of the C 37 LCAs were 4.3 (±3.5) µg m −2 d −1 for M1U, 1.2 (±0.9) µg m −2 d −1 and 0.4 (±0.2) µg m −2 d −1 for M2U and M2L, respectively, and 2.9 (±5.1) µg m −2 d −1 and 1.2 (±2.0) µg m −2 d −1 for M4U and M4L, respectively. The concentrations of the C 37 LCAs in the underlying surface sediments ranged between 0.02 and 0.41 µg g −1 . At M4, the two total mass flux peaks at the end of April and during October/November were also clearly pronounced in the C 37 alkenone fluxes (Figs. 3d, e, 5g), as well as the in-creased signal in the cup reflecting the beginning of May, which followed the cup which recorded the peak in total mass flux at the end of April. The U K 37 varied from 0.87 to 0.93, corresponding to 25.1 to 27.0 • C (Fig. 6b) for three out of seven surface sediments in which the C 37 : 3 was above the quantification limit. The flux-weighted average SSTs were 26.1 • C for M1U, 25.7 and 26.4 • C for M2U and M2L, respectively, and 28.2 and 27.5 • C for M4U and M4L, respectively (Fig. 6). SST variations per sediment trap were generally within a 2-3 • C range (Fig. 4) with no apparent trends.

Cariaco Basin
We analyzed LCDs for two time series (May 1999-May 2000 and July 2002-July 2003) from the upper (Trap A; 275 m) and the lower (Trap B; 455 m) trap in the Cariaco Basin. The main LCDs detected for both time series are the C 28 1,14-, C 30 1,14-, C 30 : 1 1,14-, C 28 1,13-, C 30 1,15-and C 32 1,15-diols, as well as the C 29 12-OH fatty acid. For some samples we did not compute the LDI, as the C 28 1,13-diol co-eluted with cholest-5-en-7-one-3β-ol. In a similar fashion to the Mozambique Channel, the C 28 1,14-diol was not affected by this co-elution due to its much higher abundance compared with the C 28 1,13-diol; therefore, the diol index was therefore

LCD sources and seasonality
The 1,14 diols can potentially be derived from two sources: Proboscia diatoms (Sinninghe Rampen et al., 2007) or the dictyochophyte Apedinella radians (Rampen et al., 2011). The non-detection of the C 32 1,14-diol, which is a biomarker for Apedinella radians (Rampen et al., 2011), and the detection of the C 30 : 1 1,14 diol and C 29 12-OH fatty acid, which are characteristic of Proboscia diatoms , suggests that Proboscia diatoms are most likely the source of 1,14-diols in the tropical North Atlantic, the Mozambique Channel and the Cariaco Basin.
In the Cariaco Basin, the diol index shows a strong correlation (visually as correlation analysis was not possible due to differently spaced data in time) with primary production rates, suggesting that Proboscia productivity was synchronous with total productivity (Fig. 8), although for the 1999-2000 time series there is a disagreement during January/February. Primary productivity in the Cariaco Basin is largely related to seasonal upwelling which occurs between November and May when the ITCZ is at its southern posi- tion. Hence, the diol index seems to be an excellent indicator of upwelling intensity in the Cariaco Basin.
The index also shows considerable variation over time in the Mozambique Channel (Fig. 7b). Previous studies have shown that upwelling occurs in the Mozambique Channel between ca. 15 and 18 • S (Nehring et al., 1984;Malauene et al., 2014), i.e., at the location of our sediment trap. Upwelling is reflected by cool water events and slightly enhanced chlorophyll a levels; Malauene et al. (2014) observed cool water events at ca. 2-month intervals although periods of 8 to 30 days were also noted. The two main potential forcing mechanisms for upwelling in the Mozambique Channel are the east African monsoon winds and the mesoscale eddies migrating through the channel. Fallet et al. (2011) showed that subsurface temperature, current velocity and the depth of surface-mixed layer all revealed a dominant periodicity of four to six cycles per year, which is the same frequency as that of the southward migration of mesoscale ed-dies in the channel (Harlander et al., 2009;, implying that eddy passage strongly influences the water mass properties. Wavelet analysis of the diol index for the 2003-2007 period (Fig. S1 in the Supplement) revealed short periods, occurring around January of 2004, 2005 and 2006, of significant (above the 95 % confidence level) variability at about bimonthly frequencies (60-day period). Both the frequency (bimonthly) and the timing (boreal winter) of the observed time periods of the enhanced diol index variability are similar to those of the cool water events as observed by Malauene et al. (2014), associated with upwelling (Fig. 7b). The strongest variability of the diol index at about bimonthly frequencies occurred in the first half of 2006. During the same period, salinity time series showed the passage of several eddies that had a particularly strong effect on the upper layer hydrography . Malauene et al. (2014) showed that neither upwelling-favorable winds, nor passing eddies, can independently explain the observed Figure 6. Flux-weighted average (annual) proxy results for the sediment traps compared with the underlying sediments (crosses) and annual mean SST (red line; specific for the coordinates of the surface sediments; World Ocean Atlas 2013 1/4 • grid resolution). Panels (a), (b) and (c) show the LDI, U K 37 and TEX 86 temperature results, respectively. Triangles reflect sediment trap results (red represents upper/∼ 1200 m; blue represents lower/∼ 3500 m), and crosses represent surface sediments. In the case of the U K 37 and TEX 86 , the green and purple triangles and grey crosses reflect the temperatures calculated using the BAYSPLINE and BAYSPAR models (Tierney and Tingley, 2014, 2015, 2018, whereas the other temperatures were calculated using the Müller et al. (1998) and Kim et al. (2010;TEX H 86 ) calibrations, respectively. Panel (d) shows the flux-weighted average diol index values for the sediment traps and the diol index estimates for the surface sediments. upwelling along the northern Mozambique coast. The two processes may act together, and both strongly influence the upper water layer and the organisms living there, potentially including the LCD producers.
The least (seasonal) variation in the diol index is observed at M2 in the tropical North Atlantic (Fig. 5b, c), which is likely due to its central open ocean position, associated with relatively stable, oligotrophic conditions . In contrast, M4 and M1 are closer to the South Amer- ican and west African coast, respectively, and thus are potentially under the influence of Amazon river runoff and upwelling, respectively, and specific wind and ocean circulation regimes (see Sect. 2.1.1). However, at M4, the diol index is also low (max. 0.1), suggesting low Proboscia productivity (Fig. 5e, f). At M1, in contrast, we observe enhanced values for the diol index of up to ∼ 0.3 during spring (Fig. 5a).
Most likely, an upwelling signal at this location is associated with the seasonal upwelling of the Guinea Dome. This upwelling is generally most intense between July and October (Siedler et al., 1992), due to the northward movement of the ITCZ and the resulting intensified Ekman upwelling. Specifically, during this period, the trade winds are weaker, atmospheric pressure is lower and the regional wind stress is favorable to upwelling of the North Equatorial Undercurrent (Voituriez, 1981). Indeed, a decrease in wind speed and increased precipitation from summer to autumn was observed ( Fig. 5a) which confirms that during these seasons the ITCZ was indeed in a northern position, and that during 2013 the upwelling associated with the Guinea Dome was most favored between July and October. The timing of the diol index peak, i.e., between March and June, is consistent with previous sediment trap studies elsewhere which have shown that Proboscia diatoms and 1,14-diols are typically found during pre-upwelling or early upwelling periods (Koning et al., 2001;Smith, 2001;Sinninghe Damsté et al., 2003;Rampen et al., 2007). The surface sediment at 22 • W just east of M1 also reveals the highest diol index (0.22), likely due to its closer vicinity to the Guinea Dome center. Several studies have reported P. alata diatoms offshore of northwest Africa (Lange et al., 1998;Treppke et al., 1996;Crosta et al., 2012;Romero et al., 1999), pointing to P. alata as a plausible source organism. The sedimentary annual diol indices compare well with the sediment trap indices (Fig. 6d), which is consistent with the results of Rampen et al. (2008). Our results clearly show that the diol index reflects different things in different regions. This is due to the ecology of Proboscia spp. where blooms occur during stratification, during early upwelling and post-bloom, and from high nutrients to low nutrients (see Rampen et al., 2014a;references in Table 1). Therefore, the type of conditions reflected by the diol index is specific for every region.
To assess variations in the seasonal production of 1,13and 1,15-diols in the tropical Atlantic, for which we have the most complete data set, we calculated the flux-weighted 1,13-and 1,15-diol concentrations for the different traps, and summed these per season (Fig. 9). Highest production is observed in autumn, followed by spring and summer, with the lowest production during winter (∼ 60 % compared with autumn). This is in agreement with Rampen et al. (2012) who observed, for an extensive set of surface sediments, the strongest correlation between LDI and SST for autumn, suggesting that production of the source organisms of the LDI mainly occurs during autumn. At M4, there are two evident peaks in the 1,13-and 1,15-diol fluxes at the end of April and October 2013. These maxima correlate with peaks in other lipid biomarker fluxes (i.e., 1,14-diols, C 37 alkenones and iGDGTs), total mass flux, calcium carbonate (CaCO 3 ), OM (organic matter) and the residual mass flux which includes the deposition flux of Saharan dust . According to Guerreiro et al. (2017), the maximum in total mass flux at the end of April 2013 is likely caused by enhanced export production due to nutrient enrichment as a result of wind-forced vertical mixing. The peak at the end of October 2013, is likely associated with discharge from the Amazon River. Moreover, both peaks are concomitant with prominent dust flux maxima, suggesting that Saharan dust also acted as nutrient fertilizer Guerreiro et al., 2017). Guirreirro et al. (2017) suggested that dur- ing the October-November event the Amazon River may not only have acted as nutrient supplier, but also as buoyant surface density retainer of dust-derived nutrients in the surface waters, resulting in the development of algal blooms within just a few days, potentially explaining the peak 1,13-and 1,15-diol fluxes, as well as the peak fluxes of the other lipid biomarkers. However, they might also partially result from enhanced particle settling, caused by factors such as dust ballasting or faecal pellets of zooplankton (see Guerreiro et al., 2017, and references therein). This agrees with the results of Schreuder et al. (2018a) that show that the n-alkane flux also peaks concomitantly with the peaks in total mass flux and biomarkers, whereas n-alkanes are terrestrially derived (predominantly transported by dust); therefore, increased deposition can not result from increased primary productivity in the surface waters.
The C 37 alkenone flux at M4U also reveals these two distinct maxima at the end of April and October during 2013 (Fig. 5g). Interestingly, this flux, as well as the alkenone flux at M2U, is consistent with coccolith export fluxes of the species Emiliania huxleyi and Gephyrocapsa oceanica . In fact, when we combine the coccolith fluxes of both species, we observe strong correlations with the C 37 alkenone fluxes for both M2U and M4U ( Fig. 5d and g, respectively; r = 0.77 and 0.92 for M2U and M4U, respectively; p-values < 0.001). This implies that these two species are the main LCA producers in the tropical North Atlantic, which agrees with previous findings (e.g., Marlowe et al., 1984;Brassell, 2014;Conte et al., 1994;Volkman et al., 1995).

Preservation of LCDs
The sediment trap data from the North Atlantic can be used to assess the relative preservation of LCDs, as well as other proxy lipid biomarkers, by comparing the flux-weighted concentration in the traps with the concentrations in the surface sediments. For all four biomarker groups, i.e., C 37 alkenones, iGDGTs, 1,14-diols and 1,13-and 1,15-diols, we observe that the flux-weighted concentrations are generally higher in the upper traps (ca. 1200 m) compared with the lower traps (ca. 3500 m; Fig. 2) by a factor of between 1.2 and 4.4, implying degradation during settling down the water column. The concentrations in the surface sediments are 2 to 3 orders of magnitude lower (i.e., between 0.1 %-1.5 % of upper trap signal), implying that degradation of lipids mainly takes place at the water-sediment surface rather than in the water column. A similar observation was made for levoglucosan in these sediment traps (Schreuder et al., 2018b). Both are functionalized polar lipids with alcohol groups and thus are chemically relatively similar when compared to species such as fatty acids (carboxyl group) or n-alkanes (no functional groups). These degradation rates are likely linked to the extent of the oxygen exposure time (Hartnett et al., 1998;Hedges et al., 1999) at the seafloor (Hartnett et al., 1998;Sinninghe Damsté et al., 2002), as during settling the lipids are exposed to oxygen for weeks, whereas for surface sediments this is typically decades to centuries. Our results compare well with several other sediment trap studies which showed that LCDs, LCAs and iGDGTs generally have a preservation factor of around 1 % (surface sediment versus trap) (e.g., Prahl et al., 2000;Wakeham et al., 2002;Rampen et al., 2007;Yamamoto et al., 2012).
We also identified the C 30 and C 32 1,15-keto-ol in the Atlantic as well as in the Mozambique and Cariaco sediment traps and surface sediments. These lipids are structurally related to LCDs, occur ubiquitously in marine sediments (e.g., Versteegh et al., 1997Versteegh et al., , 2000Bogus et al., 2012;Rampen et al., 2007;Sinninghe Damsté et al., 2003;Wakeham et al., 2002;Jiang et al., 1994) and were inferred to be oxidation products of LCDs (Ferreira et al., 2001;Bogus et al., 2012;Sinninghe Damsté et al., 2003). We did not detect 1,14-ketools, which supports the hypothesis of Ferreira et al. (2001) and Sinninghe  that the silica frustules of Proboscia diatoms sink relatively fast; thus, they are exposed to oxygen for a shorter period than the producers of 1,13-and 1,15-diols, and are therefore less affected by oxidation. Alternatively, the keto-ols are not oxidation products but are produced by unknown organisms in the water column. In fact, Méjanelle et al. (2003) observed trace amounts of C 30 1,13-and C 32 1,15-keto-ols in cultures of the marine eustigmatophyte Nannochloropsis gaditana. Thus, an alternative explanation for the non-detection of 1,14-keto-ols is that, in contrast to the 1,15-keto-ols, they were not produced in the water column.
For both the tropical Atlantic and the Cariaco Basin, we observe highly similar LDI values for the upper and the lower traps. In the Atlantic there is no statistical difference between the upper and lower traps that are 2200 m apart (two-tailed p > 0.8), but we have insufficient data for the Cariaco Basin for statistical comparison (Figs. 6a,8a,e). This suggests that degradation in the water column does not affect the LDI proxy. This is in agreement with Reiche et al. (2018) who performed a short-term degradation experiment (< 1 year) and found that the LDI index was not affected by oxic exposure on short timescales. However, the oxygen exposure time on the seafloor is much longer; Rodrigo-Gámiz et al. (2016) showed for sediments in the Arabian Sea (deposited under a range of bottom water oxygen conditions) that different LCDs had different degradation rates, which compromised the LDI ratio. For the three sites in the tropical North Atlantic, we calculated the flux-weighted average proxy values for every sediment trap and compare these with the underlying surface sediments (Fig. 6a-c). For all indices, i.e., diol index, LDI, U K 37 and TEX 86 , we observe very good correspondence between the sediment trap and surface sediment values, implying minimal alteration of the proxies after settling and during burial. Similarly, for the Mozambique Channel, the mean diol index and LDI from the sediment trap (i.e., 0.41 and 0.97, respectively) are very similar to the surface sediment values (i.e., 0.42 and 0.95, respectively). In agreement with the consistent diol indices, we observe that all individual LCDs are also preserved relatively equally in the tropical Atlantic (1.2 %-4.3 % at station M1, 0.1 %-2.9 % at station M2 and 0.03 %-0.16 % at station M4). This contrasts with Rodrigo-Gámiz et al. (2016) who found that the 1,15diols have the highest degradation rate, followed by the 1,14and 1,13-diols. Only the C 32 1,15-diol seems relatively better preserved than the other LCDs at all three North Atlantic mooring sites (Fig. 2), suggesting that the C 32 1,15-diol is less impacted by degradation. The C 32 1,15-diol likely partially derives from the same source as the other 1,13-and 1,15-diols, but is also produced in fresh water systems (e.g., Versteegh et al., 1997Versteegh et al., , 2000Rampen et al., 2014b;de Bar et al., 2016;Lattaud et al., 2017a, b). Hence, the different preservation characteristics might be the result of a different source for this LCD.

Relationship between LDI and SST
In the tropical Atlantic and the Mozambique Channel, the LDI-derived SSTs show minimal variability (< 2 • C), whereas in the Cariaco Basin we observe much larger changes that range from 22.0 to 27.2 • C (Fig. 8). Both time series in the Cariaco Basin show low temperatures between November and May, associated with the seasonal upwelling and surface water cooling, and significantly higher temperatures during the rainy summer. However, during the warmest periods, the LDI temperatures are generally lower than those measured at the surface by CTD, whereas during the colder phases, the LDI agrees well with the measurements. The LDI calibration reaches unity at 27.4 • C; therefore, is not possible to resolve the highest temperatures which are between ca. 28 and 30 • C. However, the LDI-derived temperatures are sometimes well below 27.4 • C where the CTD data suggest SSTs > 28 • C. Consequently, the LDI-based temperatures agree with CTD-based SSTs within calibration error for most of the record, but during summer when the SST is highest, they are offset outside the calibration error ( T ∼ 2.5 • C). Interestingly, the U K 37 -and TEX H 86 -derived temperature trends show the same phenomenon (Turich et al., 2013;Fig. 8), where the proxy temperatures are cooler than the measured temperatures during the warmer months. However, in contrast to the U K 37 and LDI, the TEX H 86 also overestimates the SST during the cold months. For U K 37 , Turich et al. (2013) pointed out that a time lag between synthesis, export and deposition could potentially explain the difference between the proxy and CTD temperatures. However, previous analysis of plankton biomass, primary productivity, bio-optical properties and particulate organic carbon fluxes for the same time period , as well as the total mass and terrigenous fluxes assessed by Turich et al. (2013) showed the best correlation at zero-time lag on the basis of their 14-day sample interval. We compared our LDI temperature estimates with monthly CTD measurements between a depth of 0 and 50 m, the temperature at the depth of maximum primary productivity and the temperature at the chlorophyll maximum (Turich et al., 2013; http://www.imars.usf.edu/cariaco, last access: November 2018) (Fig. 10). During the upwelling season, temperatures are significantly lower due to the upward migration of isotherms, whereas during the non-upwelling period, temperatures are higher, particularly in the upper 20 m, and the water column is more stratified (Fig. 10). LDI underestimates the SST during stratification, which suggests that the LCD producers may thrive at depths of ca. 20-30 m. During upwelling, LDI temperatures agree better with SST, implying that the habitat of the LCD producers is potentially closer to the surface, coincident with the shoaling of the nutricline and thermocline (Fig. 10). However, these absolute differences in LDI temperatures are generally within the calibration error (2 • C); thus, these seasonal variations in LDI temperatures should be interpreted with caution. Turich et al. (2003) found that the U K 37 -derived temperatures agreed reasonably well with the measured temperatures at the chlorophyll maximum, which is generally found below a depth of 20 m (average depth of 30-34 m; ranging between 1 and 55 m) in the Cariaco Basin. The LDI temperatures are almost always higher than the temperatures at the chlorophyll maximum (Fig. 10), and higher than the temperatures at a depth of 30 m, implying that the LDI producers may reside in the upper 30 m of the water column; this is consistent with the results of  which showed that LDI-derived temperatures have the strongest correlation with the temperatures in the upper 20 m of the water column. This also agrees with Balzano et al. (2018) who observed the highest LCD abundances within the upper 20 m of the water column in the tropical Atlantic.
In the Mozambique Channel, the LDI temperature variations are much smaller (< 2 • C; Fig. 7a) than the seasonal SST variation -ranging between ca. 24.5 and 30.5 • C. Accordingly, during the warmest months of the year, the difference between LDI-derived and satellite-derived SST is outside the calibration error (i.e., > 2 • C). However, this is sim- ilar to the U K 37 and TEX H 86 which also did not reveal seasonal variations. This lack of seasonality was explained by the lateral advection and resuspension of fine sediment material by migrating mesoscale eddies which ending up in the deeply moored sediment trap (Fallet et al., 2011(Fallet et al., , 2012. Most likely, this also explains the lack of seasonal variation in our LDI record (Fig. 7a). Nevertheless, the average LDI temperature for the sediment trap of 26.4 • C agrees reasonably well with the annual mean satellite-derived SST of 27.6 • C for the sampled years. Additionally, there is a good agreement with the average LDI temperature of 26.0 • C for the two underlying surface sediments, as well as with the decadal average SST of 26.7 • C for 1955-2012 (Locarnini et al., 2013) given by the World Ocean Atlas (2013). For the North Atlantic, we also observe rather constant LDI temperatures during the year (Fig. 4) which contrasts with seasonal variations in satellite SSTs of ca. 3 to 5 • C. Nevertheless, differences are mostly within the calibration error, except at M1 and M2 where LDI-derived temperatures are between 0.5 and 2.8 • C higher than satellite SSTs during winter and spring. Similar to the LDI, the TEX H 86 and U K 37 -derived SSTs for the tropical Atlantic sediment traps also do not reveal clear seasonal variation. As all three proxies show minimal seasonal variability, this might indicate that the lipids are potentially allochtonous and partially derive from distant regions, resulting in an integrated average temperature signal, similar to the Mozambique Channel. Nevertheless, the flux-weighted annual LDI temperatures of the tropical Atlantic sediment traps (26.6 for M1 and 27.1 • C for M2 and M4) agree well with the annual mean satellite-derived SSTs of 26.1, 26.0 and 27.5 • C for M1, M2 and M4, respectively. Moreover, the LDI-derived temperatures in the underlying sediments (26.5, 26.6 and 26.7 • C, respectively) do not only agree well with those found in a single year in the sediment traps, but they also agree with the decadal average SSTs for 1955 to 2012 (26.2, 27.1 and 26.3 • C, respectively; Locarnini et al., 2013;Fig. 6a).

Conclusions
In this study we evaluated LCD-based proxies, particularly the LDI, in sediment trap time series from five sites in the tropical North Atlantic, the Cariaco Basin and the Mozambique Channel. For the North Atlantic we found that ca. 25 %-85 % of the export of these lipid biomarkers was preserved during settling from 1200 to 3500 m in the water column, and that generally less than 2 % was preserved in the surface sediments. Despite substantial degradation at the seafloor, likely linked to the prolonged oxygen exposure time, LCD-derived temperatures from the sediments are generally very similar to the annual mean LCD-derived temperatures in both the deep and shallow traps as well as to the annual mean SST for the specific sampling year and on decadal timescales for the specific sites. In the Cariaco Basin we observe a seasonal signal in the LDI linked to the upwelling season reflecting temperatures of the upper ca. 30 m of the water column. The LDI temperatures in the Mozambique Channel and the tropical Atlantic reveal minimal seasonal change although the seasonal SST contrasts amount to 3-5 • C. For the Mozambique Channel this is likely caused by the lateral advection of resuspended sediment by mesoscale eddy migration, a signal not substantially altered by diagenesis. Seasonal variations in the diol index are minimal in the central and western North Atlantic and 1,14-diol concentrations are rather low, implying little Proboscia diatom productivity. However, in the eastern Atlantic, closest to the African continent, the diol index attains a clear spring maximum that is likely associated with upwelling in the Guinea Dome during summer to autumn, suggesting the diol index reflects a pre-upwelling signal, consistent with the current knowledge on Proboscia ecology. In the Cariaco Basin, controlled by seasonal upwelling, the diol index reveals the same clear seasonal trend observed in primary productivity, arguing that the diol index is an excellent indicator of upwelling intensity for this location.
Author contributions. MWdB, JSSD and SS designed the experiments, and MWdB carried them out. JU carried out the time series analysis. JBWS, GJAB and RCT deployed sediment traps and collected sediment trap materials. MWdB prepared the paper with contributions from all coauthors.