Coccolithus pelagicus subsp. pelagicus versus Coccolithus pelagicus subsp. braarudii (Coccolithophore, Haptophyta): A proxy for surface subarctic Atlantic waters off Iberia during the last 200 kyr

A Multivariate Morphon Analysis (MMA) of Coccolithus pelagicus (s. l.) coccoliths was performed on 178 samples from an oceanic core recovered off the Portuguese margin (MD95-2040), in order to define the morphometric boundaries of its morphotypes and their behaviour during the last two glacial cycles. The recurrent occurrence of the smaller morphotype of this taxon, C. pelagicus subsp. pelagicus, was recognized for the first time off the Portuguese coast. Three sections, around Heinrich Layers 1 and 6 and Termination II-Eemian, were selected to establish high resolution comparisons between C. pelagicus subsp. pelagicus and C. pelagicus subsp. braarudii and two independent proxies for cold water masses: the records of the planktonic foraminifera Neogloboquadrina pachyderma (sinistral) and ice-rafted detritus (IRD). There is an overall negative correlation between C. pelagicus subsp. pelagicus and C. pelagicus subsp. braarudii as expressed by MMA Factor 1, the scores of which respond similarly to those of two non-cocolithophore proxies and consequently, is proposed as a calcareous nannoplankton proxy for the influence of subpolar waters in the region. Detailed analysis, however, showed occasional decoupling amongst these three proxies, which are used to highlight significant changes between the cooling–warming sequences of distinct events off Iberia. D 2005 Elsevier B.V. All rights reserved.

Coccolithus pelagicus (s. l.) has been one of the most commonly and easily interpreted calcareous nannofossil species, being used as a proxy for cold waters (McIntyre and Bé, 1967;Okada and McIntyre, 1979;Winter et al., 1994;Baumann, 1995;Andruleit, 1997). However, studies performed off Iberia showed that the biogeography of C. pelagicus (s. l.) also included upwelling regions (Cachão and Moita, 2000). This apparent contradiction started to be resolved when Geisen et al. (2002) presented life cycle evidences for the existence of two extant subspecies of C. pelagicus: C. pelagicus subsp. pelagicus, the subarctic form, and C. pelagicus subsp. braarudii, the temperate form, related to moderate fronts, namely those driven by upwelling conditions (Cachão and Moita, 2000). In 2003, Sáez et al. concluded that there were also genetic differences between these two entities (Sáez et al., 2003). Important for micropalaeontological studies, these subspecies can also be recognized by the size of their placoliths (Baumann, 1995), the former being smaller (7-10 Am) and the latter larger (10-16 Am) (Geisen et al., 2002;Parente et al., 2004). More recently, based on detailed morphometric analysis, Parente et al. (2004) confirmed the existence of three morphotypes of C. pelagicus (s. l.): the smaller morphotype related to C. pelagicus subsp. pelagicus, the intermediate size morphotype related to C. pelagicus subsp. braarudii, and a larger new one described as C. pelagicus subsp. azorinus (see Jordan et al., 2004). In this work, this taxonomic subdivision will be adopted.
Several authors showed already that the Polar Front position changed through time, playing an important role in the Iberian palaeoclimate (e.g. Fatela, 1995;Zazo et al., 1996). For instance, during the last glaciation, Ruddiman and McIntyre (1981) suggested that maximum southward advance of the Polar Front reached 428N, whereas the studies of Duprat (1983) established a limit of 388N for the same period. Later, Bard et al. (1987) concluded that during the earliest retreat of the last glaciation, the Polar Front moved from 358N to 558N. In contrast, Fatela et al. (1994) showed that during the last glaciation the Polar Front did not migrate south of 428N.
The present work will discuss the influence of the migration of the polar system closer to Iberia through the presence of C. pelagicus subsp. pelagicus, the subarctic species, during the two last glacial cycles, and to compare its (factorial scores) record with the records of the two well known independent proxies for cold water masses: planktonic foraminifera Neoglobo-quadrina pachyderma (sinistral variety) and ice-rafted detritus (IRD) during selected time intervals. The selected intervals correspond to the Heinrich Layers 1 and 6 and Termination II-Eemian when placoliths of C. pelagicus (s. l.) are particularly abundant and with distinct patterns.

Core location and oceanographic setting
This study was based on core MD95-2040, recovered in July 1995 east of the Oporto Seamount off the Portuguese margin (latitude 40834.91VN, longitude 09851.67VW) ( Fig. 1) (Bassinot et al., 1996).
The studied sector is a hydrodynamically active area, where water masses of different sources and properties converge and interact, following a dynamic vertically layered pattern, established as a function of temperature, salinity and dissolved oxygen content (de Abreu, 2000). This core provides a continuous pelagic record apparently unaffected by winnowing or significant chemical dissolution (de Abreu, 2000). In this eastern part of the North Atlantic subtropical gyre, the properties of modern surface water are influenced by the descending branch of the North Atlantic Drift (Portugal Current) and by a seasonal upwelling regime during summer months. This upwelling system is associated with the increase in the intensity and steadiness of northerly winds (Portuguese Trade Winds), as a consequence of the Azores anticyclone intensification and its westward movement during summer (Fiú za et al., 1998). In the northeastern border of the subtropical gyre, acting as a continuous link to the south-eastern branch of Gulf Stream, flows the Azores Current (Klein and Siedler, 1989). South of the Azores Islands the Azores Current coincides with the Azores Front, which is characterized by a zone of strong hydrographic transition, in terms of temperature (Gould, 1985) and water column structure (Fasham et al., 1985). Towards the East Atlantic margin, the geographic positions of these two entities are separated; the Azores Current continuing eastwards to the Strait of Gibraltar, while the Azores Front degenerates into two weaker transitions, one lying to the west of the Gulf of Cadiz and another lying further to the south flowing between the Canary Islands and North Africa (Johnson and Stevens, 2000). The region is also influenced by the North Atlantic Oscillation (NAO) in such a way that low NAO indices are related to higher atmospheric moisture in Western Iberia, coeval with warm temperatures over Greenland, while high NAO values are associated with Iberian aridity and cold temperatures over Greenland (Hurrel, 1995;Rodó et al., 1997).

Materials and methodology
MD95-2040 core, collected during the IMAGES MD101 cruise aboard the R/V Marion Dufresne, was recovered with a Calypso giant piston corer from a depth of 2465 metres below sea-level (mbsl), and is 35.24 m long (Bassinot et al., 1996).
A sample set of this core (178 samples) was used to analyse morphometric changes during marine isotope stages (MIS) 1 to 7 (de Abreu, 2000;de Abreu et al., 2003). Three sections with 58, 34 and 51 samples, corresponding to the Heinrich Layer 1 (HL1), Heinrich Layer 6 (HL6) and to the interval from Termination II to the Eemian (TII-E), respectively, were subsequently studied in detail due to the higher relative abundances of C. pelagicus subsp. pelagicus.
Smear slides were prepared and observed under an optical polarizing microscope (OLYMPUS BX-40), at a 1250Â magnification, connected to a digital camera (OLYMPUS DP11). For morphometry, 100 placoliths were randomly selected throughout the slide of each sample, and their maximum diameters (length) measured (Fig. 2).
The image analysis program Scion-Image, equipped with a suite of routines (www.nhm.ac.uk/hosted _ sites/ ina/CODENET/Coccobiom) and adapted to PC, was used to: (1) group the images in mosaics, so that multiple coccolith images could be seen simultaneously; (2) measure the main axis of each coccolith; and (3) copy the results to a spreadsheet for subsequent statistical analysis. The morphometric parameters were directly measured on frame-grabber captured digital images at magnifications of about 2500Â. The measurement-associated error is F 0.1 Am. For statistical analysis of the palaeoecological distributions specimens were assigned to 1 Am size integer intervals, hereafter referred to as morphons. For example, morphon 10 includes all specimens which have a maximum diameter within the interval [10.0, 11.0[ Am.
A morphometric data matrix, with samples as rows and morphons as columns, was subjected to R-mode Factor Analysis. This analysis, herein referred to as Multivariate Morphon Analysis (MMA), tests which contiguous morphons show high positive correlation and thus define a specific morphotype. A factor may be characterized by a single morphotype, with a particular and distinguishable behaviour along the entire data set, or by the opposite behaviour between two distinct morphotypes. Based on the percentage of the variance (of the initial data matrix) that each factor represents, a factor is considered relevant or not. This procedure allows: (1) to statistically test the existence of distinct morphotypes in the data set; (2) to define their morphometric boundaries and how they may change; and (3) to analyse the relative behaviour of the morphotypes along the time series. All statistical data processing used the program STATISTICAR version 7.
The age model of the MD95-2040 core, established from a combination of oxygen isotope stratigraphy, 14 C dating and synchronisation of the sea-surface temperature (SST) records and the GISP2 d 18 O data, is discussed in de Abreu et al. (2003). Isotopic stages are recognized and ages assigned in accordance with the standard curve SPECMAP of Martinson et al. (1987). The detailed stratigraphy for the last deglaciation and the Holocene is consistent with the planktonic d 18 O record from the nearby core SU81-18 (37846VN, 10811VW) described in Bard et al. (1989) (de Abreu, 2000de Abreu et al., 2003).
The palaeotemperature estimates for MD95-2040 are from de Abreu (2000) and were obtained with both the CLIMAP transfer function equation FA20 (Imbrie and Kipp, 1971) and with a modern analogue technique SIMMAX 28 (Pflaumann et al., 1996).

Overall MD95-2040 core
Factor analysis performed on the overall morphometric data (MMA) of MD95-2040 (Appendix 1; http:// correio.fc.ul.pt/~mcachao/papers/morphon/Appendix1. pdf) resulted in two factors, each related to a highly correlative combination of morphons, and which defines three morphotypes (Fig. 3, Table 1). Factor 1, with a variance of 43%, is strongly influenced by morphons 5 to 9 (5 to 8 with loadings higher than 0.7 and 9 with loadings of 0.63) in negative correlation to morphons 12 and 13 (with loadings higher than 0.7), reflecting an opposite behaviour between these two clusters of morphons, the two morphotypes associated to C. pelagicus subsp. pelagicus and C. pelagicus subsp. braarudii, respectively.
Factor 2, with a variance of 15%, is directly related to morphons 14 (with loadings of 0.7) and 15 (with loadings of 0.67), consequently describing the presence of a distinct and potentially meaningful larger morphotype with coccolith lengths longer than 14.0 Am.
The comparison of F1 scores against sea surface temperatures (warm and cold) shows a negative correlation between the smaller/intermediate morphotypes and the temperature of surface waters ( Fig. 4a and b).

Heinrich Layer 1
MMA performed on the HL1 data matrix (Appendix 2; http://correio.fc.ul.pt/~mcachao/papers/morphon/ Appendix2.pdf) shows there is just one factor for this time slice (Fig. 5, Table 2). This factor, with a variance of 63%, describes the behaviour of a small morphotype, characterized by morphons 4 to 9 (with loadings higher than 0.7), in perfect opposition to an intermediate one, characterized by morphons 11 to 14 (with loadings also higher than 0.7). Morphon 10 is characterized by almost null loadings thus defining the morphometric boundary between these two morphotypes (Fig. 5).

Heinrich Layer 6
MMA performed on the HL6 data matrix (Appendix 3; http://correio.fc.ul.pt/~mcachao/papers/morphon/ Appendix3.pdf) shows that the specimens' behaviour is somehow more complex, needing two factors to describe it (Fig. 6, Table 3). Factor 1, with a variance of 50%, is strongly influenced by the opposite behaviour between morphons 5 to 9 (small morphotype) and 11 to 13 (intermediate morphotype), all with absolute loadings higher than 0.7. Again it depicts strong opposite behaviour between these two morphotypes. Factor 2, on the one hand, with a variance of 15%, explains the independent behaviour of the morphons 14 and 15 (the first one with loadings of |À 0.61| and the last one with loadings of |À 0.74|), characterizing a large morphotype together with the occurrence of coccoliths, the lengths of which are in the limit between the small and the intermediate morphotypes (10 Am; with loadings of 0.79).

Termination II-Eemian
MMA performed on the TII-E data matrix (Appendix 4; http://correio.fc.ul.pt/~mcachao/papers/morphon/ Appendix4.pdf) depicts the appearance of the three distinct morphotypes whose behaviour can also be explained through two factors (Fig. 7, Table 4). Factor 1, with a variance of 49%, describes again the opposite behaviour between the small and the intermediate morphotypes by disclosing important negative correlation between morphons 5 to 9 and morphons 11 to 13 (all with absolute loadings higher than 0.7). Factor 2, with a variance of 18%, is related to morphons 14 and 15 (with absolute loadings higher than 0.7), thus reflecting Further study consisted of plotting and analysing the small/intermediate morphotype relationship (F1 factor scores) against depth, together with the two cold water proxies, the planktonic foraminifera N. pachyderma (sin.) and IRD, obtained from previous studies (de Abreu, 2000;de Abreu et al., 2003) (Fig. 8). Although Factor 1 represents the opposite behaviour between the smaller and the intermediate coccolith bearing morphotypes, its scores can be interpreted as an indication of the relative importance of C. pelagicus subsp. pelagicus in a region that otherwise registers the development of C. pelagicus subsp. braarudii.
In general, factor scores show increments during cold MIS 2, 4, and 6, being the most important peaks coincidal with the highest peaks of the other proxies, associated with the Heinrich Layers (e.g. HL1 and HL6). A gradual increase of these scores is also observed from Termination II to the beginning of isotope interglacial stage 5 (Eemian), although it does not completely match the records of the other two proxies for cold water masses.
Thus, there is a positive moderate correlation between Factor 1 scores and N. pachyderma (sin.) (r = 0.57; n = 84), while with IRD this correlation, although positive, is less important (r = 0.37; n = 84) for the entire time series.

Heinrich Layer 1
During HL1 (Fig. 8a) the relative abundances of N. pachyderma (sin.) and C. pelagicus subsp. pela-  gicus present a very similar pattern, which is also demonstrated by the high correlation between them (r = 0.93; Table 5a). When compared with these two proxies, the IRD record presents, in general, a similar trend, although with more pronounced oscillations. As expected, the correlation coefficients between the two first proxies and IRD are lower, although still representative (Table 5a). The comparison of the different records allows to distinguish three phases: -phase I is characterized by a precocious increase of N. pachyderma (sin.) and C. pelagicus subsp. pelagicus relative to the record of the IRD, synchronously with the onset of the cooling trend (SSTs; see Fig. 8); -phase II, corresponding to the interval between 17.5 and 15.6 kyr for which the estimated SSTs ( Fig. 8) register the lowest values, is characterized by all three proxies being persistently and abundantly represented. The oscillating pattern of IRD is responsible for the lower coefficient correlation values between this proxy and the other two (Table 5b); -phase III is characterized by all three proxies disclosing a similar decreasing behaviour until the end of this interval, at which interpreted SSTs (Fig. 8) rise again. The representativity of their correlation is expressed by coefficient values on Table 5b.

Heinrich Layer 6
During HL6 (Fig. 8b) the N. pachyderma (sin.) and C. pelagicus subsp. pelagicus relative predominance also depicts a similar but less close pattern in three  phases. In fact, the correlation among the three proxies is representative for the overall interval, but with negative values between C. pelagicus subsp. pelagicus and the other two proxies (Table 6a).
-phase I, as during HL1, N. pachyderma (sin.) and C. pelagicus subsp. pelagicus reacted to the cooling shortly before the arrival of IRD. Due to the reduced number of samples, correlations although high are not meaningful (Table 6a); -phase II, displaying the lowest estimated SSTs (Fig. 8, time interval 61.4 to 59 kyr), and in spite of the three proxies being well represented and with an overall similar trend, correlations are low and not meaningful (Table 6b). Thus, C. pelagicus subsp. pelagicus relative abundance exhibits a pattern with relatively strong oscillations, namely at the base and the top of these interval peaks, which moderately correspond to the major peaks of the IRD record. Similarly, a non-meaningful anti-variation between C. pelagicus subsp. pelagicus and N. pachyderma (sin.) might indicate a moderate offset between these two proxies; -phase III is characterized by a synchronous decrease on the abundance of the three proxies. Although showing an identical behaviour only the positive correlation between N. pachyderma (sin.) and IRD record is meaningful (Table 6b).

Termination II-Eemian
From Termination II to the beginning of isotope interglacial stage 5 (Eemian), the relative behaviour of C. pelagicus subsp. pelagicus does not accompany the records of the other two proxies (Fig. 8c). This decoupling is responsible for the registered null correlations   (Tables 7a, b and c). In this interval five different phases can be observed: -phase I, as during the previous intervals, is characterized by the simultaneous increase of N. pachyderma (sin.) and C. pelagicus subsp. pelagicus, both registering the cooling trend earlier than IRD; -phase II, from 134 to 125 kyr, both IRD and N. pachyderma (sin.) display a significant increase timidly followed by C. pelagicus subsp. pelagicus; -phase III is defined this time by lower and irregular IRD values and almost null N. pachyderma (sin.) percentages, while C. pelagicus subsp. pelagicus continuous to increase its record; -phase IV, from 117.6 to 114.8 kyr, N. pachyderma (sin.) and IRD are almost absent while C. pelagicus subsp. pelagicus is still increasing its importance; -phase V defines the interval during which C. pelagicus subsp. pelagicus relative abundance returns to minimal values.
This complete decoupling between C. pelagicus subsp. pelagicus and the other proxies are well depicted by their correlation values for phases II to V, though during phase I, as mentioned, C. pelagicus subsp. pelagicus and N. pachyderma (sin.) behaved similarly.

Importance of the Multivariate Morphon Analysis (MMA)
By tabulating morphometric measurements of C. pelagicus (s. l.) placoliths into 1.0 Am classes (morphons 4 to 16 Am), MMA disclosed clusters of mutually correlated contiguous morphons, each cluster defining a single morphotype.
Concerning the definition of morphotype boundaries, our data indicates that the morphometric interval characteristic of each morphotype narrows from the smaller to the larger morphotype (Fig. 9), although small changes in time may occur. For instance, during HL1 ( Fig. 5), the smaller morphotype is defined by a strong correlation between all morphons 4 to 9, indicating that all placoliths between 4 and 10.0 Am equally contribute to its definition. However, during HL6 (Fig. 6), morphon 4 (placoliths with [4.0; 5.0[ Am length) do not correlate with the others and so should not be used to define it. Causes for this can be related to its lower abundance because: i) it was not well formed during this older cold stage or ii) its presence in the region was too erratic. During TII-E (Fig. 7) morphon 4 regained some importance for the definition of the smaller morphotype. However, its factorial load being less than |À 0.4|) means that a significant part of the measured specimens with that size range have no correlation with the rest of the cluster, and thus should not be used for the morphotype tabulation. As shown, which of these specimens covary and thus can be used with the rest to define a specific morphotype can only be defined through MMA.
Results show that during these three time slices the upper morphometric limit between the smaller and the intermediate morphotypes is kept consistently around the size range of morphon 10 ([10.0; 11.0[ Am), as displayed by its almost null factor loadings (see Figs. 5-7). Thus, this dimensional interval reveals to be useful to separate placoliths attributed to C. pelagicus subsp. pelagicus and C. pelagicus subsp. braarudii, respectively, confirming previous interpretations from Geisen et al. (2002).
During HL1 a single factor explains the overall morphometric behaviour of C. pelagicus (s. l.), indicating that only the smaller and intermediate morphotypes are present, with a clear opposite behaviour between them (see Fig. 5). This allows a simpler comparison of the MMA method with the more traditional single histogram visual analysis. As seen in Fig. 10, frequency histograms clearly display the 10 Am morphometric boundary. Samples 293 and 218 cm have low F1 scores meaning that the two morphotypes have an equal importance as shown by their histograms. Sample 257 cm has a higher positive score. Since the higher positive loadings of Factor 1 characterize the smaller morphotype, high positive   Table 5a Correlation coefficient values (r) among ice-rafted detritus (IRD), N. pachyderma (sin.) and C. pelagicus subsp. pelagicus during the entire Heinrich Layer 1 and the first phase of this interval Bold correlations are significant at p b 0.05. Table 5b Correlation coefficient values (r) among ice-rafted detritus (IRD), N. pachyderma (sin.) and C. pelagicus subsp. pelagicus during the second and third phases of the Heinrich Layer 1 Bold correlations are significant at p b 0.05.
Intermediate stages in the steps described above were not presented for the sake of clarity. In fact, the analysis and graphic representation of a large number of histograms altogether is cumbersome and is one of the advantages of substitution by the integrated approach provided by the MMA method. However, because the MMA method is based on bthe behaviour of the sizesQ it can not be applied to isolated or to a limited number of samples. The minimum number of samples should necessary be equal to the number of morphons (i.e. tabulated morphometric classes) that are considered for a specific taxa.
The usefulness of MMA is even better evidenced when one attempts to define the morphometric boundary between the intermediate and the larger morphotypes. While during HL1 C. pelagicus subsp. braarudii was formed by all morphons 11 to 14 ([11.0; 15.0[ Am) (Parente et al., 2004), during HL6 (Fig. 6) the morphon 14 ([14.0;15.0[ Am) has both positive scores on Factor 1 and negative scores on Factor 2. This indicates that part of the placoliths that integrate this morphon behave coherently with the 11-14 Am sized placoliths, which constitute C. pelagicus subsp. braarudii, while another part shows a distinct and independent behaviour in phase with the other larger placoliths, which are attributed to the larger morphotype C. pelagicus subsp. azorinus. The same also occurs during TII-E, although to a lesser Table 7a Correlation coefficient values (r) among ice-rafted detritus (IRD), N. pachyderma (sin.) and C. pelagicus subsp. pelagicus during the entire interval TII-Eemian and its first phase Bold correlations are significant at p b 0.05. Table 7b Correlation coefficient values (r) among ice-rafted detritus (IRD), N. pachyderma (sin.) and C. pelagicus subsp. pelagicus during the second and third phases of the interval TII-Eemian Bold correlations are significant at p b 0.05. Table 6a Correlation coefficient values (r) among ice-rafted detritus (IRD), N. pachyderma (sin.) and C. pelagicus subsp. pelagicus during the entire Heinrich Layer 6 and the first phase of this interval Bold correlations are significant at p b .05. Table 6b Correlation coefficient values (r) among ice-rafted detritus (IRD), N. pachyderma (sin.) and C. pelagicus subsp. pelagicus during the second and third phases of the Heinrich Layer 6 Bold correlations are significant at p b .05.
degree. This indicates that the morphometric boundary between the intermediate and the larger morphotypes is less clearly defined due to a higher degree of overlapping between them (Fig. 9), which, in turn, may reflect a younger phylogenetic separation. In these circumstances, to distinguish these morphotypes based only on single histograms is virtually impossible. Fig. 11 shows the relative behaviour of the three morphotypes during HL6, expressed by their respective MMA factorial scores and selected frequency histograms. Sample 1187 cm has a positive F1 score and a negative F2 score. As shown before (see Fig. 6) this indicates a strong presence of both the intermediate and the larger morphotypes. In fact, the histogram shows a minor contribution of the smaller morphotype relative to the larger placoliths (11 to 15 Am), but do not show a clear distinction between the larger morphotypes. As stated before, this latter distinction can only be made by analysing the behaviour of the morphons along the selected time series, and not on a single sample. The following selected sample (1160 cm) has a negative F1 score and a positive F2 score. As shown before this indicates that the smaller morphotype is better represented than the intermediate morphotype (as clearly shown by its histogram), and the larger morphotype should be absent. This last aspect is less clear to observe in the histogram of the sample 1160 cm, mainly because Factor 2 is also influenced by the presence of placoliths of size [10,11[, which MMA analysis indicates do not contribute to the smaller neither to the intermediate morphotypes, but anti-vary with the largest sized placoliths, a still not completely understood relationship. The almost null F1 scores for sample 1148 cm is indicative of the equatitative presence of the smaller and the intermediate morphotypes, an aspect that is clearly seen in the histogram. As for sample 1187 cm its negative F2 scores also indicate the presence of the large morphotype. Positive F1 and F2 scores for sample 1136 cm are indicative of increased importance of the intermediate morphotype, relative to the others. The last selected sample (1106 cm) is a more extreme situation than those previously described as its histogram also clearly shows.

Palaeoecological and palaeoceanographical implications
5.2.1. Comparison C. pelagicus subsp. pelagicus versus C. pelagicus subsp. braarudii Several mechanisms have been proposed to explain the iceberg discharges into the North Atlantic, known as bHeinrich eventsQ (Broecker et al., 1992), such as insulation changes (Heinrich, 1988), sea level changes (Bond and Lotti, 1994), global temperature changes (Bond and Lotti, 1995), and ice-sheet dynamics (Broecker, 1994;McManus et al., 1999). During the subsequent trend to increase global tem- Fig. 9. Diagram illustrating the relative contribution and variability of C. pelagicus (s. l.) placolith lengths for the three morphotypes. The dimensional interval that characterizes them decrease from the smaller to the larger morphotype, while the dimensional boundary between the smaller and the intermediate morphotypes is much better defined than between the intermediate and the larger ones. Table 7c Correlation coefficient values (r) among ice-rafted detritus (IRD), N. pachyderma (sin.) and C. pelagicus subsp. pelagicus during the fourth and fifth phases of the interval TII-Eemian Bold correlations are significant at p b 0.05. perature, the southward migration and melting of icebergs at medium latitudes as indicated by several evidences (Cayre et al., 1999;Bard et al., 2000;de Abreu et al., 2003) would induce the cooling of the surrounding water masses. Entire coccospheres (or just isolated placoliths) of C. pelagicus subsp. pelagicus may thus have extended its geographic range, towards the Iberian margin reflecting this cold water influx.
The systematic increment of C. pelagicus subsp. pelagicus over C. pelagicus subsp. braarudii during Heinrich events shows that their opposite behaviour, expressed throughout the last 200 kyr of core MD95-2040 by MMA F1 scores (see Figs. 3, 5, 6 and 7), can be used as a proxy for cold waters, similarly to the record of N. pachyderma (sin.) and IRD. This is corroborated by negative correlations, although moderate, between F1 scores and estimated palaeotemperatures (r = À 0.51 and r = À 0.43 for cold and warm season SSTs, respectively).
The consistently opposite behaviour between C. pelagicus subsp. pelagicus and C. pelagicus subsp. braarudii off the Western Iberia margin is interpreted as follows. During glaciations the southward shift of the Polar Front closer to North Iberia moved the meridional limit of the ecological niche of C. pelagicus subsp. pelagicus southwards. This shift seemed to have conditioned at certain specific moments the development of C. pelagicus subsp. braarudii, which otherwise was experiencing a fair development due to on one hand an increase on upwelling conditions during glacial times and on the other hand a restriction by sea level retreat from the Western Iberia continental shelf, over which has its present day niche. Thus, while the presence of the smaller morphotype off Western Iberia is linked to a cooling trend, the intermediate morphotype (C. pelagicus subsp. braarudii) seems to be favoured by non-glacial upwelling conditions (see Cachão and Moita, 2000). The behaviour of the larger morpho-type (when present) is always characterized by a distinct factor, reflecting that its behaviour is unrelated to the others and thus its presence off Iberia must be linked to a distinct mechanism, most probably related to the activity of the Azores Front (Parente et al., 2004).

Comparison C. pelagicus subsp. pelagicus versus N. pachyderma (sin.) and IRD
Since N. pachyderma (sin.) and IRD (Fig. 8) are commonly used to better characterize periods influenced by cold meltwater, the significant overall similarity of F1 scores to the records of these two proxies, particularly evident during MIS 2 and 4, confirms the positive response of C. pelagicus subsp. pelagicus to the arrival of subpolar water masses off Western Iberia. The limited contribution of C. pelagicus subsp. pela- Fig. 11. Time series and histograms of Factor 1 and Factor 2 scores for HL6 for selected samples (1187, 1160, 1148, 1136 and 1106 cm). Positive F1 scores indicate a stronger presence of C. pelagicus subsp. braarudii (C.p.b.z-C. pelagicus subsp. braarudii increase). Negative F1 scores indicate a stronger presence of C. pelagicus subsp. pelagicus (C.p.p.z-C. pelagicus subsp. pelagicus increase). Samples with almost null F1 scores are characterized by the presence of these both subspecies. Negative F2 scores indicate a stronger presence of C. pelagicus subsp. azorinus (C.p.a.z-C. pelagicus subsp. azorinus increase).
Detailed analysis for the HL1, HL6 and TII-Eemian intervals occasionally show small discrepancies in the timings of these three proxies. This can be partially understood if one considers that the present day occurrence of sinistral N. pachyderma is mainly confined to the Polar region itself (e.g. Hemleben et al., 1989;Bauch et al., 2003) while C. pelagicus subsp. pelagicus is attributed to the bordering Subarctic biogeographic province (McIntyre and Bé, 1967;Raham fide Roth, 1994;Winter et al., 1994). Since the two taxa do not share the exact same water masses their records may reflect the position of their respective biogeographical boundaries.
For all three studied time intervals (HL1, HL6 and TII-E) the increase of both C. pelagicus subsp. pelagicus and N. pachyderma (sin.) shortly before the rise of IRD abundances (bottom of phase I; see Fig. 8a,b and c) is interpreted as reflecting an increase in the cooling trend and maximum meridional ice front development, which gradually shifted southwards their subpolar and polar meridional biogeographic boundaries. Their synchronicity indicates, on the other hand, that during this paroxysmal cold stage (phase I) the meridional limits of the Planktonic Foraminiferal Polar and the Calcareous Nannoplankton Subarctic Zones coincided, while migrating over our core location, off Western Iberia. This overlapping is also confirmed by high correlations between F1 scores and N. pachyderma (sin.) during all phase I (see Tables 5a, 6a and 7a). With the onset of terminations, ice cap break up and southward movement of the iceberg armadas, IRD begins to arrive at our core location, and defines the lower limit of all phase II. During this stage, both C. pelagicus subsp. pelagicus and N. pachyderma (sin.) increase significantly their relative abundance, reflecting an enlargement of their biogeographic domains in this region. During HL1 C. pelagicus subsp. pelagicus and N. pachyderma (sin.) records of phase II correlate with each other significantly (r = 0.71; Table 5b) but between C. pelagicus subsp. pelagicus and IRD it is not so good. This seems to indicate that, at this stage, both types of organisms shared the same palaeobiogeography but were not favoured during direct iceberg discharges. However, during all phase II of HL6 and TII-E, C. pelagicus subsp. pelagicus and N. pachyderma (sin.) display negative moderate correlation to each other. Although sharing the same palaeobiogeography, it is interpreted that they developed in distinct water masses, keeping their own identity, as reflected by this present day distinct Northern Atlantic biogeographic provinces.
Phase III of HL1 and HL6, characterized by the gradual decreasing trend of C. pelagicus subsp. pelagicus, N. pachyderma (sin.) and IRD, are interpreted as reflecting the disappearance of glacial-subglacial water conditions from offshore Iberia and the re-establishment of a North Atlantic current system similar to today. Again, during HL1 all three proxies behave quite similarly (correlation coefficients are significantly high; Table 5b), while during HL6 C. pelagicus subsp. pelagicus and N. pachyderma (sin.) display moderate anti-variation. This suggests that during HL1 the glaciation-deglaciation cycle was much more intense than during HL6, for which either species, although reflecting the same general retreat pattern maintained their individuality of inhabiting slightly distinct water masses. Phase III of TII-E is quite distinct from the previous ones. In it IRD experiences a less intense and more irregular second increase, N. pachyderma (sin.) practically disappears while C. pelagicus subsp. pelagicus increases its relative abundance. This complete decoupling between the two taxa indicates that during Termination II the Planktonic Foraminiferal Polar province retreated from off Iberia much faster than the Calcareous Nannoplankton Subarctic province, sooner even than the complete iceberg meltdown. On the other hand, the Subarctic province persisted much longer than for the other time intervals, as indicated by the increasing pattern of C. pelagicus subsp. pelagicus even when the other two proxies were already absent (phase IV). This is interpreted as reflecting a much slower return to normal conditions after TII around Western Iberia, compared to what happened during Termination I (HL1). This interpretation implies that obtained SST's for this interval along Western Iberia are overestimated due to the precocious retreat of N. pachyderma (sin.), one of their transfer function estimators.
The above interpretations fail, however, to justify the reason why other Heinrich Layers (HL 2 and 5) present a significant increase of N. pachyderma (sin.) and IRD but no meaningful record of C. pelagicus subsp. pelagicus. This question will continue to be followed through the undergoing study of new cores in the region.

Conclusion
From the results of the statistical morphometric analysis of C. pelagicus (s. l.), here defined as Multivariate Morphon Analysis (MMA), applied to core MD95-2040 located off Western Iberia, a small morphotype (with coccolith lengths less than 10.0 Am) known as C. pelagicus subsp. pelagicus could be recognized. This taxon depicted in all situations a clear anti-variation to C. pelagicus subsp. braarudii as expressed by the most important factor retrieved by MMA. F1 scores revealed an overall similar response to N. pachyderma (sin.) and IRD and thus have a potential to be used as a palaeoceanographic proxy for the influence of Subpolar North Atlantic water masses. However, when analysed in much greater detail, the coupling or decoupling between C. pelagicus subsp. pelagicus and the other two cold water proxies allows more precise palaeoceanographic local reconstructions. Thus, for each of the three higher resolution sections (Heinrich Layer 1, Heinrich Layer 6 and the interval from Termination II to the Eemian) we conclude that: 1. The significantly coupling (in phase and positively correlated) between C. pelagicus subsp. pelagicus and N. pachyderma (sin.) during HL1, together with the fact that the morphometric variability of C. pelagicus subsp. pelagicus is mainly explained by a single factor, indicates that the Last Glacial Maximum and Termination I strongly conditioned palaeoecological conditions off Western Iberia, during which there seems to be an overlapping of both Polar and Subarctic biogeographic provinces; 2. The moderate decoupling (in phase but slightly negatively correlated) of C. pelagicus subsp. pelagicus and N. pachyderma (sin.) during HL6, together with a more complex pattern of morphometric variability, indicates that during HL6 the cooling phase was intense but the warming occurred more gradually, allowing the two taxa to retreat to the global changing conditions but according to their own specific slightly distinct (palaeo)ecological requirements; 3. The complete decoupling of C. pelagicus subsp.
pelagicus and N. pachyderma (sin.) during and after TII-E indicates a much gradual return to normal interglacial conditions, in which each taxon kept their own identity. This allows a distinction to be made between the first retreat of the Polar province followed, 7.6 kyr later, by the Subarctic biogeographic domain retreat from off Western Iberia.