Recent sea-level changes in the southern Bay of Biscay: transfer function reconstructions from salt-marshes compared with instrumental data

1 departamento de Geologia, Facultade de Ciências da universidade de Lisboa. Edificio C6, 3o piso, Campo Grande, 1749016, Lisboa, Portugal. E-mail: eduleorri@yahoo.es 2 Sociedad de Ciencias aranzadi, Zorroagagaina kalea 11, 20014 donostia-San Sebastian, Spain. 3 departamento de Estratigrafía y Paleontología, Facultad de Ciencia y tecnología, universidad del País Vasco/EHu, apartado 644, 48080 Bilbao, Spain.

available and provide new estimates of ~3 mm yr -1 (Cabanes et al., 2001;Cazenave and nerem, 2004;Leuliette et al., 2004).these differences could be related to the acceleration in sea-level rise found by Church and White (2006) for the period 1870-2000, or represent decadal variability (nerem et al., 2006).Furthermore, an important question to address is to identify when the modern rate of sea-level rise began.therefore, estimating sea-level rise from sedimentary sequences of salt marshes has received increasing attention, since this research offers great potential to supplement the temporal and spatial global database of instrumental (tide-gauge) observations of sealevel change.these sequences provide a benchmark against which one must measure the additional sealevel rise that has occurred over the last 100-150 years (Church and White, 2006;Holgate, 2007). in fact, recent geologically-based research from the north atlantic has provided the first indication that modern rates of relative sea-level (rSL) rise (last 100 years) in this region may be more rapid than the long-term rate of rise (over the last 800-1000 years), and that the timing of this acceleration may be indicative of a link with human-induced climate change (Gehrels et al., 2002(Gehrels et al., , 2004(Gehrels et al., , 2005;;donnelly et al., 2004).advances in high-resolution sea-level reconstruction have been made in the last few years through the development of foraminifera-based transfer functions (Gehrels et al., 2002;Edwards et al. 2004;Gehrels, 2004;Hayward et al., 2004;Boomer and Horton, 2006;Horton and Edwards, 2006).Fossil foraminifera have been used as "proxies" for elevation by quantifying the relationship between faunal data (relative abundance of individual species) and environmental data (elevation) in the modern salt-marsh environment.these modern relationships are then applied to cores to reconstruct past tide levels from fossil assemblages within sedimentary sequences in order to reconstruct palaeomarsh surface elevation.in the southern Bay of Biscay, Leorri et al. (2008a) developed a foraminifera-based transfer function from four Basque marshes.the relationship between observed and foraminifera-predicted elevation was strong, allowing precise reconstructions of former sea levels (error ranged from 0.11 to 0.19 m). the transfer function was used to calibrate the foraminiferal assemblages collected from a 50-cm salt-marsh core.the resulting relative sea-level curve was in very good agreement with regional tide-gauge data, suggesting a rate of rSL rise of approximately 2 mm yr -1 for the 20th century.
However, we argue that a single core analysis could reflect local rather than regional forcing factors.Consequently, we hypothesise that regional signals should be recorded in different marsh areas and be coincident with tide-gauge data.therefore, we apply here the transfer function developed previously by Leorri et al. (2008a) to a 50 cm-long core recovered from the muskiz marsh (Fig. 1).We describe the application of this transfer function and also assess its performance in reconstructing palaeomarsh-surface elevation.We compare the derived relative sea-level curve with both the previously reconstructed sealevel curve from the nearby ostrada marsh (Leorri et al., 2008a) and regional Santander tide-gauge data in order to provide a quantitative assessment of the potential of intertidal foraminifera for rSL studies in the Bay of Biscay.matEriaLS and mEtHodS a 50 cm sediment core was collected from the muskiz marsh in november 2003.Compaction of the sediment during sampling was negligible due to the thin marsh lithosome, the dominance of minerogenic sediments, and the proximity to the basal contact with mesozoic-Cenozoic rocks that provide little opportunity for additional accommodation space due to autocompaction (loss of porosity due to the load of overlying sediments) (Leorri et al., 2008a).the muskiz marsh (Fig. 1) is located in the middle reaches of the Barbadun estuary and characterised by halophytic plants: Sarcocornia perennis, Sarcocornia fruticosa and Salicornia ramossissima together with Atriplex portulacoides and Puccinellia maritima, among others (anonymous, 1986).the Barbadun estuary forms the tidal part of the Barbadun river.it has a total surface area of 204 ha, a length of 4.4 km and an average width of 5-10 m (anonymous, 1999).aerial and historical photography indicate that the selected area does not have a history of reclamation.Furthermore, sedimentological investigations did not reveal any agricultural layers as identified in many other salt marshes of northern Spain (e.g.Cearreta et al., 2002).two PVC tubes (12.5 cm diameter) were inserted into the sediment in order to obtain sufficient material to determine grain size, benthic foraminifera, sediment geochemistry, organic matter, and 137 Cs and 210 Pb geochronologies.the core was described, photographed and X-radiographed before being sliced into samples of one centimetre thickness.We measured topographic elevation (Leica station; elevation error: ± 0.005 m) for the core and this information is presented relative to the local ordnance datum (lowest tide at the Bilbao Harbour on 27th September 1878).
Since the transfer function has been developed from four different marsh areas, the inundation frequency at each study area must be calculated and elevations are standardised relative to the tidal range.Hence, the elevations are expressed as a standardised water level index (SWLi; Hamilton and Shennan, 2005): where SWLI n is the standardised water level index for sample n, h n the elevation of sample n, h MTL the mean tide level elevation, and h MHHW the mean higher high water elevation, all of them expressed in metres above local ordnance datum.this produces a stand-ardised water level index (SWLi) for each modern sample, with 100 representing the mean tide level and 200 the mean higher high water.SWLi for each buried sample is then converted back to an elevation above local ordnance datum at the field site by reversing the calculations (Hamilton and Shennan, 2005).

Microfaunal study
We analysed the samples of the muskiz core for foraminiferal content at 1 cm intervals.Samples were sieved through a 1 mm sieve (to remove large organic fragments) and a 63 micron sieve and washed to remove clay and silt material.tests were picked until a representative number of more than 300 individuals was obtained, and then studied under a stereoscopic binocular microscope using reflected light.the total number of samples analysed in this core was 23, and more than 7100 foraminifera were identified.Foraminiferal species identified in this core with relative abundances greater than 2% are presented in table 1.

Radiometric analysis
Core sub-samples were dried, homogenised and then placed in sealed 9 ml Petri dishes in order to obtain the radioactive equilibrium between 226 ra and 222 rn daughters over 60 days.the concentration of existing radioactive isotopes in each sample was determined by means of a semi-planar detector (EGSP 2200-25-r, Euro-ySiS measures) coupled with a multi-channel analyser (8000 channels) at the university of Bordeaux (France).
the concentrations of the radioactive elements 137 Cs, 226 ra and 210 Pb were determined by the number of counts beneath the corresponding photopeaks, taking into account background noise and spectrum base line.the photopeaks considered here are: 661 KeV from the 137 Cs, 46.5 KeV from the 210 Pb, 352 and 611 KeV from the 214 Pb and 214 Bi, daughter products of the 222 rn in equilibrium with the 226 ra (Lederer et al., 1967).Each sample was counted for a period of 24 hours.the detection efficiency of the measuring system was calculated using samples of known activity prepared with the same geometry as the samples to be measured.the uncertainties of the measured concentrations are due mainly to the statistical counting error, and depend on each value.

Geochemical analysis
the sediments were sieved through a 1 mm sieve, oven-dried at 45ºC and mechanically homogenised in an agate mill to avoid metal contamination.metal concentrations were determined by inductively coupled plasma-optic emission spectrometry (iCP-oES) after microwave digestion with aqua regia.Lowest detection limits were 0.1 mg kg -1 for Sc; 1 mg kg -1 for mn, Zn, Cu and ni; 2 mg kg -1 for Pb and Cr; 3 mg kg -1 for as; and 0.01% for Fe and Ca.
organic matter analysis was performed following the Walkey method (Hesse, 1971).

Statistic analysis
the performance of the transfer functions is assessed in terms of the root-mean square of the error of prediction (rmSEP) and the squared correlation (r 2 ) of observed versus predicted values.the rmSEP indicates the systematic differences in prediction errors, whereas the r 2 measures the strength of the relationship of observed versus predicted values.these statistics are calculated as "apparent" measures in which the whole training set is used to generate the transfer function and assess the predictive ability, and the data were also jack-knifed (also known as 'leaveone-out' measures).jack-knifing is a measure of the overall predictive abilities of the dataset.However, it does not provide sample-specific errors for the palaeomarsh elevation reconstructions of each core sample (Birks, 1995).Bootstrapping can be used to derive a standard error of prediction (SEpred; Birks et al., 1990;Line et al., 1994), which varies from sample to sample depending upon the composition of the core assemblage and the presence or absence of taxa with a particularly strong signal for the environmental variable of interest (Birks, 1995).SEpred was estimated using 1000 cycles.We will use here the model 3 (Wa-PLS; component 3, see table 2) transfer function described by Leorri et al. (2008a).model 3 uses samples above a standardised water level index of 160, thus removing lower elevation samples (see Hamilton and Shennan, 2005 for discussion), and has low rmSEP, although it also has low r 2 jack because of the lower number of samples and the reduced length of the elevational gradient (table 2). the precision of the transfer function is comparable with other foraminifera-based transfer functions from the northern atlantic ocean, with a precision ranging from ±0.12 m to ±0.29 m (see Leorri et al., 2008a).Following back-transformation of the SWLi values, the reconstructions have a precision of ±0.14 m.
We also employed the modern analogue technique (mat; juggins, 2004) to evaluate the likely  (Horton and Edwards, 2006).rESuLtS the muskiz core was located at 4.52 m above local ordnance datum in the higher vegetated subenvironment of the muskiz marsh (Fig. 1). the uppermost 16 cm of the core are composed of dark brown laminated mud (silt and clay) with a small sand content (average 11%; range 8-16%) and are highly organic (average 3.6%; range 0.9-10.9%).Below this depth, the sediment is dominated by sand and bioclasts (average 66%; range 23-75%) with a low organic matter content (average 1.1%; range 0.6-1.6%).
the micropalaeontological analysis showed forty six species and two distinct foraminiferal assemblage zones (FaZ) in this core.the basal 31 cm (47-16 cm depth interval) are characterised by the dominance of transported, allochthonous foraminifera (average 74%).the foraminiferal assemblage is dominated by Cibicides lobatulus, and secondary Quinqueloculina lata, Quinqueloculina seminula and Cribroelphidium williamsoni (table 1).FaZ1 contains a high number of both tests g -1 (average 277; range 110-524) and number of species (average 25; range 17-31).the sand content is also high and decreases significantly in the upper 5 cm of this zone when agglutinated, marsh foraminifera enter the assemblage.By comparison with modern assemblages, FaZ1 has been interpreted as deposited in an intertidal sandy environment with normal marine salinity conditions, which become progressively shallower throughout the last 5 cm of this zone.Finally, the top sixteen cm represent current environmental conditions in the vegetated marsh area, with extremely abundant foraminiferal tests (average 1299 tests g -1 ; range 114-3809) and a very low-diversity foraminiferal assemblage (average 5 species; range 2-7).the organic matter content increases rapidly upwards in this zone from 0.9% at the base to 10.9% at the surface.allochthonous forms are very rare (average 3%) and only the agglutinated species Jadammina macrescens and Trochammina inflata are dominant.table 1 and Figure 2 summarise the micropalaeontological data.
table 3 summarises the radiometric data. 210Pb activity declines with depth and shows a small er-ror range.the sedimentological change at -16 cm prevents the use of the constant rate of supply (CrS) model (Goldberg, 1963), since this environment has probably undergone intense sediment mixing (Leorri et al., 2008b).the data should be normalised to account for the granulometric differences (Cearreta et al., 2000), making it difficult to obtain a reliable estimation of the supported 210 Pb.We therefore used the constant initial concentration (CiC) model, which provides a sedimentation rate of 2.1±0.1 mm year -1 for the upper 16 cm. 137Cs shows a clear subsurface maximum in activity at -16 cm.ascribing this subsurface activity maximum to 1963 gives an accretion rate of 4.0 mm yr -1 .However, there is no clear evolution with depth and significant values coexist with others that are well below the detection limit (table 3).
the aim of this study is not to analyse the geochemical composition of the sediments but to provide additional support for the radiometric age model.We therefore summarise the geochemical results using the Zn value as representative of the heavy metal variations, following Cearreta et al. (2000), although  137 Cs content (Bq kg -1 ), metal concentrations and metal background values (mg kg -1 ) in the muskiz core.metal background values were calculated for the Basque estuaries by Cearreta et al. (2000).CiC age model with assigned errors and metal concentration-inferred ages are indicated.age model is calculated from the sedimentation rates obtained from the combination of 210  Pb and as values are also provided since they have been recently used to support radiometric chronologies.Heavy metal concentrations are fairly constant below 16 cm and lie within background values (table 3).From this depth upwards heavy metal values increase, attaining a subsurface sharp peak at -8/-6 cm, before declining towards the surface, although at -3/-2 cm there is an additional increase, peaking again at the surface.Forty-six species were found in the muskiz core (table 1), but only J. macrescens and T. inflata were dominant throughout FaZ2 (Fig. 2).these species are indicative of a high salt-marsh environment.We selected model 3 transfer function, which consists of modern samples from higher elevations (table 2), to calibrate the foraminiferal assemblages because the micropalaeontological data suggest a salt-marsh environment for the dated section of the muskiz core.Furthermore, the buried foraminiferal data exhibit optima high in the tidal frame (above mHHW).the calibration process assigns a palaeomarsh-surface elevation to each core sample together with a bootstrapped standard error of prediction, which varies between 15 and 16 cm (Fig. 2). the foraminiferal assemblages are fairly similar throughout the FaZ2.Consequently, the transfer function predicts that the marsh surface at the sample site has varied little during deposition of FaZ2, which indicates that the marsh surface has built up largely parallel to the sea-level rise.Slight variations in the palaeomarsh elevation reflect the presence of varying abundances in secondary species.FaZ1 is dominated by calcareous foraminifera, whose modern analogues are not included in model 3 transfer function.in contrast, mat indicates that all FaZ2 samples possess good modern analogues in the training set.diSCuSSion Higher high-marsh elevations provide the most reliable sea-level reconstructions since they have a lower influx of inorganic sedimentation due to reduced frequency, duration, and depth of tidal inundation and their dense root mats prevent sediment mixing.Lower elevations are most affected by sediment mixing and pulsating events (e.g.storms), reducing their temporal acuity and providing less accurate sea-level reconstructions when compared with tide-gauge data (Gehrels, 2000;Leorri et al., 2008b).therefore, only FaZ2 in the muskiz core can be reliably dated and used to produce a relative sea-level curve.
recent sedimentation trends (ca.1850 ad to present) typically employ short-lived radionuclides ( 210 Pb, 137 Cs). 210Pb is a naturally occurring radionuclide whose vertical distribution allows ages to be ascribed to sedimentary layers based on the known decay rate of 210 Pb (see appleby and oldfield (1992) for a discussion of the 210 Pb method).this technique is restricted to the last 120 years (Gottgens et al., 1999) and needs to be supported by other chronological markers (Smith, 2001). 137Cs (half-life 30 years) is an artificially produced radionuclide present in the environment since 1954.its presence in the study area is likely to be dominantly derived from nuclear weapons testing, with peak fallout in 1963.However, the use of 137 Cs fallout peaks is increasingly viewed as uncertain due to post-depositional migration (abril, 2004).this effect could be responsible for the discrepancy with the CiC model and the lack of a clear pattern with depth, which prevents the use of 137 Cs in this core.to further support the CiC model chronology, we also analysed heavy metal concentrations that have recently been used as chronological markers in high resolution sea-level studies (e.g.donnelly et al., 2004;Gehrels et al., 2006;Leorri et al., 2008a). in core sediments with continuous records, heavy metal concentration increases gradually from background values, reflecting the increase in emissions as a result of the industrial revolution (~1800 ad). the initial divergence of metal concentrations from background values has been dated in core sediments between 1800 and 1850 (see Leorri et al., 2008a for references).therefore, FaZ2 should be younger than 1850 since all metals show values above those determined as background in the area by Cearreta et al. (2000).on the other hand, maximum metal concentration has been dated between 1969 and 1975 by numerous authors (e.g.Weiss et al., 1999;Bindler et al., 2001;Eades et al., 2002;renberg et al., 2002;Farmer et al., 2006), being most likely related to the peak in pollution reached in the early 1970s (renberg et al., 2002).ascribing this date to the peak identified at -7 cm represented by the Zn and as profiles, we obtain a sedimentation rate of 2.3 mm yr -1 , which is in excellent agreement with the CiC model-derived sedimentation rates (see table 3 for summary of data and chronologies).
Figure 3 represents the sea-level reconstruction since the 1920s.Each foraminiferal sampling point is plotted as a sea-level index point using: where RSL is the former relative sea level for sample i, E is sample elevation in metres above the local datum for sample i, and PME is the palaeomarsh elevation in metres above the local datum for sample i reconstructed by the transfer function.
Figure 3 shows an excellent agreement between the muskiz rSL reconstruction and the ostrada core inferred trend (Plentzia estuary in Leorri et al., 2008a). in fact, regression analysis through the mid-point of the reconstruction provides a general trend of 2.3±0.4 mm yr -1 for the period 1936-2003, which approximates the results obtained from the ostrada core that varied from 2.4±0.4 mm yr -1 to 2.0±0.3 mm yr -1 for the period 1884-1994, depending on whether or not the uppermost sample was included in the analysis, since it was considered a significant outlier (Leorri et al., 2008a).to further explore the reliability of the core reconstructions, we compared the results with the Santander tide-gauge data.We selected Santander due to its proximity to the study area and its record length (since 1943).the Santander tide-gauge linear trend is 2.18±0.41mm yr -1 for the period 1944-2001.However, the temporal resolution of the core reconstructions cannot Fig. 3. -relative sea-level curve for the muskiz core (Barbadun estuary) derived from both the reconstruction of the elevations produced by the transfer function and the age model chronology.Key: a-muskiz sea-level reconstruction with assigned errors and the ostrada (Plentzia estuary) sea-level trend (derived from regression analysis, 95 % confidence; Leorri et al., 2008a).B-annual and five-year sea-level obtained from the Santander tide gauge (data obtained from the Permanent Service for mean Sea Level website).C-Comparison between the sea-level index points (derived from both ostrada cores [circles] and muskiz cores [squares]) and the Santander tide gauge (envelope based on the five-year reconstruction).d-Comparison between the composite reconstruction of sea-level (sea-level index points and instrumental data) and sea-level general trends from northeastern north atlantic: dark grey- Gehrels et al. (2005), light grey-donnelly et al. (2004).
be directly compared to instrumental data that provide a five-minute resolution signal.therefore, we averaged the tide gauge data at five-year segments providing the mean and the standard deviation (1σ) for each segment (Fig. 3B). the resulting envelope is then used to compare with the reconstructed sealevel index points (see Fig. 3C). the conversion of SWLi to elevation data was done following back the calculations of Equation 1 but using mtL and mHHW corresponding to the Santander tide gauge area so that the instrumental data and transfer function reconstructions could be directly compared (Fig. 3C).as observed in Figure 3C, all but one index point fall within the five-year sea-level envelope, indicating that the reconstructions are consistent with the tide gauge record.Furthermore, the excellent agreement between the two core reconstructions and the tide gauge provides a linear trend of rSL rise of 1.9±0.3mm yr -1 since 1923.also, it seems to imply a regional significance of the rSL signal, allowing meaningful comparisons with other areas in the north atlantic region, as proposed by Leorri et al. (2008a).this composite sea-level curve for the southern Bay of Biscay resembles those reported in northeastern north america (donnelly et al., 2004;Gehrels et al., 2005) (Fig. 3d), indicating that the modern rate of sea-level rise started at the turn of the century and is roughly coincident with the temperature increase during the 20th century (iPCC, 2007).However, it is still unable to correctly identify the precise timing of the recent sea-level acceleration because of the low number of sea-level index points during the 19th century, and further sediment sequences should therefore be investigated.

ConCLuSionS
accelerating rates of GSLr are increasingly viewed as one of the most devastating impacts of future climate change.However, our understanding of multi-decadal climate-ocean relationships is poor.Geologically-based research can potentially provide accurate estimates of sea-level rise at sub-centennial scales that can be used as benchmarks to measure the additional sea-level rise that has occurred during the last century, supplementing the temporally and spatially limited instrumental data base and indicating when the modern rate of sea-level rise started.this paper has assessed the regional significance of saltmarsh sea-level reconstructions in the southern Bay of Biscay by comparing two different sediment cores recovered from two nearby estuaries and combining tide-gauge and foraminifera-based transfer function reconstructions of rSL. in this geographical area salt-marsh foraminifera show a strong vertical zonation and the use of transfer functions has allowed the first regional quantitative estimate of sea-level index points for the Bay of Biscay.We placed the reconstruction of palaeomarsh elevations into a temporal framework to produce former rSLs.the resulting rSL curve is in very good agreement with local and regional tide-gauge records providing a regional sealevel rise of 1.9±0.3mm yr -1 since 1923.aCKnoWLEdGEmEntS dr.Leorri was supported by the contract SFrH/ BPd/44750/2008 (Fundação para a Ciência e a tecnologia; Portugal).We thank nieves González for digitalising the data set.this work was partially funded by the unESCo06/08 and GiC07/32-it-332-07 research contracts.it represents a contribution to iGCP project #495.maria alday carried out micropalaeontological sampling preparation and foraminiferal counting as part of her Phd dissertation research supported by a doctoral grant from the Basque Government (Spain).two anonymous referees greatly improved the original manuscript with their comments and suggestions.
rECEnt SEa-LEVEL CHanGE in tHE Bay oF BiSCay • 289

Fig. 1 .
Fig. 1. -Location of the muskiz marsh (highlighted by the white border) in the Barbadun estuary and localities referred to in the text.Key: 1-Barbadun estuary; 2-ostrada marsh; 3-Santander.White triangle in the muskiz marsh indicates location of core.

Table 2 .
-Statistics summary of the performance of Wa-PLS (weighted averaging partial least squares) for the foraminiferal assemblages from the Basque marshes corresponding to model 3 (fromLeorri et al., 2008a).
Fig.2.-relative abundance of main foraminiferal species, predicted palaeoelevation produced by the transfer function and sedimentation rates-derived age (table 3) with depth (cm) referred to standardised water level index (SWLi).age model chronology is represented as 2σ error (dark grey area).Cross indicates heavy metal concentration-derived age.FaZ, Foraminiferal assemblage zone.