Holocene interdependences of changes in sea surface temperature, productivity, and fluvial inputs in the Iberian continental shelf (Tagus mud patch)

1. Introduction

[2] Until recently the present interglacial, the Holocene, has been considered as a period of stable climate [McManus et al., 1999]. However, high-resolution studies of paleoproxies in marine sediments [Bianchi and McCave, 1999; Cacho et al., 2001; Frigola et al., 2007] and ice cores [O'Brien et al., 1995], as well as dust records [Jackson et al., 2005], have revealed the occurrence of suborbital millennial-scale climate variability during the last 11 cal ka B.P. of the Holocene [Bianchi and McCave, 1999; Bond et al., 1997, 2001; Cacho et al., 2001; Dokken and Jansen, 1999; Frigola et al., 2007; Lorenz et al., 2006; Sarnthein et al., 2003]. Since the early studies of the Holocene in the North Atlantic Ocean, the occurrence of a 1,500-year cycle has been attributed to solar activity [Bond et al., 2001], to ocean current intensity variation [Bianchi and McCave, 1999], or yet to atmospheric processes linked to the North Atlantic Oscillation [Giraudeau et al., 2000]. During this period, sea surface temperature (SST) reconstructed by the alkenone method shows long-term cooling in the northeast Atlantic (36°N to 75°N) and in the Mediterranean Sea, involving drops of 1.2–2.9°C between the Holocene Optimum and the present [Marchal et al., 2002]. This temperature decrease is consistent with other climate proxy records and with model simulations based on Milankovitch forcing [Lorenz et al., 2006; Marchal et al., 2002]. At 10.2 cal ka B.P., a SST fall of 1–1.5°C marks the Preboreal oscillation in the North Atlantic, which in turn corresponds to a southward advance of the polar front by about 500 km [Ruddiman and McIntyre, 1981]. The most extreme cold episode of this present interglacial was the 8.2 cal ka B.P. “event” detected in North Atlantic marine deep sea sediments by several climate proxy data. This ‘8.2 event’ is often attributed to a meltwater outflow from the Hudson Basin (Canada) deglaciation into the North Atlantic Ocean and the consequent slowdown of the North Atlantic Deep Water formation [Alley et al., 1997; Alley and Ágústsdóttir, 2005; Barber et al., 1999; Clark et al., 2002; Mayewski et al., 2004]. The introduction of large amounts of freshwater into the North Atlantic triggers an important decrease of SST that occurs earlier and lasts longer than the isotopic signal recorded in the Greenland ice cores [Alley et al., 1997; Clark et al., 2001; Kleiven et al., 2008], as shown by long climate cooling anomalies of multicentennial scale between ∼8.9 and 8 cal ka B.P. [Ellison et al., 2006; Naughton et al., 2007a; Rohling and Pälike, 2005].

[3] Climate changes identified during the late Holocene have been less severe like the Roman Period (RP, A.D. 0–400 [Lamb, 1985]), the Medieval Warm Period (MWP, A.D. 800–1300 [Hughes and Diaz, 1994]) and the subsequent Little Ice Age (LIA, A.D. 1350–1900 [Bradley and Jones, 1993]).

[4] All the above referred studies and many other available in the literature provide an oceanic perspective of such climate changes since they were performed in cores retrieved from open sea sites. The continental response, although also reconstructed from marine sediment cores [Naughton et al., 2007b; Roucoux et al., 2006; Sánchez-Goñi et al., 2002], has been mainly studied in other paleoarchives such as ice cores [O'Brien et al., 1995], lake sediments [Cranwell, 1973], speleothems [Baldini et al., 2008] and tree rings [Briffa et al., 2002]. Marine and continental approaches give complementary information, but the dependence of the respective climate reconstructions from different paleoarchives restricts the possibilities of an integrated understanding of the overall changes. The use of the same archive in interface areas between the ocean and the continent has a strong potential for providing a better combined picture of the climate processes than those emerging from separate studies. In this perspective, climate reconstructions have also been done from continental shelf sediments. Besides variations in SST or salinity (SSS), these environments also record major variations in discharge of continental materials that affect the areas under fluvial influence [Amorosi et al., 2005; Müller and Stein, 2000; Rivera et al., 2006; Stein et al., 2004; Tanabe et al., 2006; Xiao et al., 2006] and experience significant coastline variations [Lambeck and Chappell, 2001; Siddall et al., 2003, 2004, 2006; Yokoyama et al., 2000].

[5] Studies of marine cores close to continental landmasses have been performed in sites such as the Arabian Sea (ODP Site 723A [Gupta et al., 2003]), western tropical Atlantic (GeoB3129-3911 [Weldeab et al., 2006]), eastern tropical Atlantic (GeoB 6518 [Schefuß et al., 2005; Weijers et al., 2007]), Gulf of Guinea (MD03-2707 [Weldeab et al., 2007]), eastern equatorial Pacific Ocean (EEP [Leduc et al., 2007]), Timor Sea (MD01-2378 [Xu et al., 2006]) and Cap Blanc (ODP Site 658C [deMenocal et al., 2000b]). Those sites are located at water depths of 807, 830, 962, 1295, 1619, 1738 and 2263 m, respectively, which is obviously deeper than the continental shelf. On the other hand, there is a large number of studies concerning the variability of sediment supply in different coastal areas, like the Laptev Sea [Müller and Stein, 2000; Rivera et al., 2006], the Amazon Basin [Maslin and Burns, 2000], the Yangtze River [Liu et al., 2004; Xiao et al., 2006], the Kara Sea [Stein et al., 2004], the Mahakam Delta [Storms et al., 2005], the Po River [Amorosi et al., 2005] or the Red River delta system [Tanabe et al., 2006], but they lack the information from the climate proxies performed in open sea sites.

[6] Until now, few previous studies have considered both climate change and river contribution to the continental shelf during the late Holocene (e.g., the Tagus mud patch, water depths 88–96 m [Abrantes et al., 2005; Lebreiro et al., 2006]; the Muros Ria, water depths 33–38 m [Diz et al., 2002; Lebreiro et al., 2006]).

[7] The Tagus River crosses the Iberian Peninsula and discharges a major load of suspended organic-rich sediments into the continental shelf off Lisbon forming the Tagus mud patch [Cabeçadas and Brogueira, 1997; Gaspar and Monteiro, 1977; Vale, 1981; J. H. Monteiro and I. M. C. Moita, personal communication, 1971]. Variations in river contribution and seasonal changes in coastal hydrology are described for the Cape Roca upwelling filament area for the last 2 ka [Abrantes and Moita, 1999; Abrantes et al., 2005; Fiúza, 1983; Moita, 2001]. Going further, the present work extends the initial evaluation of Abrantes et al. [2005] back to 11.5 ka. Biomarkers have been analyzed at high-resolution using the U₃₇^K′ index to determine SST. Primary productivity is estimated from the alkenone concentration and coccolithophore's assemblages and TOC, while river input is reconstructed from the concentration of Fe and higher plant n-alkanes and n-alkan-1-ols. The combined study of the these proxies contributes to understand the Holocene climate evolution in eastern Atlantic coastal areas providing unambiguously time-related information on changes occurring in the marine and continental environment and their interaction.

2. Material and Methods

[8] The study area is located in the western Portuguese continental shelf, between Cape Raso and Cape Espichel (Figure 1), in a region strongly influenced by the Tagus River. The Tagus is the longest river in the Iberian Peninsula. Periodic flood events lead to major discharge of suspended and bed load sediments [Vale, 1981]. The sediments transported in suspension are preferentially deposited at the inner and middle shelf, and are characterized by organic-rich silty clays (>80%) [Gaspar and Monteiro, 1977; Monteiro and Moita, personal communication, 1971].

[9] Two sedimentary sequences are studied in this paper: a composite of box core PO287-26B, gravity core PO287-26G and piston cores D13902; and the piston core D13882 (Table 1 and Figure 1). The box core and the gravity core were retrieved in May 2002 during the PALEO 1 campaign on board the RV POSEIDON [Monteiro et al., 2002]. The piston cores were collected during the Discovery cruise 249 in August–September 2000 using the National Oceanography Centre long-piston coring system (Table 1). The first 189 cm of core D13902 were disturbed and half-filled. As age determinations confirmed this disturbance, the zero depth of the core was moved downward to the 189 cm liner level [Abrantes et al., 2005].

[10] Iron (Fe) concentrations in counts per 30 s (cts) were determined at 1 cm intervals in cores PO287-26 and D13882 and at 2cm intervals in core D13902 by X-ray fluorescence core scanning for nondestructive semiquantitative analysis of major elements [Jansen et al., 1998] at Bremen University.

[11] For biomarkers analyses and TOC measurements, sediments were sampled every 1 cm in the box core, every 2–5 cm in cores D13902 and PO 287-26G, and every 5 cm in piston core D13882.

[12] Haphtophyte-synthesized C₃₇ alkenones and higher-plant biomarkers, such as C₂₃-C₃₃n-alkanes and n-alkan-1-ols C₂₀-C₃₀, were analyzed following the methods described by Villanueva et al. [1997]. Briefly, sediment samples (2 g) were freeze-dried and extracted by sonication with dichloromethane and then hydrolyzed with 6% potassium hydroxide in methanol. After derivatization with bis(trimethylsilyl)trifluoroacetamide, the extracts were analyzed with a Varian gas chromatograph Model 3400 equipped with a septum programmable injector and a flame ionization detector. The concentrations of each compound were determined using nonadecan-1-ol, hexatriacontane and tetracontane as internal standards.

[13] Selected samples were analyzed by gas chromatography coupled to mass spectrometry (GC-MS) for compound verification and identification of possible coelutions. The GC-MS used was a Fisons MD800 (THERMO Instruments, Manchester, UK). The carrier gas was helium (flow 2.1 mL/min). Injection and transfer line temperatures were 300°C. The quadrupole mass spectrometer was operated in EI mode (70 eV), scanning between m/z 50–650 in 1 s. The ion source temperature was 200°C.

[14] SST determinations were based on the alkenone unsaturation index, U₃₇^K′, which was calibrated to the temperature equation of Müller et al. [1998]. Replicate injections and sample dilution tests allowed assessing measurement errors to less than 0.5°C [Grimalt et al., 2001].

[15] Total organic and inorganic carbon was determined on aliquots (2 mg) of dry and homogenized sediment subsamples of the same levels used for biomarker determination. Total carbon was analyzed in the ground sediment with a CHNS-932 Leco elemental analyzer. Aliquots of those same samples were burnt at 400°C in an oven and further analyzed for determination of the inorganic carbon weight %. Organic carbon (OC) weight % was calculated by difference between the two measurements. The relative precision of repeated measurements of both samples and standards was 0.03 wt%.

[16] Smear slides for coccolithophore identification were prepared as described by Flores and Sierro [1997]. A light polarizing microscope (1250x) was used for observation and counting.

3. Chronology

[17] The age models are based on 18 accelerator mass spectrometry (AMS) ¹⁴C dates of Turritela and other mollusk shells and 16 ²¹⁰Pb measurements in the top part of the sequence. These ¹⁴C dated samples were chosen from core D13882 (six), core D13902 (eight) and core PO287-26G (three) and one in the bottom of box core PO287-26B (Table 2). All samples were collected at levels with no evidence of reworking, that is, the shells selected were the ones found in their living position, well integrated in the sediment level and in good degree of preservation. When available, complete bivalve mollusks still articulated were selected.

[18] The age model for the last 2 cal ka B.P. of cores D13902 and PO287-26B/G is reported by Abrantes et al. [2005]. The top part of core D13882 was correlated to D13902 via the Fe and MS records (Table 2 and Figure 2). Furthermore, six AMS-¹⁴C dates from core D13882 and one from D13902, using the MARINE04 calibration data and CALIB 5.0 program, were used for the rest of the sequence [Hughen et al., 2004; Stuiver and Reimer, 1993; Stuiver et al., 1998]. As such, all ages listed through the rest of the text correspond to calendar ka (ka = 1000 years) B.P. but will be referred as ka for simplification.

[19] The age model of box core PO287-26B, elaborated from the ²¹⁰Pb measurements and using the values determined in two levels collected from D13902 as background level, result in a sedimentation rate of 0.47 cm a⁻¹ and a temporal resolution of 2 years per centimeter. The same resolution was found in the upper 75 cm of piston core D13902 and the top 100 cm of gravity core PO287-26G.

[20] In cores D13902 and PO287-26G sedimentation rates were 0.11 and 0.14 cm a⁻¹, allowing resolutions of 9 and 7 years, respectively [Abrantes et al., 2005]. In core D13902, the section between 95 cm and 76 cm showed an instantaneous deposit of reworked material as consequence of the Lisbon earthquake and its associated tsunami [Abrantes et al., 2005]. Considering the last dated level below this event and the year of this earthquake (A.D. 1755) it was assumed that 355 years of record were eroded. The same event was found in core PO287-26G where approximately 160 years of the core were lost. For core D13882, sedimentation rates vary between 0.02 cm a⁻¹ and 0.2 cm a⁻¹ allowing a temporal resolution between 4 and 48 years, respectively.

4. Results

4.1. Sea Surface Temperature

[21] The U₃₇^K′-SST record of the composite Holocene sediment sequences (Figure 3a) fluctuates between 19°C and 14°C, showing maximum values between 11–6 ka and a minimum at the beginning of the 1990s. Furthermore, it exhibits a 4°C-decreasing trend between 10 ka and the present but is interrupted between 2 and 0.6 ka by the RP and the MWP. At the end of the MWP, there is an additional 2°C cooling, to an average temperature of 15.5°C, marking the beginning of the LIA. The last 130 years exhibit a strong variability with marked minima at A.D. 1930–1940 and A.D. 1990, 14.5°C and 14°C, respectively, and maxima of about 17°C at A.D. 1905, A.D. 1950–1970 and A.D. 1995. These changes involve SST drops and increases of 3°C in periods of 10 years or less (Figure 3a).

4.2. Paleoproductivity

[22] Total alkenones have been used as paleoproductivity indicators in open ocean sites [Schubert et al., 1998; Schulte et al., 1999; Villanueva et al., 1998; Villanueva et al., 2001]. In coastal areas, rivers input both particles and nutrients. While the particles will dilute the flux of sediments relative to the alkenone carrying organisms, the nutrients may enhance algal productivity. Off the Tagus, coccolithophores are dominant most of the year [Moita, 2001], governing the phytoplankton composition during winter and early spring when influence of nutrients from the river is high. Only during the upwelling season (May to September) are these algae outcompeted by diatoms [Abrantes and Moita, 1999]. On this basis alkenones may not be good markers for upwelling related productivity.

[23] In our sediments, the concentration of total alkenones is low (200 ng/g, or less than 20 ng/cm²/a¹) during the early Holocene, starting to rise at 5.5 ka and becoming more pronounced (up to 800 ng/g, or 300 ng/cm²/a¹; Figures 4b and 4c) after 3 ka. This same trend is observed for the total organic carbon (TOC) that ranges between 0.3 and 2% (Figure 4e). The strong similarity between these two proxies in D13882 supports a response to high amounts of river-transported nutrients to the coastal area, in particular during the last 3 ka.

[24] Examination of the distribution of coccolithophores reveals Gephyrocapsa (G.) as the most abundant genus, with dominance of the small sized Gephyrocapsa (<3 μm) followed by G. oceanica and G. muellerae (Figure 4f). While the group of small Gephyrocapsa and G. oceanica is known to live in nutrient rich environments [Brand, 1994; Wells and Okada, 1997; Winter et al., 1994], G. muellerae is characteristic of cold waters [Flores and Sierro, 1997].

4.3. River Contribution

[25] The influence of the Tagus River discharge in the studied area is easily observed in the satellite and space shuttle images [Abrantes et al., 2005]. C₂₃-C₃₃n-alkanes and C₂₀-C₃₀n-alkan-1-ols originate from higher plants [Eglinton and Hamilton, 1967] and their concentrations in the continental shelf can be used as an indicator for fluvial terrigenous input [Grimalt et al., 1990]. In addition, changes in the concentration of lipid compounds give information on the type of continental vegetation close to the study area. The alkanes index (C₂₉/C₃₁) has been defined on the basis of the existing relationship between the two main n-alkanes compounds following the suggestions made by previous studies [Cranwell, 1973; Poynter and Eglinton, 1990; Tareq et al., 2005]. This index allows us to know about the kind of vegetation that dominates in the river's hydrographic basin, in the adjacent continent. An index <1 reflects higher abundance of the alkane with 31 carbon atoms and indicates the dominance of grass and herbaceous plants. An index >1 results from a larger amount of alkanes with 29 carbon atoms, and so, the dominance of trees and shrubs in the adjacent continent.

[26] These Tagus sequences show a clear prevalence of C₂₉ during the early Holocene (Figure 5g). The concentrations of n-alkane-1-ols (Figure 5c) are currently higher (up to 11,000 ng/g) than those of n-alkanes (up to 4,000 ng/g, Figure 5e). However, both terrigenous proxies exhibit a similar trend, pointing to a common origin. The concentrations of these biomarkers are also parallel to the Fe content as measured by XRF (Figure 5f). This metal is a good marker for terrigenous input and therefore for river contribution. These three markers together validate clear trend toward higher terrigenous input in the last 3 ka, coincident with the onset of Neoglaciation [Andersson et al., 2003; Bauch and Weinelt, 1997].

5. Discussion

5.1. Climate Variability and River Evolution During the Holocene

[27] The continuous SST decreasing trend observed between 10.5–7 ka and the present (Figure 3a) follows the summer insolation at 37°N (Figure 3b). Similar trends have also been detected in U₃₇^K′-SST records from extratropical sites of the Atlantic Ocean (between −0.69 and +4.41°C [Marchal et al., 2002]), contrary to the warming progress observed in the Pacific Ocean (between −0.31 and +1.08°C [Kim et al., 2004]) and in tropical areas (between +0.19 and +1.05°C [Lorenz et al., 2006]). This divergent SST evolution in the Atlantic and the Pacific is attributed to a dipole pattern generated by the interaction between the positive Pacific North American Oscillation (PNA) and the negative North Atlantic Oscillation (NAO) phases by Kim et al. [2004], while Lorenz et al. [2006] attribute the divergent Holocene climate trend between northern and tropical locations to seasonally opposing insolation changes.

[28] The strong cooling of 4°C observed in the Tagus record is of similar intensity to that found at high latitudes, such as the North Atlantic [Marchal et al., 2002], but differs from the cooling trends of lower intensity observed in other similar latitudes sites [Cacho et al., 2001; Marchal et al., 2002]. This strong SST decrease may reflect higher sensitivity to seasonal changes by continental influence, e.g., an additional cooling effect due to the fact that in the cold seasons, river waters are cooler than seawater. An influence reflected by a clear contrast between river and sea surface temperature, recognized in satellite images (Figure 6) (P. Oliveira et al., personal communication, 2007). As such, the river output of colder waters resulted in a relatively stronger Holocene SST cooling gradient in this area. A gradient that is also noticeable in the terrigenous markers which show a continuous increase toward the Late Holocene (Figures 5b–5f).

[29] The period between 10 and 8 ka, reflects the highest sea surface temperatures of the Holocene in the western Iberian margin, and is also marked by the maximum extent of forested area, with the predominance of pine forests along the Portuguese coast [Queiroz, 1999], oak in high altitudes, like in the Serra da Estrela mountains [Van der Knaap and Van Leeuwen, 1985] and by oak and pine in the midaltitude and low-altitude sites of the northwestern Iberian Peninsula [Naughton et al., 2007b]. This suggests that both the Iberian margin and the adjacent landmasses have experienced the highest sea and atmospheric temperature values at the beginning of the Holocene. The C₂₉/C₃₁ ratio shows a remarkable coincidence with the abundance of semidesert plants [Naughton et al., 2007b; Turon et al., 2003]. In the Early Holocene, the C₂₉ is dominant and the alkane index is higher than 1 indicating forest cover with high soil stability and consequently, lower drainage. From 9 to 6 ka the concentration of terrigenous markers and Fe content is continuously low, accompanying the SST decrease, and in good agreement with the gradual contraction of the forest defended by Naughton et al. [2007b] and Turon et al. [2003].

[30] In terms of sea level, at ¹⁴C 10 ka the regional curve of Dias et al. [2000] indicates a sea level 60 m below its present level, resulting in a water depth of 40 m at the study area. On the lower Tagus valley, river sediments of that age correspond to channel deposits [Vis et al., 2008], and is only after 9 ka that the retention of sediments inside the estuary becomes important. A retention that caused a major decrease in the amount of terrigenous material transported to the deep sea through the Lisbon canyon (F. Abrantes et al., Sedimentation processes on the Tagus river system (Portuguese Margin) during the Holocene—SEDPORT main results, paper presented at European Geosciences Union General Assembly, Vienna, 1–7 April 2006), and possibly also to the shelf, although not reflected in the sedimentation rate (SR) (Figure 2).

[31] Between 6 and 3 ka, changes in the vegetation patterns on land are associated with deforestation. Soils became unstable and contribute more to river input. Concentrations of n-alkanes and n-alkan-1-ols start to rise progressively at 6 ka to a rapid increase around 3 ka, apparently coincident with the onset of the Atlantic Mutidecadal Oscillation (AMO) mode of variability [Lorenz et al., 2006]. High concentrations are maintained up to the present. Compared to the Holocene Climatic Optimum, during the two last warm periods, the Roman and the Medieval Warm Periods, the terrigenous biomarkers show prominent concentrations along with higher Fe contents. These enrichments are not coincident with SST drops. In fact, between 3 and 1.5 ka, pollen records from the southern Tagus River basin confirm an increase in agriculture activity and suggest deforestation and possibly higher soil erosion [Mateus, 1992; Queiroz, 1999]. Therefore, a higher concentration of terrigenous biomarkers during these warm periods seems a more direct response to an increase in soil use at times of amerioration than to river flow enhancement and climate cooling or high sea level stands.

6. Conclusions

[37] This high-resolution study of sedimentary sequences covering the last 11.5 ka provides the first integrated picture of the marine and continental climate changes at the inner to middle Portuguese continental shelf site under the influence of the Tagus River. The obtained proxy data reveals climatic events comparable with existing open sea and inland sites, and help to clarify the land-sea interaction during the Holocene.

[38] The general decrease trend in Holocene SST, from 19°C at 11–7.0 ka to 15°C at present, follows the general summer insolation trend and is in agreement with observations in other midlatitude marine sites. However, the Tagus mud patch sequences exhibits higher amplitude of variation, involving a 4°C decrease, likely to reflect a continental climate-dependent influence besides the marine variability at the site. Furthermore, the higher amounts of terrigenous markers at lower SST suggest a land effect by means of colder river outflow.

[39] The occurrence of rapid cooling periods of 1–2°C within 250 years at 11.1, 10.6, 8.2, 6.9 and 5.4 ka, within the general Holocene SST decreasing trend, match the ages of similar events recorded in the NE Atlantic–Mediterranean area showing that they correspond to a common widespread climate feature. Higher-amplitude SST variation observed during the last 1.5 ka, relative to the preceding Holocene times, exhibit the warm Roman Period and MWP followed by a 2°C decrease during the LIA.

[40] Decadal to subdecadal cold episodes are also recorded in the twentieth century, e.g., a 2°C decrease at A.D. 1925–1950 and a 3°C drop at A.D. 1980–1990. These proxy-defined episodes are coincident with low instrumental values measured during NAO negative phases and suggest that the known climate sensitivity of the Tagus marine area to this index is superbly recorded in the sediments deposited during the 150 years of instrumental record.

[41] River influence is observed to increase toward recent times, when comparing the early with the later Holocene, including the RP and MWP, and might reveal that AMO variability has operated since 5–3 ka. This change goes together with the climate cooling observed in the SST records but may also reflect deforestation by agricultural activities. Higher river influence results in an increase in the amount of nutrients brought from the continent to the sea, and consequently higher river-induced marine productivity.