Proxy calibration to instrumental data set: Implications for paleoceanographic reconstructions

Authors


Abstract

[1] High-resolution proxy data analyzed on two high-sedimentation shallow water sedimentary sequences (PO287-26B and PO287-28B) recovered off Lisbon (Portugal) provide the means for comparison to long-term instrumental time series of marine and atmospheric parameters (sea surface temperature (SST), precipitation, total river flow, and upwelling intensity computed from sea level pressure) and the possibility to do the necessary calibration for the quantification of past climate conditions. XRF Fe is used as proxy for river flow, and the upwelling-related diatom genus Chaetoceros is our upwelling proxy. SST is estimated from the coccolithophore-synthesized alkenones and U37k′ index. Comparison of the Fe record to the instrumental data reveals its similarity to a mean average run of the instrumentally measured winter (JFMA) river flow on both sites. The upwelling diatom record concurs with the upwelling indices at both sites; however, high opal dissolution, below 20–25 cm, prevents its use for quantitative reconstructions. Alkenone-derived SST at site 28B does not show interannual variation; it has a mean value around 16°C and compares quite well with the instrumental winter/spring temperature. At site 26B the mean SST is the same, but a high degree of interannual variability (up to 4°C) appears to be determined by summer upwelling conditions. Stepwise regression analyses of the instrumental and proxy data sets provided regressions that explain from 65 to 94% of the variability contained in the original data, and reflect spring and summer river flow, as well as summer and winter upwelling indices, substantiating the relevance of seasons to the interpretation of the different proxy signals. The lack of analogs and the small data set available do not allow quantitative reconstructions at this time, but this might be a powerful tool for reconstructing past North Atlantic Oscillation conditions, should we be able to find continuous high-resolution records and overcome the analog problem.

1. Introduction

[2] Forecasts of future climatic trends depend on widespread accurate and quantitative records of past climate. Briffa and Osborn [2002], on a revision of a new tree ring–derived temperature record for the past millennium, state “We need more independent reconstructions, based on improved proxy records, and we need to know why it was once so warm and then so cool, before we can say whether 21st-century warming is likely to be nearer the top or the bottom of the latest IPCC range of 1.4°C to 5.8°C” [Watson et al., 2001]. Furthermore, in 2002, recommendation 3 of the U.S. National Research Council report on “Abrupt Climate Change” [Committee on Abrupt Climate Change, 2002] reinforces the need for (1) coordinated projects to produce especially robust, multiparameter, high-resolution histories of climate change and ecological response; (2) better geographic coverage and higher temporal resolution; and (3) additional proxies, including those that focus on water (e.g., droughts, floods, etc). Additional emphasis is put on the need for records of the last 2,000 years, so that warming and associated changes of the last 100 years can be assessed in context.

[3] Given that the instrumental record is short, only about 100 years at most, longer climate reconstructions depend on proxy data that needs calibration. Up to now, much of the sediment proxy data calibration has been of two types: (1) definition of empirical correlations based on the comparison of spatial distribution of a certain proxy and a certain property and (2) the establishment of direct relationships through sediment trap studies. Nevertheless these studies are recent and the available time series are yet too short to allow totally reliable calibrations. The existence of high-resolution proxy data from key areas characterized by specific phenomena and from areas where instrumental record data is available is of great importance to estimate the relative magnitude of past changes. In addition it will also allow the scaling of proxy data against 100 years of instrumental observations.

[4] In this article, we use paleoenvironmental high-resolution records recovered from the inner-shelf deposition center of the suspended matter transported by the Tagus river flow to assess the climate events of this century. To do so, we present what we believe to be a first attempt to calibrate sediment proxy data for both continental climate and oceanic productivity to the existing 100 years instrumental time series of data for sea surface temperature (SST), precipitation, river runoff and upwelling strength.

2. Regional Setting

[5] The North Atlantic Ocean is one of the key areas in the workings of global climate [Broecker, 1997], and on multidecadal, decadal and subdecadal time scales, the North Atlantic Oscillation (NAO) has been identified to have the most recurrent atmospheric teleconnection determining Atlantic climatic conditions. The first instrumental index of the NAO [Hurrell, 1995] is based on pressure measurements from Lisbon (Portugal) and Stykkisholmur (Iceland) and the Iberian precipitation and the Tagus (Iberia's longest river) flow has been shown to be highly determined by this single atmospheric circulation mode [Trigo et al., 2002, 2003].

[6] From late fall to early spring rain falls in the Tagus drainage basin contributing to the Tagus flow. The subsequent runoffs transport terrigenous materials to the coast and offshore, where they accumulate in the shelf area off Lisbon. The presence in those sediments of continental materials can be used as a proxy for the Tagus River flow and winter-spring precipitation allowing the reconstruction through time of the determining winter atmospheric circulation mode, that is, the NAO phases.

[7] From late spring to late summer, the occurrence of strong northerly winds generates coastal upwelling along the western Portuguese coast. During this upwelling season, a front between the cold recently upwelled waters and the offshore waters becomes visible in satellite images that also show the existence of upwelling filaments mainly associated to the capes [Sousa and Bricaud, 1992; Relvas et al., 2007].

[8] The upwelling jet rooted at Cape of Roca, has an approximate N-S orientation, it is a persistent feature and influences mainly the NW half of the Tagus mud patch. As such, primary production at site 28B is upwelling derived, while at the location of site 26B, primary productivity is mainly associated with the Tagus River discharge of nutrients as demonstrated by Cabeçadas et al. [1999]. Although both winter-spring river input of nutrients and summer upwelling feed primary productivity in the area and result in the export of organic carbon, calcareous and siliceous microfossils to the sediments, the combination of a river flow proxy with proxy(ies) for upwelling conditions should allow the reconstruction of the processes that result from these two seasons' main atmospheric conditions.

3. Material and Methods

3.1. Core Characterization

[9] PO287 cores were retrieved in May 2002, during the PALEO1 campaign aboard the German RV Poseidon (Figure 1). Box and gravity cores PO287-26B (52 cm) and 26G (327 cm) were recovered from 38°33.49′N, 9°21.84′W and 96 m water depth, while box core PO287-28B (52 cm) was collected at 38°37.47′N, 9°30.87′W and 105 m water depth. Piston core D13902 (600 cm) was recovered from 38°33.24′N, 9°20.13′W from 90 m water depth during the research cruise D249 on board RV Discovery.

Figure 1.

Geographic location of the studied area in relation to the Iberian Peninsula and sites where the instrumental river flow is available. The blowup map shows also the location of box cores PO287-26B and 28B, as well as of gravity core PO287-26G and piston core D13902. Reconstructions of the Tagus River flow and upwelling index are done through the application of the equations encountered by the proxy-instrumental data correlation to the D13902 data used by Abrantes et al. [2005].

3.2. Methodology Used in the Analyses of the Various Parameters

[10] Magnetic susceptibility (MS in K10−6 SI) was continually measured on a multiparameter logging system and bulk sedimentary iron (Fe) was measured, as counts per second (cps), with a profiling X-ray fluorescence scanner at 1 cm spacing [Jansen et al., 1998]. Both measurements were done at the University of Bremen.

[11] All other analyses, including grain size, alkenones with 37 carbon atoms [37alq] and other terrigenous biomakers such as n-alkanes and n-alcohols ([n-OH]) and diatoms, were done continuously on 1 cm sediment slices. The methods used for each determination were the ones described by Abrantes et al. [2005].

[12] Diatom assemblages were defined on the basis of the identification and counting of a minimum of 300 specimens. For samples with very low diatom abundances, which increase the counting time to more than 2 days per sample, the number of identified specimens was reduced to 100 on the basis of the work of Fatela and Taborda [2002].

3.3. Instrumental Sea Surface Temperature

[13] Instrumental data for SST and Precipitation are the ones compiled at the CRU Web site (http://www.cru.uea.ac.uk/cru/data/). The SST data corresponds to mean mensal sea surface temperature for an area of 5° × 5° and cover the years 1870 to 1999.

3.4. Tagus River Flow

[14] River flow data has been obtained from the Portuguese National Service for Hidric Resources (SNIRH) (http://snirh.inag.pt). In this paper we use a 103 years data set that results from the combination of the data from different stations, Vila Velha de Rodão (VVR) which covers 1901 to 1974, and Santarém for which there is data between 1944 and 2004. The fact that both stations cover the years 1944 to 1974 allowed for us to verify the agreeability of these two stations' measurements through the calculation of the correlation between the two variables for the winter months (December to March (DJFM) and January to April (JFMA); Figure 2 (left)). Besides, we have verified how this longer data set (103 years) compares with the 82 year record of the Fratel hydrologic station, a data set considered of higher quality and used by Trigo et al. [2002]. Figure 2 (right) shows the regression line and high correlation of those data sets.

Figure 2.

Correlation between the Tagus River total flow in 105 dam3 as instrumentally measured at Vila Velha de Rodão (VVR) and Santarém, and between those and the Fratel station. The values compared are the mean values calculated for the NAO Winter months, December to March (DJFM, solid squares) as well as the mean values for the same number of months but with a 1-month delay, that is, January to April (JFMA, open squares).

3.5. Upwelling Index

[15] Upwelling indices were determined at the National Oceanic and Atmospheric Administration (NOAA) as follows: first, geostrophic winds were calculated from the FNMOC 6-hourly pressure analysis, which is on a 1-degree grid spacing after 1969, a 3-degree grid spacing after 1947, and a 5-degree grid spacing after 1898. Wind stress and Ekman transport were then calculated from these geostrophic winds. Finally, the Ekman transport is rotated to get the offshore component, provided that the orientation of the coast is known, and following Bakun [1973] (http://las.pfeg.noaa.gov/las/doc/global_upwell.html).

[16] The 1-degree index was calculated for four different geographic locations in the vicinity of our sites (38.5°N 9.5°W; 38.5°N 10.5°W; 37.5°N 10.5°W and 39°N 9.5°W), but their comparison showed very good agreement for the overall region (Figure 3). As so, we have chosen the data for 38°N, 9°30′W, the closest to the study sites, for comparison to the proxies for upwelling-derived primary productivity.

Figure 3.

Upwelling indices' (m3/s/100 m of coastline) variability through time (1967 to 2002) as estimated for four different locations on the Portuguese coast following Bakun [1973].

3.6. Multiple Stepwise Regressions

[17] One of the first steps in proxy calibration is the study of the existence of possible “links” between instrumental data and the properties that we can measure in the sediments [Lopes et al., 2006]. These “links” are measured and expressed as statistical values of significant correlations, strong regressions and robust calibrations and/or validations [Crosta and Koç, 2007]. The ultimate goal of these statistical analyses is to develop transfer functions (statistical models that allow us to reconstruct the instrumental measurements back in the past).

[18] As previously mentioned, the temporal length of the instrumental data and the resolution of the sediment records are not suitable enough, at this point, to develop transfer functions (as they are defined [e.g., Crosta and Koç, 2007]). However, regression coefficients can be found and point us on the right direction in order to achieve this goal.

[19] The multiple stepwise regressions were done using the instrumental data as the independent variables and the several sediment measured properties as the dependent variables. The stepwise regression is a process in which at each step the addition or not of a variable is evaluated by an F test at the 95% confidence level [Pisias et al., 1997]. The final equations are thus evaluated by the final regression coefficient and the RMSE (root mean squared error).

4. Chronology

[20] An age-depth model for both cores has been constructed by integrating the 210Pb determinations done at the Nederlands Instituut voor Onderzoek der Zee (NIOZ) for 9 levels in each box core and three deeper samples from core D13902 for base activity determinations.

[21] AMS 14C analysis was preformed on a sample from a 36–37 cm depth in box core 26B, reservoir corrected by 400 years [Abrantes et al., 2005] converted to calendar ages with the INTCAL04 data set [Reimer et al., 2004]. Calibrated ages are presented in years anno Domini (A.D.).

[22] Correlation between the two cores 28B to 26B was further done on the basis of the MS record (Figure 4). Furthermore, as discussed by Abrantes et al. [2008], the minimum in total 210Pb found in both box cores represents a 5 to 8 cm “tsunamite” emplaced at an estimated age of 1970 ± 4, and is attributed to the 1969 A.D. earthquake-related tsunami felt in the Lisbon area.

Figure 4.

The 210Pb (mBq/g), magnetic susceptibility (MS − K × 10−6 SI), and mean grain size (μm) distribution along box cores (top) PO287-28B and (bottom) PO287-26B. Gray band marks the levels disturbed by the 1969 earthquake and tsunami [Abrantes et al., 2008].

[23] The definition of the 2,000 year long spliced sequence age model (D13902 and PO287 26B, G) followed the line of reasoning presented by Abrantes et al. [2005] although the conversion of the conventional AMS 14C ages into calendar ages is based on the new INTCAL04 data set [Abrantes et al., 2008].

[24] The final estimated sedimentation rates vary between 0.14 and 1.45 cm/a with a mean value of 0.73 cm/a in core 26B, while in 28B they oscillate between 0.24 and 1.41 cm/a with a mean value of 0.63 cm/a.

5. Results and Discussion

5.1. Precipitation and River Input

[25] The precipitation regime in the Iberian Peninsula is characterized by large interannual variability as well as high variability in the amount and distribution of rainfall [Fiuza, 1984; Trigo and DaCamara, 2000], but most precipitation occurs mainly between November and April. The Tagus is the longest river of the Iberian Peninsula and satellite-derived estimations of its plume suspended load, between May 1992 and January 1993 [Williams, 1994], indicate that the Tagus suspended load is high throughout the year, as actually revealed also by a September space shuttle image [Abrantes et al., 2005, Figure 8]. However, short but strong rainfall can lead to flood events that can cause a river flux up to 12000 m3/s, 106 times higher than the month average (0.9 m3/s), and, 30 times higher than the mean annual flux (400 m3/s). During the February 1979 flood event, the suspended load reached 300 mg/l and a stationary value of 20 mg/l was found for the following month [Vale, 1981].

[26] As part of the essentially fine-grained particles transported by the river, one can include Fe particles (resultant from the weathering of continental rocks and soils), and continental derived organic matter. The Fe content of the sediments has been used as a simple chemical proxy for the input of land-derived materials and Fe variations are considered to be a direct measure of rainfall and river input by several authors such as Haug et al. [2001] and Itambi et al. [2009]. On this basis, the Fe content of this sedimentary sequence has been assumed to be mainly determined by continental precipitation and thus able to portray winter precipitation conditions on the continent. Therefore it can be calibrated to winter mean river flow (mean value of the considered month's total flow measured in Santarem and/or VVR) and precipitation. A comparison of the Fe (cps) content on both cores, to the Tagus winter (JFMA) mean flow anomaly and a 5 point mean average run of this anomaly for the winter mean flow is presented in Figure 5. Although not comparable point by point, a good general agreement is observed in terms of trend and the correlation coefficients between Fe and the mean river flow at all seasons for each site as well as both sites is significantly related to the river flow (Table 1) and consequently to the magnitude of winter precipitation. On this basis, quantitative estimations of the river paleoflow at this location should be possible through the application of regression equations that are obtained from the comparison of the instrumental data set to the proxy data of 26B and 28B, and its application to the data covering the last 2,000 years in the D13902 and PO287-26G cores [Abrantes et al., 2005].

Figure 5.

Distribution of the Fe content (cps, brown thick line), the winter (JFMA) Tagus River Mean Total Flux anomaly (green thin line), and a five-point mean average run of the (JFMA) Tagus River Mean Total Flux anomaly (brown thin line) along cores (top) PO287-26B and (bottom) PO287-28B. Data relative to the tsunami-disturbed layer are not included.

Table 1. Correlation Coefficients Found Between Proxies and Instrumental Data for a 99.9 Confidence Levela
 River Flux VVR and SantaremRiver Flux FratelUpwelling 1 × 1Upwelling 3 × 3Upwelling 5 × 5Instrumental SST
Winter (D-M)Winter (J-A)AnnualWinter (D-M)Winter (J-A)SpringSummerFallUpseasonAnnualSummerWinter (D-M)Winter + SpringSummerSummerAnnualUpseasonSummerWinter (D-M)Spring
  • a

    N represents the number of observations. Numbers in italic represent significant correlation coefficients at p > 0.01, and numbers in bold refer to the most significant correlation coefficient, either positive or negative.

26BN = 70 N = 53      N = 24   N = 70N = 70N = 70   
   SST (°C)0.080.140.090.050.070.070.190.010.130.350.350.030.170.08−0.21−0.14−0.18−0.25−0.310.03
   [n-alk] ng/g−0.06>−0.01−0.10−0.10−0.080.060.10−0.120.09−0.48−0.17−0.53−0.580.13−0.18−0.15−0.09−0.09−0.04−0.08
   [n-OH] ng/g0.00>0.03−0.02−0.09−0.070.040.07−0.120.10−0.54−0.37−0.51−0.740.210.000.230.180.080.070.26
   [37alq] ng/g0.06>0.020.010.060.09−0.09−0.270.03−0.26−0.47−0.15−0.55−0.570.340.310.190.110.100.240.10
28BN = 79 N = 79      N = 36   N = 79N = 79N = 79    
   SST (°C)−0.12−0.06−0.06−0.05−0.040.080.20−0.150.140.370.210.450.470.01−0.10−0.24−0.14−0.12−0.11−0.19
   [n-alk] ng/g0.090.090.110.150.120.140.080.020.110.280.06−0.030.190.19−0.18−0.43−0.28−0.27−0.20−0.37
   [n-OH] ng/g0.030.030.090.130.120.07−0.140.00−0.070.15−0.11−0.170.010.280.21−0.09−0.050.010.00−0.03
   [37alq] ng/g0.00−0.03−0.030.040.03−0.10−0.360.01−0.300.11−0.13−0.11−0.030.170.210.070.030.010.060.11
   [38alq] ng/g0.020.010.090.170.14−0.07−0.220.06−0.220.270.00−0.190.010.230.16−0.03−0.04−0.060.030.02
26B+28BN = 149 N = 132      N = 60   N = 149N = 149N = 149    
   SST (°C)0.030.080.050.020.040.060.17−0.020.110.350.270.220.270.06−0.16−0.17−0.17−0.20−0.12−0.04
   [n-alk] ng/g0.000.02−0.02−0.02−0.04−0.030.090.000.03−0.05−0.08−0.18−0.130.04−0.15−0.09−0.03−0.03−0.09−0.02
   [n-OH] ng/g0.010.02−0.01−0.03−0.04−0.040.040.040.01−0.15−0.18−0.26−0.280.11−0.020.120.130.070.080.16
   [37alq] ng/g0.01−0.010.010.040.060.05−0.16−0.02−0.09−0.020.17−0.04−0.030.120.15−0.01−0.04−0.040.03−0.03
26BN = 35 N = 24      N = 16   N = 19N = 36     
   FWDiatom Ab−0.18−0.13−0.080.08−0.11−0.04−0.05−0.08−0.02−0.40−0.76−0.65−0.69−0.360.12     
   Total Diatom Ab−0.10−0.030.070.200.250.26−0.16−0.13−0.110.650.530.450.500.040.34     
   Upwelling Forms Ab−0.09−0.040.030.130.270.33−0.14−0.09−0.130.780.650.680.640.040.33     
28BN = 37 N = 31      N = 12   N = 19N = 38     
   FWDiatom Ab−0.10−0.050.01−0.040.010.03−0.110.03−0.04−0.55−0.40−0.17−0.15−0.030.37     
   Total Diatom Ab−0.03−0.040.090.050.15−0.06−0.440.20−0.42−0.64−0.27−0.59−0.520.010.33     
   Upwelling Forms Ab0.070.050.180.150.220.01−0.340.25−0.34−0.440.05−0.53−0.440.000.20     
26B + 28BN = 72 N = 55      N = 28   N = 38N = 74     
   FWDiatom Ab−0.13−0.07−0.02−0.02−0.02−0.01−0.090.03−0.03−0.51−0.47−0.30−0.31−0.130.16     
   Total Diatom Ab−0.05−0.040.080.050.13−0.02−0.350.13−0.30−0.010.04−0.160.010.030.28     
   Upwelling Forms Ab0.020.020.130.100.170.02−0.280.16−0.270.150.27−0.010.120.010.16     
Fe 26B (N = 39)0.580.600.580.490.460.630.450.290.52           
Fe 28B (N = 36)0.350.310.510.450.530.44−0.010.230.16           
Fe 26B + 28B (N = 75)0.430.400.470.410.450.470.140.210.26           

5.2. Upwelling and Productivity

[27] A qualitative and quantitative study of the water column phytoplankton along the Portuguese coast at the four seasons by Moita [2001], confirms the findings of Sousa and Bricaud [1992] and Abrantes and Moita [1999] in identifying coastal upwelling as the determinant process for the primary production patterns observed off Portugal. On the basis of the same work, the phytoplankton assemblage associated with that hydrographic process, dominant during both spring and summer, was mainly composed of chain forming diatoms, of small and medium size like Chaetoceros. A compilation of the observations for the Tagus prodelta region in particular, is presented in Table 2 and reveal that here, as along the entire coast, diatoms only dominate the phytoplankton community during upwelling periods [Moita, 2001].

Table 2. Contribution of Coccolithophores and Diatoms to the Water Column Total Phytoplankton Biomass at the Two Site Locationsa
 Chl a (mg/m3)Total Phytoplankton (#cells/L)Diatoms (#cells/L)Coccolithophores (#cells/L)Dinoflagellates (#cells/L)
26B     
   Summer 19853.86.4E+043.6E+031.4E+047.7E+03
   Winter 19860.42.4E+043.2E+031.5E+041.5E+03
   Spring 19865.82.0E+041.0E+031.6E+045.8E+02
28B     
   Summer 19852.58.3E+044.4E+032.8E+042.4E+04
   Winter 19860.28.0E+038.8E+026.4E+034.8E+02
   Spring 19865.02.1E+041.1E+031.2E+043.6E+02

[28] Abrantes and Moita [1999] compared the distribution pattern of total phytoplankton biomass, diatoms and coccolithophores abundance (# cells/l and % abundance) as well as diatom assemblages in the water column, during a nonupwelling and an upwelling situation, and the same groups distribution in the sediments. Their results indicate that the total diatom distribution pattern in the sediments is a clear record of the diatom dominance of the phytoplankton during the upwelling season. Furthermore, also the coccolithophores, even though clearly more important in winter phytoplankton, show a distribution in the sediments that is connected to the distribution of the same group during upwelling periods. On the basis of these results the authors conclude that, on the Portuguese margin diatoms produced during blooms are preserved with greater efficiency than those produced during nonbloom periods as defended also by Nelson et al. [1995].

[29] Upwelling conditions through time should then give a reasonable indication of the related productivity at the study sites. As a measure of the upwelling conditions, we have used the upwelling index calculated by NOAA following Bakun [1973]. In the sediments, total diatoms and Chaetoceros resting spores can be the proxies for productivity generated by upwelling. Figure 6 shows the comparison of the upwelling-related genus Chaetoceros abundance along the two box cores and the various upwelling indices. The first observation that can be noticed is the higher diatom abundance in core 28B, an expected result considering the continuous influence of upwelling conditions at this site. A second very important observation is the fact that diatoms disappear in both sites below 1950, that is, 20–25 cm downcore, certainly due to important dissolution. When we compare the diatom data to the upwelling indices, a good agreement is visible for the 1-degree index, but it becomes less and less good when the lower-resolution indices are used, an evaluation confirmed by the correlations to the upwelling indices (Figure 6 and Table 1).

Figure 6.

(top) Distribution of upwelling-related diatoms' abundance (#valves/g) in box cores PO287-26B (thin gray) and PO287-28B (thick gay) and the estimated upwelling indices, 1967–2001, 1 × 1° (thick orange); 1947–2003, 3 × 3° (thin orange); 1900–2005, 5 × 5° (thin green and different scale). (bottom) Distribution of the abundance (#valves/g) of upwelling-related diatoms in box cores PO287-26B (thin gray) and PO287-28B (thick gray), now on different scales, and the higher-resolution estimated summer upwelling index (1967–2001, 1 × 1°, thick orange).

[30] These results call our attention for the need to use high-resolution instrumental data if a comparison to high sedimentation rate sequences is the aim.

5.3. Sea Surface Temperature

[31] Instrumentally recorded sea surface temperature (SST) variability over the last century shows that, on a regional scale, interannual to decadal climate variability is the norm [Rodrigues et al., 2009]. Estimates from the sediment record are based on the U37k′ index, an index derived from the alkenones, a biomarker produced by coccolithophores. The index calculation followed the equation proposed by Müller et al. [1998], for which the error of estimate is of 1.5°C. Visual observation of its variability during the last century (Figure 7) shows little oscillation around 16°C at site 28B. At site 26B on the contrary, and even though the average value is also 16°C, the amplitude of variability is up to 4°C. When these estimated SST records are compared with the longer available series of instrumental SST from the SW coast of Portugal, the 28B alkenone-derived SST mainly represents winter temperature values, while the high variability observed for 26B closely follows the summer upwelling variability. Nevertheless, no significant correlations exist between the alkenone-derived SST and the instrumental SST. However, if correlations to all instrumental data sets are calculated, the PO287-26B alkenone-derived SSTs do show a positive significant correlation to the summer 1-degree resolution upwelling index (Table 1), indicating both the interplay between upwelling and the lower temperature of river waters as also visible in satellite images [Rodrigues et al., 2009, Figure 6]. That is, in years in which the strong upwelling conditions affect the normal river flow, warmer waters dominate site 26B. As for core 28B, the only significant correlation found is a positive correlation for the 1-degree resolution upwelling index calculated for the winter and spring months (Table 1). Thus, at core 28B location, the alkenone producers dominate the phytoplankton (as it can also be seen in Table 2) mainly during the winter-spring seasons and appear to be generated during winter upwelling pulses, therefore recording winter upwelling waters temperature.

Figure 7.

A. Comparison of the alkenone-derived SST along PO287-26B (thick blue) and 28B (thin blue) with the instrumentally measured winter SST (thick red). B. Comparison of the alkenone-derived SST along PO287-26B (thick blue) and the higher resolution estimated summer upwelling index (1967–2001, 1 × 1° - thick orange).

[32] According to Moita [2001] at site 26B location coccolithophores abundance (as number of cells/L) in the summer plankton is equal to the values found in winter and spring [Moita, 2001, Table 2]. During upwelling conditions the increase in new nutrients and the uncoupling observed between phytoplankton and zooplankton blooms work together toward the occurrence of exceptional export conditions. In addition, the data of Abrantes and Moita [1999] shows that there is better preservation capacity of organisms produced during bloom conditions. Therefore, at site 26B, where the influence of upwelling is intermittent, occurring only in years of strong upwelling and/or low river flow, it is to be expected that the U37k′ temperature (alkenone-derived SST) signal found in the sediments be weighted toward the upwelling-season months.

5.4. Reconstructions

[33] Although the good agreement found between the various independent traditional semiquantitative methods allowed us to reconstruct the history of river flow and primary productivity during the last 2,000 years [Abrantes et al., 2005], for those findings to be useful for climate modeling we need quantitative reconstructions. One way to quantify the existing relationship between each one of the acting processes that we want to reconstruct and the various sediment proxies that reflect them is through the calculation of existing regressions. In the regression equations analysis we allowed all the measured variables from the sediments to interact. The reasoning for this approach is that there is more than one physical process interacting in our study area. That is, strong upwelling conditions during some months and strong river influence during others. Since each sediment sample represents in average two years, the processes mentioned before will be represented by all the measured properties, in different proportions.

5.4.1. River Flow

[34] Figure 8 presents the spring and summer river flow (RF) reconstructed values using the multiple regression equations found. For the river flow, we obtained the following regressions, considering the full time series or a 3 year averaged time series:

equation image
equation image
equation image
Figure 8.

Spring and summer river total flow reconstructed from the application of functions (1), (2), and (3) to the data of piston core D13902 (lines with dots) and gravity core PO287-26G (lines) [Abrantes et al., 2005]. Fe content as well as the δ18OW variation downcore are also plotted for comparison. (1) RFsummer = 2172.58 − 5.66[37alq] + 3.37[n-alk] (28B) − 91%, (2) RFspring = −14873.83 + 3.0352[Fe] (28B) − 55%, and (3) RFspring = −8673.46 + 2.0790[Fe] + 0.6680[n-OH] − 2.1293[n-alk] (26B) − 94%.

[35] Surprisingly [Fe] does not enter in equation (1). In this case, the summer river flow signal is given by a combination of C37 total alkenones concentration ([37alq]) and total n-alkane concentration ([n-alk]). This is most probably a reflection of a combination of a lower transport capacity by the river but also of low sediment availability.

[36] Downcore (Figure 8), the highest Fe contents (as cps units) are observed above 1200 A.D., while a broad minimum occurs between 500 and 1200 A.D. When the two equations for spring river flow are applied quite close values are retrieved after 1200 A.D., but diverge for older periods and give negative values. Considering the co-occurrence of these negative values with low Fe contents, those negative values are interpreted as a reflection of a lack of instrumental data for low river flow conditions, that is, a no-analog problem. Although, the occurrence of negative values does not instill much confidence on the estimated values, at least the regression equations using independent proxies yield similar predictions that are also comparable with the water oxygen isotopical signal (δ18OW), another independent proxy for river flow presented by Abrantes et al. [2005] (Figure 8).

5.4.2. Upwelling Conditions

[37] The main proxy for upwelling are the diatoms, in particular the upwelling-related genus Chaetoceros; however, the disappearance of diatoms in both cores below ±20 cm is indicative of the strong opal dissolution observed in this region.

[38] For the upwelling indexes (UI) the following regressions were found:

equation image
equation image

[39] Equation (4) relates the upwelling diatom genus Chaetoceros and Fe to the 5-degree resolution summer upwelling index; however, as it includes diatoms, its utilization is limited to the top levels, that is, where diatoms are present (Figures 6 and 9).

Figure 9.

Summer and winter spring upwelling index reconstructed from the application of functions (4) and (5) to the data of core D13902 [Abrantes et al., 2005]. (4) UIsummer = 140.02 + 3.47 × 10−5 (Chaetoceros abundance) − 0.0171[Fe] (28B) − 64% and (5) UIwinter+spring = 573.19 − 0.0346[n-OH] − 0.266[37alq] (26B) − 64%.

[40] A second function (5) relates the 1-degree Winter+Spring upwelling to [37alq] and total concentration of n-alcohols ([n-OH]). Its application downcore to data measured by Abrantes et al. [2005] along the PO287-26G and D31902 composite sequence results in the curve presented in Figure 9 and although the small number of data points in the Little Ice Age (LIA) interval, the existing values indicate lower upwelling conditions as compared to the Medieval Warm Period (MWP).

[41] The trends found downcore when equations based on independent proxies for both river flow and upwelling confirm the indications given by the direct proxies, that is, Fe for river flow and diatoms for upwelling. However, the objective of quantification is not achieved, since different regressions lead to totally different values, demonstrating that we cannot trust the reconstructed flow in quantitative terms.

[42] Given the high correlation between the atmospheric conditions along the Iberian margin and the NAO index [Hurrell, 1995], the inferred low river flow and relatively higher upwelling during the MWP but higher river flow and lower upwelling during the LIA have been interpreted as a sign of dominant NAO minima during the LIA, and a more long-term positive NAO state in the MWP [Abrantes et al., 2005]. In this study, one of our final goals was to extend the reconstruction of the NAO index through the use of marine sediments proxy data. However, the disturbing effect of the sedimentary sequence by historical earthquakes and tsunamis prevents this otherwise high sedimentation rate site to be a good continuous site. Furthermore, the lack of instrumental data for extreme conditions generates a no-analog problem and prevents a trustable reconstruction.

6. Conclusions

[43] We describe an attempt to reconstruct the Tagus winter river flow (JFMA) using a series of proxies analyzed along two high resolution sediment box cores recovered from the Tagus mud patch, off Lisbon (Portugal). These cores cover the last 100 years, and the comparison of the Fe (cps), alkenone-derived SST and upwelling-related diatoms to the Tagus River flow, SST and upwelling index prove correct the use of those sediment components as proxies to the listed climate-related variables.

[44] Alkenone-derived SST values are essentially similar to winter temperatures and pretty much constant at site 28B but highly variable at site 26B, which is a reflex of the mutual exclusive interplay of the river flow and upwelling and the different temperatures of the waters associated to each process. Furthermore, it proves that the occasional influence of a strong process, such as summer upwelling conditions at site 26B, is more important for the definition of the sediment record than the continuous influence of the river flow.

[45] Stepwise regression analyses of the instrumental and proxy data sets provided regressions that can be indicative of possible later transfer function development. The equations reflect spring and summer river flow, as well as summer and winter upwelling indices.

[46] Reconstructions obtained through the application of the encountered equations to a previously published 2,000 year composite record of cores D13902 and PO287-26G, confirm conditions estimated from proxy data. The two equations for spring river flow retrieve quite close values after 1200 A.D.; however, they diverge for older periods and give negative values, indicating an analog problem, that is, lack of instrumental data for low river flow conditions. On the contrary, the upwelling index reconstructions give positive values at all times, again indicating a lack of instrumental data for stronger upwelling conditions.

Acknowledgments

[47] This work was conducted as a follow-up of the Holsmeer project (EVK2-CT-2000-00060), with further financial support from LNEG (former INETI) and Fundação para a Ciência e Tecnologia through the INGMAR project. The authors acknowledge the captain and crew of the PO287 and D134 campaigns, C. Monteiro and A. Inês for laboratory support, and three anonymous reviewers for their constructive comments.