Coccoliths produced by F. profunda (29–56 %) were high in November 2012 and late July 2013, with lower fluxes in October 2012, late March, late August, and late September 2013. In terms of relative abundances, F. profunda was most common in early November 2012 and least common in late April 2013 (Fig. 6a). A similar seasonal pattern was shown by G. flabellatus (18–47 %) but with a less abrupt decrease in the transition to spring compared to F. profunda. Lower fluxes of G. flabellatus occurred in October 2012, March, and October–November 2013, whereas its relative abundances generally followed the same seasonal pattern of the fluxes (Fig. 6b). Seasonality in Umbellosphaera spp. (2–8 %) and R. sessilis (0.2–2 %) was somewhat similar to that of G. flabellatus. Maximum fluxes of Umbellosphaera spp. occurred in December 2012 and September 2013, and lower fluxes occurred from late winter to early spring, with higher percentages in October and December 2012 and later in September 2013 (Fig. 6d). Flux maxima of R. sessilis occurred in late December 2012 and late July and early October 2013, and flux minima occurred in October 2012, late March, late April, and late May 2012 (Fig. 6h).

Rhabdosphaera spp. (0–3 %) and Umbilicosphaera spp. (1–3 %) revealed similar seasonal patterns, with the highest fluxes recorded during spring and summer and maxima in late July 2013 and later in April and June 2013 (Fig. 6e and g). In the case of Umbilicosphaera spp., flux increases were also noticed in early December 2012 and November 2013. The lowest fluxes of both taxa were recorded from January to March 2013. The seasonal pattern of their relative abundance generally followed that of the coccolith export fluxes. In spite of some similarity with the latter taxa, Helicosphaera spp. (0.1–2 %) was clearly more abundant in late April and late July, with the lowest fluxes from late summer to late winter (Fig. 6f), whereas E. huxleyi (0.1–2 %) showed two distinct and very sharp peaks in late April 2013 and in October–November 2013 (Fig. 6c).

Gephyrocapsa muellerae (0–25 %) occurred in persistently low fluxes and percentages throughout the year but remarkably increased in January 2013 (Fig. 6i). Seasonality in G. oceanica (0.1–9 %) was somewhat in between G. muellerae and E. huxleyi, but it was slightly more abundant from December 2012 until April 2013 and strikingly increased in October–November 2013 (Fig. 6j).

In comparison to the western site where different species revealed distinct seasonal variations, at the more central site M2 most of the taxa revealed a very similar seasonality, with the highest fluxes in late October 2012 and early July 2013 and lower fluxes in early November 2012, late May, and August 2013 (Fig. 6) The exceptions were G. flabellatus (Fig. 6b), which was the only one of the most abundant species without a flux increase in early March 2013, and Helicosphaera spp. (Fig. 6f), which were clearly more abundant from late winter to early summer 2013.

F. profunda (53–72 %) stood out for being much more common at site M2 compared to M4 (Fig. 6a), and species within Umbellosphaera spp. (4–10 %), H. carteri (1–9 %), Rhabdosphaera spp. (1–5 %), Umbilicosphaera spp. (1–5 %), and R. sessilis (1–4 %) were also generally more common in this area than at M4 (Fig. 6d–h). By contrast, G. flabellatus (4–16 %), E. huxleyi (3–13 %), G. oceanica (0–2 %), and G. muellerae (0–1 %) revealed much lower fluxes and percentages as well as comparatively low month-to-month variation in fluxes compared to M4 (Fig. 6b, c, i, and j).

Fluxes recorded at station M4 were high not only in comparison to station M2, but also to several other locations in the Atlantic Ocean, including open-ocean temperate and subtropical settings (e.g. Knappertsbusch and Brummer, 1995; Broerse et al., 2000; Sprengel et al., 2002), areas in the vicinity of islands (Sprengel et al., 2002), and more marginal regions, even when under the influence of coastal upwelling (Beaufort and Heussner, 1999; Köbrich and Baumann, 2009). Furthermore, most of these flux studies referred to the opportunistic E. huxleyi as being the dominant species, while the unexpectedly high total coccolith fluxes we found in the presumably oligotrophic western equatorial North Atlantic were mostly due to the LPZ dwellers F. profunda and G. flabellatus. Mean fluxes of G. flabellatus, in particular, were almost 14 times higher at M4 than at M2 (Fig. 6b, Table 2) and contributed importantly to Factor 1, explaining the highest percentage of the variability within the taxa and environmental parameters (i.e. 30 %) compared to the other factors (16, 10, and 9 % for F2, F3, and F4, respectively; Fig. 7). That F1 is clearly characterized by the opposition between the central station M2 and the western station M4 highlights the much higher abundances of G. flabellatus further west as the most important feature or difference between the two sites and indicates the presence of spatial variability shown by the LPZ flora as the statistically more relevant factor explaining our flux records. This finding is consistent with what it is known about the well-constrained enhancement of this species in surface sediments from the western equatorial Atlantic and Brazilian continental margin (Kinkel et al., 2000; Boeckel et al., 2006). That Umbellosphaera spp., Rhabdosphaera spp., Umbilicosphaera spp., E. huxleyi, and gephyrocapsids also produced higher coccolith fluxes in this area points to enhanced productivity throughout the entire photic zone at M4.

The persistent and overwhelming dominance of the LPZ flora suggests profiting from some subsurface and year-round nutrient supply. Forced by the trade winds, the westward deepening of the equatorial mixed layer and associated nutricline (see Hastenrath and Merle, 1987; Longhurst, 1993; Philander, 2001) could have promoted higher production of the LPZ flora in the western equatorial Atlantic (Fig. 10). In situ CTD and nutrient measurements at both stations do indicate a consistently deeper nutricline at M4, particularly during spring when the deep chlorophyll maximum (DCM) was found deeper at station M4 compared to station M2 (data not shown; Roepert and Brummer in Stuut et al., 2016).

In addition to the westward tilting of the nutricline, changes in the depth range of the Antarctic Intermediate Water (AAIW) flowing in from the south-west (Reid, 1994) may have also contributed to the enhanced fluxes of the LPZ flora further west (see http://whp-atlas.ucsd.edu and Fig. 10). Originating from the surface region of the Antarctic circumpolar layer, this water mass is known to follow the South Atlantic subtropical gyre to enter the western equatorial Atlantic (Stramma et al., 2003; Stramma and England, 1999). It crosses the Equator and spreads along the Brazilian shelf (Talley, 1996), contributing high nutrients and low oxygen to the Gulf Stream and North Atlantic Current (Reid, 1994). Furthermore, Poulton et al. (2017) have recently reported that F. profunda and Gladiolithus spp. thrive below the depth at which light is thought to be sufficient to support photosynthesis in equatorial waters, probably by mechanisms of mixotrophy and/or phagotrophy. This means that these two species may be able to live even deeper in the water column than originally thought.

Higher fluxes at station M4 may also be related to the higher production of faecal pellets by zooplankton grazers acting as vehicles for the downward flux of coccolithophores in this area. Recent observations by Knebel (2016) report higher fluxes of spinose planktonic foraminifera at station M4 compared to M2. Knappertsbusch and Brummer (1995) argued earlier that the export of coccolithophores is intimately related to day-to-day fluctuations in faecal pellet production by migrating zooplankton and nekton in the overlying mesopelagic zone. Therefore, higher zooplankton grazing in the western equatorial North Atlantic may have increased the coccoliths' transfer efficiency, hence contributing to the much higher coccolith fluxes compared to the central equatorial North Atlantic (Fig. 10).

Spatial environmental variability in the equatorial North Atlantic is most clearly expressed by the much higher fluxes and the pulsed maxima of the more opportunistic species E. huxleyi, G. muellerae, and G. oceanica at station M4 compared to their persistently low abundances and weak seasonality at station M2 (Figs. 6c, i, j, and 5). Their occurrence in the form of short-term high-flux events in January, April, and October–November 2013 at station M4 was associated with strikingly increasing ratios between these species and the LPZ flora (Fig. 9). Stoll et al. (2007) have used similar ratios as indicators of upwelling in the Bay of Bengal based on the idea that the decrease in the LPZ flora reflects the shallowing of the nutricline (Molfino and McIntyre, 1990; Beaufort et al., 1997). The sporadic sharp increase in the UPZ/LPZ ratio at station M4 hence suggests that the nutricline was temporarily shallower, resulting in a fast production response of the more opportunistic placolith-bearing species to increased nutrient input. Higher fluxes of these UPZ taxa at station M4 support the argument that the equatorial North Atlantic becomes generally more productive further west.

The pronounced maximum of G. muellerae in January 2013 (samples M4-6 and M4-7; Figs. 6i and 9a) was strong enough to be reflected in Factor 3, explaining 10 % of the variance (Fig. 7). This event occurred when the wind had just started to intensify and PAR was gradually increasing, as expected from the fall–winter transition in the equatorial North Atlantic (Fig. 2). This probably resulted in some degree of vertical mixing, as suggested by the decrease in SST, which in turn may have led to the supply of nutrients from below the pycnocline to shallower levels in the photic zone (see Sect. 5.1). The lack of a significant response by E. huxleyi and G. oceanica during this period suggests that nutrient availability was lower and/or not persistent enough for these species to bloom. Previous studies reported G. muellerae to have an affinity for intermediate to higher nutrient conditions in more transitional water conditions where competition with E. huxleyi and G. oceanica is lower (e.g. Giraudeau and Bailey, 1995; Boeckel et al., 2006; Guerreiro et al., 2013) and to occupy a deeper position in the photic zone than other species of the genus Gephyrocapsa (Boeckel and Baumann, 2008; Guerreiro et al., 2013). Low surface Chl a concentrations (Fig. 9a) and the absence of any significant increase in biogenic mass fluxes during this January period (Korte et al., 2017; Fig. 9b) point to a scenario in which only a few species profit from short-term nutrient supply and moderate light intensities at intermediate levels in the photic zone.

Similar conditions appear to have recurred in April 2013 (sample M4–12), with a pulsed maximum of E. huxleyi occurring under maximum PAR and lower SST following a period of more persistent wind strength increase that lasted from early winter until late spring (Figs. 2, 6c, and 9). This suggests that the water masses were stratified enough but still nutrient enriched due to previous wind-forced vertical mixing, hence promoting favourable conditions for blooming species to rapidly develop in the UPZ. The observed response of E. huxleyi is consistent with previous observations from other open-ocean areas reporting this species to dominate spring coccolithophore blooms (Haidar and Thierstein, 2001) and to induce enhanced coccolith fluxes (Broerse et al., 2000; Sprengel et al., 2002) following the water mixing and concomitant nutrient replenishment of the photic zone typical of late winter and early spring. That E. huxleyi responded in April but not in January suggests that it was more efficiently growing under presumably higher nutrient availability and maximum light levels during spring. This is consistent with previous observations of E. huxleyi bursting into a bloom within only a few days in response to nutrient availability and clear sky conditions in surface waters (Guerreiro et al., 2013). The high tolerance of this species to high levels of light has been considered crucial for its capacity to dominate coccolithophore assemblages (Nanninga and Tyrrell, 1996; Tyrell and Merico, 2004). Although to a lesser extent compared to E. huxleyi, other taxa that also increased during this April period include Helicosphaera spp., Rhabdosphaera spp., and Umbilicosphaera spp. This event was strong enough to be reflected in Factor F4, explaining 8 % of the variance (Fig. 7) in enhanced total coccolith fluxes and species H′ (Fig. 3), and by a striking flux increase in CaCO3, organic matter, and biogenic silica (Korte et al., 2017; Fig. 9b). This suggests that nutrient and light conditions were sufficiently favourable and persistent enough to promote the development of the entire UPZ plankton community during this short period.

The pulsed maxima of G. oceanica and E. huxleyi in October–November 2013 (sample M4-24) were also strong enough to be revealed by Factor 4 (Fig. 7) and by a slight increase in the total coccolith export production (Figs. 3 and 6c, j). Higher fluxes of these species were accompanied by increased surface Chl a concentrations (Fig. 9a), pointing to high productivity in the uppermost photic zone (Figs. 6c, j and 9a). During the SEM counting of coccoliths throughout sample M4-24, extremely high amounts of large diatom fragments were found, strongly supporting a scenario in which nutrient enrichment at the surface and clear sky conditions promoted the development of more competitive phytoplankton species. This is further supported by a striking flux increase in organic matter, biogenic silica, and to a lesser extent, CaCO3 during the same period (Korte et al., 2017; Fig. 9b). The ability of E. huxleyi and G. oceanica to compete with diatoms in surface waters and to rapidly respond to nutrient availability was observed in coastal waters off central Portugal, confirming its capacity for rapid population growth in nutrient-rich environments (Guerreiro et al., 2013).

During the fall 2013 event, SSS had dropped to a minimum of 33.9 at station M4 compared to ∼36.25 recorded at station M2 during the same period (Figs. 2a, b and 9a, b). Since both areas are under the influence of the ITCZ with a comparable precipitation regime (see Fig. 8) and given that higher precipitation during the fall of 2012 at M2 was associated with much higher SSS than at M4 for the same period (Fig. 2c and d), the distinctly lower salinity at M4 in October–November 2013 likely reflected advected Amazon River water (Figs. 9 and 10). A strong positive correlation between high surface Chl a concentrations and the salinity minimum revealed by Factor 2 (Fig. 7) along with the sharp increase in surface-dwelling opportunistic species, biogenic silica, and organic matter during this period (Fig. 9; Korte et al., 2017) suggest that this nutrient-rich buoyant plume promoted the development of phytoplankton at the surface of the western equatorial North Atlantic. The presence of the plume at station M4 was also noticed in a CTD profile taken during the recovery of the sediment trap mooring in late November 2013, showing a relatively shallow chlorophyll maximum associated with Amazon-related phytoplankton productivity (unpublished data, not shown). Previous satellite observations from this area have shown the north-eastward dispersion of the Amazon River Plume forced by the NBC retroflection at around 5–10∘ N, 50∘ W, typically from August to December (e.g. Muller-Karger et al., 1988; Molleri et al., 2010; see Sect. 2). The plume resulted in a dramatic surface salinity contrast (Wilson et al., 1994) and promoted a gradient in environmental conditions that evolved while the plume was meandering northward and mixing with the open-ocean waters, with a strong impact on the magnitude and composition of phytoplankton communities (refs. in Goes et al., 2014). The influence of the Amazon River water on the living coccolithophore communities has been previously reported by Winter et al. (2002) in the Caribbean Sea where enhanced cell densities of E. huxleyi and G. oceanica were seen coupled with this buoyant water mass. Our trap record clearly testifies to its impact by changing the flux and species composition of the coccolithophore communities in the oligotrophic western equatorial North Atlantic. Furthermore, despite the limitations of using environmental data that only represent the atmospheric conditions and the surface of the ocean, it highlights the remarkably good statistical correlation between surface satellite data (i.e. drastic salinity decrease and associated increase in Chl a) and enhanced coccolith fluxes of G. oceanica, E. huxleyi, organic matter, and biogenic silica (Korte et al., 2017) in the trap at 1200 m of depth.

The present-day deposition of Saharan dust has been recently quantified based on a transatlantic array of four sediment trap moorings between NW Africa and the Caribbean, which included mooring traps M4 and M2 (Korte et al., 2017; van der Does et al., 2016). The increase in AOD from spring to mid-summer at both stations (Fig. 2e and f) and its positive correlation with wind speed during this time (F2 positive scores; Fig. 7) corresponds to the same period when Korte et al. (2017) found the best accordance between dust outbreaks detected from satellite and high fluxes of dust-driven lithogenic particles. This agrees with previous observations that transatlantic Saharan dust fluxes are the highest during summer (e.g. Prospero et al., 2014). Since precipitation started to increase at the beginning of July at both stations (Fig. 2c and d), wet dust deposition probably contributed to the observed enhanced coccolith fluxes (Figs. 3 and 4). Through the exposure of the dust particles to cloud processes and mixing with anthropogenic species such as HNO3 in the atmosphere, wet dust deposition is thought to provide more bioavailable (soluble) nutrients compared to dry dust deposition and hence to have a greater fertilizing effect for primary production (Ridame et al., 2014). With the exception of R. sessilis, all the tropical taxa, including both UPZ and LPZ flora, showed increased fluxes during the July period at both locations, suggesting that coccolithophores benefited from the nutrient input by dust along the entire photic zone. A positive correlation between AOD and Helicosphaera species thought to thrive in waters of moderately high fertility (Roth and Bergen, 1975; Haidar and Thierstein, 2001; Boeckel et al., 2006) points to an ecological response to increased nutrient levels in July (F2; Fig. 7). Whereas E. huxleyi and G. oceanica also slightly increased during this period, one would expect these species to have a more significant response compared to the tropical assemblage. This was, however, not the case, suggesting that the dust acted more as a ballast than as a fertilizer during summer.

Two prominent dust flux peaks not detected by satellite were recorded in April and October–November of 2013 at station M4 (Korte et al., 2017), precisely when E. huxleyi and G. oceanica revealed a pulse-like increase in abundance (Figs. 6c, j and 9). This points to a scenario in which Saharan dust also acted as a nutrient fertilizer in this area in addition to surface–ocean mixing (spring) and the discharge and advection of the Amazon River Plume (fall). In fact, our observations suggest that the Amazon acted not only as a nutrient supplier during fall, but also as a buoyant surface density retainer of dust nutrients in the surface layer. Such a combination would explain the much higher fluxes of biogenic silica reported for this period by Korte et al. (2017) and the drastic increase in Chl a at the surface of the ocean (Figs. 2a and 9a). A similar case was observed offshore of central Portugal where the stabilization of a river buoyant plume was seen providing optimum stratification and nutrient-rich conditions for phytoplankton to bloom at the surface, resulting in a striking increase in Chl a concentrations, E. huxleyi, G. oceanica, and long-chain diatoms within only a few days (Guerreiro et al., 2013). Given the huge amounts of Saharan dust continuously blown into and over the Atlantic Ocean (e.g. Prospero et al., 2014), one could argue that the massive algal blooms reported to occur in the western equatorial North Atlantic, until now interpreted as being solely associated with nutrients provided by the Amazon, could actually result from the combination of surface buoyancy and nutrients provided by atmospheric dust deposition. In addition, the fall event was also marked by high precipitation rates, possibly resulting in higher nutrient bioavailability by means of wet dust deposition. In contrast, the Amazon Plume was not yet present in the study area during spring and the nearly non-existent precipitation possibly resulted in comparably lower nutrient bioavailability by dry dust deposition (see Ridame et al., 2014). Such differences in atmospheric–sea surface conditions between the two events apparently had an effect on phytoplankton species composition, resulting in higher development of several opportunist phytoplankton species during the fall and only E. huxleyi during spring.

The observed short-term shift from a more typically tropical (K-selected) to a more opportunistic (r-selected) settling coccolith assemblage during April and October–November 2013 at station M4 supports the dust fertilization hypothesis earlier proposed by Martin (1990) and later corroborated by observations from the Amazon Basin (Bristow et al., 2010), the Gulf of Mexico, and the coast of southern Florida (Lenes et al., 2012). High fluxes of organic material recently observed in a sediment trap in the North Atlantic subtropical gyre (23∘ N, 41∘ W) have been associated with enhanced phytoplankton productivity resulting from stimulated nitrogen fixation by Trichodesmium species following the deposition of dust-derived nutrients (Pabortsava et al., 2017). The role of aeolian dust for nitrogen fixation in subtropical and tropical oligotrophic regions has been previously reported by several authors (e.g. Falkowski et al., 1998; Mills et al., 2004; Jickels et al., 2005 and refs. therein; Martino et al., 2014). In the tropical North Atlantic where phytoplankton communities are nitrogen limited, experiments conducted by Mills et al. (2004) have shown Saharan dust addition to stimulate nitrogen fixation in this area, presumably by supplying iron and phosphorous. This is supported by Baker et al. (2006), who reported soluble phosphorous concentrations to be the highest in Saharan aerosols from the North Atlantic. Guieu et al. (2014), on the other hand, report the importance of strong and short-term (pulse-like) dust deposition events for marine productivity in low-nutrient, low-chlorophyll (LNLC) areas, whereas Romero et al. (2011) add that when dust input is accompanied by turbulence (i.e. strong winds), its potential fertilizing effects on phytoplankton are prone to be amplified. This fits well with the episodic nature of the dust deposition events at station M4 and the fact that they were accompanied by some degree of wind intensification.

In spite of all this evidence, mineral dust deposition is also thought to increase carbon sequestration to the deep ocean by acting as a mineral ballast of sinking particles (Pabortsava et al., 2017). Van der Jagt et al. (2017) report more abundant and faster-sinking aggregates when formed from a natural plankton community that has been exposed to Saharan dust deposition compared to less abundant and slower-sinking aggregates when formed without dust. The same authors argue that such dust-influenced aggregates would become heavily ballasted with lithogenic material at the surface and hence without scavenging any additional particles during their settling. This could at least partially explain why the dust peaks in spring and fall were marked by maxima of surface-dwelling species but not maxima of the deep-dwelling species. To confirm whether the pulsed flux maxima of opportunistic species presented in our study truly reflected the response of living coccolithophores thriving in the overlying photic layer to such a combination of factors (ecological signal) or resulted from enhanced particle transfer efficiency (e.g. ballasting by dust; faecal pellet production by zooplankton grazers; Armstrong et al., 2002; Ziveri et al., 2007; Fischer and Karakas, 2009; Fischer et al., 2016), a comparison between settling coccolith assemblages and the living coccolithophore communities as well as with in situ atmospheric–oceanographic observations would be required.

Whereas the spring and fall r-selected pulse events were accompanied by a slight increase in the total coccolith export, such an increase was never as high as in the peaks of November 2012 and July 2013 (see Sects. 4.1 and 5.1). This suggests that r-selected transient productivity is comparably less important than tropical background productivity for the overall bulk coccolith export production in the equatorial North Atlantic. Our observations indicate that enhanced productivity and/or transfer efficiency in the LPZ in the context of highly stratified tropical regions may be at least as important from a long-term perspective as that of the fast-blooming taxa more often found at higher latitudes and within productive coastal neritic areas.

In contrast to station M4, persistently lower abundances of opportunistic coccolithophore species, in particular the gephyrocapsids, and the absence of major pulse-like increases in these species at station M2 point to comparably more stable and oligotrophic conditions in this area. A narrower range and lower values of the UPZ/LPZ ratios at station M2 suggest that the UPZ was more persistently stratified and nutrient depleted in the central equatorial North Atlantic (Fig. 9). An exception occurs in early March 2013 when an increase in the UPZ/LPZ ratio at M2 and a species composition similar to that recorded in April at M4 is observed. It coincides with a slight flux increase in dust and carbonate (Korte et al., 2017), suggesting that M2 was also subjected to some degree of wind-forced water mixing (Fig. 2c and d) combined with Saharan dust deposition (Fig. 9). There are no indications of influence from Amazon River water at M2, as revealed by persistently low Chl a concentrations at the surface and low fluxes of opportunistic taxa during the fall of 2013 (Fig. 9a). Our observations suggest that, compared to M4, M2 was less affected by continental influences whether from Amazon River water or Saharan dust. The location of station M2, which is more central and 2∘ further north than station M4 closer to the centre of the North Atlantic subtropical gyre, may have contributed to the more oligotrophic character of this station (Fig. 10).

That the evidence we found for ocean fertilization by Saharan dust is more significant at station M4 than at station M2 may be due to the presence of buoyant Amazon water retaining nutrients and phytoplankton near the surface at M4 and to higher wet dust deposition during fall 2013 at station M4 (e.g. Ridame et al., 2014; Fig. 2c and d). A longer distance to the dust source in Africa at M4 compared to M2 may have also contributed to increasing the bioavailability of dust-driven nutrients further west (see Stuut and Prins, 2014). This could also help explain the lack of clear evidence for dust fertilization in open-ocean regions west of Africa (see Fischer et al., 2016; Neuer et al., 2004) since aerosols sinking in these areas would not be fine and chemically processed enough to act as fertilizers.