Assessing sea-level data from Connecticut, USA, using a foraminiferal transfer function for tide level

Marine Micropaleontology 51 (2004) 239 255 www.elsevier.com/locate/marmicro Assessing sea-level data from Connecticut, USA, using a foraminiferal transfer function for tide level R.J. Edwards a, *, O. van de Plassche b, W.R. Gehrels c, A.J. Wright b a Departments of Geography and Geology, Trinity College Dublin, Dublin 4, Ireland b Department of Earth and Life Sciences, Vrije Universiteit, Amsterdam, The Netherlands c Department of Geography, University of Plymouth, Plymouth PL4 8AA, UK Received 1 August 2003; received in revised form 7 November 2003; accepted 10 November 2003 Abstract Concerns surrounding possible future climate change, sea level rise, and their potential impacts on coastal environments, have stimulated research seeking to elucidate the relationship between ocean levels and climate at the (sub)century-scale. The need for increasingly precise reconstructions of relative sea-level change has, in-turn, driven the development of new methodologies capable of resolving fine-scale variability within salt-marsh sedimentary sequences. The use of salt-marsh foraminifera as precise indicators of past tide levels has played a central role in this process and is exemplified by a number of detailed studies conducted in Connecticut, USA. In this paper, we apply the most recent methodological advance in the reconstruction of relative sea-levels using salt-marsh foraminifera. We develop a foraminiferal transfer function for tide level, derived from the modern foraminiferal distributions of four Connecticut salt-marshes. In contrast to existing approaches that employ site-specific interpolations of local vertical foraminiferal assemblage zones, this transfer function can provide objective, quantitative and reproducible estimates of palaeomarsh-surface elevation with explicitly stated error terms for marshes throughout Connecticut, irrespective of differences in tidal range. We demonstrate the foraminiferal transfer function approach by employing it to reconstruct changes in palaeomarsh-surface elevation from three cores recovered from Hammock River marsh, Connecticut. The resulting reconstructions, which are precise to F 0.09 m, are tested using an independent dataset from the neighbouring state of Maine, and are used to assess a pivotal record of mean high water (MHW) change from this site. We conclude that the transfer function reconstructions show good general agreement with previous records of palaeomarsh-surface elevation, but with the advantages of quantified error terms and a transparent, reproducible methodology. This replicability will assist in the comparison or combination of local records of MHW change to infer large-scale, regional sea-level variations. D 2003 Elsevier B.V. All rights reserved. Keywords: foraminifera; sea-level changes; salt-marsh; holocene; connecticut; transfer function 1. Introduction * Corresponding author. Fax: +353-1-671-3397. E-mail address: edwardsr@tcd.ie (R.J. Edwards). The salt-marshes of Connecticut (CT), northeast USA, are a locus for high-precision sea-level research seeking to elucidate the relationship between 0377-8398/$ - see front matter D 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.marmicro.2003.11.003

240 R.J. Edwards et al. / Marine Micropaleontology 51 (2004) 239 255 1 For clarity we use the term elevation when referring to height relative to a defined, contemporaneous, local tide level (e.g. MHW), and the term altitude when referring to height relative to the national geodetic datum (NGVD). ocean levels and climate at the (sub)century-scale (van de Plassche, 1991, 2000; Thomas and Varekamp, 1991; Varekamp et al., 1992; Nydick et al., 1995; van de Plassche et al., 1998a). During the course of this research, the need to produce records of the highest possible resolution has necessitated the development and application of novel methodologies capable of minimising error terms associated with age and elevation. 1 The results of these studies have stimulated debate, ranging from the validity of postulated sea level-climate relationships to the reliability of the methods employed to infer them (van de Plassche et al., 1998a,b, 2001; Varekamp et al., 1998; Thompson et al., 1999, 2000; van de Plassche, 2000). Central to this debate is the manner in which local records of mean high water (MHW) are correlated or extrapolated to infer large-scale regional sea-level change. Whilst the extent to which existing records may reproduce similar patterns of change is contested, the urgent need to collect more data to test current understanding is not (Gehrels and van de Plassche, 1999). The pioneering nature of this research, however, requires that the process of validation is accompanied by a rigorous assessment of the methods used to develop the records concerned. Ultimately, the premise that a measure of record reliability can be inferred from its reproduction elsewhere is firmly founded on the notion of methodological transparency, and the consistency or replicability of results. The high resolution of these studies is achieved by employing assemblages of salt-marsh foraminifera that are vertically zoned with respect to tide level and can therefore be used as precise sea-level indicators (Scott and Medioli, 1978). The vertical assemblage zone concept was modified by Thomas and Varekamp (1991) to produce marsh palaeoenvironmental curves (MPE) that express changes in marsh elevation relative to local MHW. These curves, when combined with chronostratigraphic data, enable the reconstruction of MHW change (Varekamp et al., 1992). One complication associated with the use of assemblage zones is the fact that their composition, elevation and vertical range are variable within and between sites. As a consequence, the precision with which former marshsurface elevations can be reconstructed varies with elevation and location (Edwards et al., in press). In addition, assemblage zones frequently cover vertical intervals of decimetres, necessitating site-specific, subjective sub-division to maintain resolution across the full vertical range of environments sampled (e.g. Nydick et al., 1995). As a consequence, the uncertainties associated with reconstructing palaeomarshsurface elevations are difficult to quantify and are often assigned indicative, but ultimately arbitrary values. This variability, whilst not precluding reliable reconstruction, introduces additional variables and uncertainty that complicate the process of record comparison. Recent developments in quantitative palaeoenvironmental reconstruction, employing foraminiferal transfer functions capable of reconstructing former tide levels, provide a means of delivering objective, quantitative and reproducible estimates of palaeomarsh-surface elevation associated with explicitly stated error terms (Horton et al., 1999). The foraminiferal transfer function approach has been successfully applied to salt-marsh cores from Maine (ME), USA (Gehrels, 2000), but spatial variability in surface and sub-surface foraminiferal assemblages may result in erroneous reconstructions when these data are applied in Connecticut, USA (Gehrels and van de Plassche, 1999; Edwards et al., in press). In light of the pivotal nature of existing sea-level reconstructions from the area, and on-going research seeking to extend and refine these data, the development of similar quantitative techniques suitable for application in Connecticut is therefore both important and timely. We present a new transfer function for tide level, developed from the modern distributions of foraminifera recorded in seven transects from four Connecticut marshes: Double Beach marsh, Hammock River marsh, Menunketesuck River marsh, and Pattagansett River marsh (Fig. 1). We describe the development and application of this transfer function, and assess its performance in reconstructing palaeomarsh-surface elevation from three salt-marsh cores from Hammock River marsh. Finally, this transfer function is used to

R.J. Edwards et al. / Marine Micropaleontology 51 (2004) 239 255 241 Fig. 1. Location map showing the key study sites referred to in the text. test an existing, pivotal record of relative sea-level change from this marsh. 2. A foraminiferal transfer function for tide level The use of microfossils in palaeoenvironmental reconstruction is founded upon the premise that the modern relationship between particular indicator organisms and the environmental variable of interest has remained constant through time. In this way, by quantifying the modern distribution of the palaeoenvironmental indicators (e.g. salt-marsh foraminifera) relative to the target variable (e.g. MHW), their past distribution can be used as a proxy for former environmental conditions. The need to discern ever more subtle variations in environment has required the adoption of quantitative techniques capable of producing precise reconstructions with associated measures of uncertainty. One way of achieving this is via the use of transfer functions, which use a range of statistical methods to define and apply biota-environment relationships. Transfer functions have been used in a range of palaeoecological studies to predict variables such as ph, temperature and salinity (Birks et al., 1990; Lotter et al., 1997; Roberts and McMinn, 1998). The foraminiferal transfer function approach was first applied in the context of sea-level research by Guilbault et al. (1996) to examine coseismic subsidence in Canada. Since then, other studies in temperate salt-marsh environments have sought to test and refine this approach with a view to constructing high-resolution records of relative sea-level change for the mid to late Holocene periods (Horton et al., 1999; Gehrels, 1999, 2000; Edwards, 2001; Gehrels et al., 2002). Whilst the precise statistical technique used varies between studies, the general method is essentially the same and can be sub-divided into three stages. Stage 1 comprises the collection and analysis of modern salt-marsh foraminifera and associated environmental data (e.g. altitude 2 ) to produce patterns of foraminiferal distribution that are expressed relative to the tidal frame. These coupled assemblage-elevation data are termed modern analogues and collectively constitute a training set. In Stage 2, these data are interrogated to produce a series of taxa-environment relationships expressed as an optimal elevation and vertical range for each foraminiferal species. These relationships are then used in Stage 3 to calibrate fossil foraminiferal assemblages and, on the basis of the relative abundances of the individual species present, infer the past elevation of a sample. Here we describe these stages in the development and application of a foraminiferabased transfer function for tide level from Connecticut, USA. We express elevation in centimetres with respect to MHW, with positive elevations referring to marsh surfaces above MHW. 2.1. Stage 1: the modern training set The compilation of an appropriate training set must strike a balance between the desire for precision and 2 Measured in metres relative to NGVD.

242 R.J. Edwards et al. / Marine Micropaleontology 51 (2004) 239 255 Table 1 Summary components of the modern training set including the tidal parameters Site Date collected No. of samples Tidal data (m NGVD) MHW MTL the need for accurate and robust reconstructions. The most precise fauna environment associations will be developed from site-specific studies that derive relationships reflecting the particular set of physiographic or climatic conditions found locally. The success of this approach rests upon the assumption that these conditions have remained constant through time and, due to the site-specific nature of these parameters, is susceptible to producing erroneous reconstructions (Horton and Edwards, in press). The collection of modern analogues from a wider range of sites experiencing differing environmental controls (e.g. vegetation, salinity or tidal range), engenders greater predictive power in the resulting transfer function since it is capable of quantifying species environment relationships under a larger range of conditions (Gehrels, 2000). This increased measure of variability inevitably results in a reduction in the precision of reconstructions however, and so attention must be focussed on maximising the range of useful modern analogues whilst avoiding those that simply introduce noise into the record. The modern distribution of salt-marsh foraminifera in Connecticut varies within and between sites (Edwards et al., in press), and differences in assemblage composition are apparent between surface and sub-surface samples from the region (Gehrels and van de Plassche, 1999). A single-site approach is therefore undesirable, and here we make use of surface distributions in the form of seven transects from four salt-marshes in Connecticut (Fig. 1). These data comprise the distributions presented in Edwards et al. (in press), coupled with modern distributions from neighbouring Menunketesuck River marsh (Gehrels and van de Plassche, 1999). The resultant unscreened training set consists of 91 samples and incorporates 11 principal foraminiferal species (see Table 1). Species contributing less than 2% of any sample are removed since significant errors in species optima and tolerance may be introduced by considering small counts or taxa with low occurrences (Horton et al., 1999). Due to their scarcity, the following six species were not considered diagnostic of a particular sub-environment and were removed from the training set: Ammobaculites balkwilli, Ammobaculites dilatatus, Eggerella advena, Polysaccammina ipohalina, Textularia earlandi, Trochammina ochracea. Death assemblages are used since these are considered most applicable to fossil foraminiferal investigations and are less susceptible to seasonal variations in distribution, diversity and abundance (Murray, 2000; Horton and Edwards, 2003). Previous studies indicate that infaunal foraminifera contribute between 2% and 5% of the sub-surface assemblages recorded in the highest marsh sediments of Connecticut (Saffert and Thomas, 1998; Edwards et al., in press). Consequently, the influence of deep infaunal activity observed in other regions is not considered significant in the sub-environments contributing to this modern training set. As a consequence of collecting modern foraminiferal samples from sites with differing tidal characteristics, it is necessary to normalise the distributions with respect to the tidal frame. Here we employ the standardised water level index (SWLI) modified after Horton (1997) and defined as: SWLI ¼ ð½ðsample Altitude MTLÞ =ðmhw MTLÞŠ 100Þþ100 ð1þ where MTL and MHW refer to the altitude of mean tide level and mean high water at the study site, respectively. 2.2. Stage 2: transfer function development No. of species Double Beach July 1999 23 1.09 0.19 9 Hammock July 1999 13 0.84 0.16 13 River Pattagansett July 1999 26 0.61 0.14 13 July 2001 Menunketesuck June 1994 29 0.71 0.07 14 The fauna environment data that constitute the modern training set are used to develop a transfer function for tide level. Analysis was performed

R.J. Edwards et al. / Marine Micropaleontology 51 (2004) 239 255 243 Fig. 2. Observed versus predicted values of the standardised water level index (SWLI) for component 2 (WA-PLS) of the Connecticutbased foraminiferal transfer function for tide level (CTTF03). using the program C2 (version 1.3, Juggins, 2003), which is capable of employing a range of statistical techniques. Preliminary analysis demonstrated that whilst all methods used produced reconstructions exhibiting the same general patterns of change, weighted-averaging partial least squares regression (WA-PLS) performed best in terms of a high coefficient of determination (r 2 ) and a small root mean square error of prediction (RMSEP). WA-PLS is a modification of the unimodal method weighted averaging (WA), which considers the variance along a single environmental gradient such as elevation (or the SWLI). Unmeasured environmental variables (e.g. salinity) will also influence foraminiferal distributions. This may distort fauna elevation relationships and cause estimated elevations to differ from their observed values. WA-PLS improves predictions by using any structure present in these differences (called residuals) and, in effect, considers the combined influence of additional environmental variables such as salinity (Ter Braak and Juggins, 1993). Transfer function development is an iterative process. The initial run of the full dataset was used to identify samples exhibiting poor vertical relationships with tide level and plotting as significant outliers (residuals greater than the standard deviation of SWLI). Four of these anomalous samples comprise mixed, highest marsh assemblages that were identified by Edwards et al. (in press) as potentially reflecting disturbance and human activity at the rear of the study sites. Only one other sample, collected from the lowest elevation at Double Beach marsh, plots as a significant outlier. Lower marsh samples are more susceptible to reworking and disturbance by biological activity (e.g. fiddler crabs) and this may account for its anomalous fauna environment relationship. These five samples exhibiting poor relationships with SWLI must be excluded from the training set since failure to remove them will decrease the predictive ability of the transfer function (Gasse et al., 1995; Jones and Juggins, 1995). A second run employing the screened dataset (86 samples) produced the final transfer function for tide level (hereafter referred to as CTTF03). The performance of CTTF03 is presented in Fig. 2 and summarised using standard statistical measures in Table 2. Component 2 is selected for use on the principle of parsimony. The error associated with SWLI reconstruction is given by the root mean squared error of prediction (RMSEP). This is calculated by a process known as jack-knifing, where individual samples are removed in turn from the training set and their measured SWLI is compared to the predicted value produced by the transfer function. The result- Table 2 Summary statistics for the Connecticut-based (CTTF03) and Mainebased (METF03) foraminiferal transfer functions for tide level developed using weighted averaging partial least squares regression (WA-PLS), showing the observed and prediction (jacknifed) measures, where: RMSE = root mean square error of prediction; r 2 = coefficient of correlation; Max Bias = maximum bias CTTF03 METF03 Comp 1 Comp 2 Comp 3 Comp 1 Comp 2 Comp 3 RMSE 15.8 12 11.7 7.8 7.1 6.4 r 2 0.61 0.78 0.79 0.77 0.8 0.84 Max Bias 38.8 25.5 27 24.7 18 16.5 RMSEP 16.9 13.5 13 8.2 7.9 7.5 r 2 (jack) 0.56 0.72 0.72 0.74 0.76 0.78 Max Bias 41.6 29 29.8 27.4 24.4 17.6 (jack) The components used in calibration are highlighted in bold.

244 R.J. Edwards et al. / Marine Micropaleontology 51 (2004) 239 255 ing magnitude of the error term will vary with tidal range, but is equivalent to 0.09 m at Hammock River marsh. 2.3. Stage 3: transfer function application The final stage of the transfer function approach is the calibration of fossil foraminiferal assemblages to reconstruct estimates of SWLI. This is achieved by applying the fauna environment relationships developed in Stage 2 via the program C2 (version 1.3, Juggins, 2003). The reliability of the resultant predictions of SWLI will be greatest when they are derived from fossil samples that are similar in composition to modern assemblages used in the training set. This is assessed via the Modern Analogue Technique (MAT), release 1.1 (Juggins, 1997) which calculates the dissimilarity between the fossil sample and the 10 most similar modern samples. The squared chord distance was selected as the dissimilarity coefficient, and samples with coefficients below the 10th percentile were considered to have good analogues in the training set (Birks et al., 1990). To illustrate this procedure, a series of fossil assemblages extracted from Hammock River marsh are calibrated using CTTF03. Analysis was performed on 28 samples from core HRM-6.5, collected immediately adjacent to core F6.5 reported in van de Fig. 3. Summary diagram showing the composition of the foraminiferal assemblages from core HRM-6.5 and the predicted values of SWLI produced by the Connecticut-based foraminiferal transfer function for tide level (CTTF03).

R.J. Edwards et al. / Marine Micropaleontology 51 (2004) 239 255 245 Plassche (2000). Alternate sediment slices 1 cm in thickness were extracted between 1.82 and 1.28 m depth in core. Material was washed through 500 and 63 Am mesh sieves after the methods described by Scott and Medioli (1980). Samples were suspended in approximately 500 ml of water and sub-divided into eight aliquots using a rotary wetsplitter (Edwards et al., in press). Replicate aliquots were then examined wet under a binocular microscope and full counts are tabulated in Appendix A. The results of these analyses and the resulting reconstructions of SWLI are summarised in Fig. 3 and Table 3. MAT indicates that all Table 3 Estimated standardised water level indices (SWLIs) produced by the Connecticut-based foraminiferal transfer function (CTTF03), and corresponding reconstructed sample elevations expressed relative to local mean high water, for fossil foraminiferal assemblages from core HRM-6 Sample Depth in core (m) SWLI Elevation relative to MHW (m) Error (m) Modern analogue HRM6-128 128 262 0.42 0.09 Good HRM6-130 130 264 0.43 0.09 Good HRM6-132 132 262 0.42 0.09 Good HRM6-134 134 264 0.44 0.09 Good HRM6-136 136 262 0.42 0.09 Good HRM6-138 138 247 0.32 0.09 Good HRM6-140 140 227 0.18 0.09 Good HRM6-142 142 236 0.25 0.09 Good HRM6-144 144 244 0.30 0.09 Good HRM6-146 146 235 0.24 0.09 Good HRM6-148 148 255 0.38 0.09 Good HRM6-150 150 258 0.39 0.09 Good HRM6-152 152 243 0.29 0.09 Good HRM6-154 154 212 0.08 0.09 Good HRM6-156 156 245 0.30 0.09 Good HRM6-158 158 253 0.36 0.09 Good HRM6-160 160 245 0.31 0.09 Good HRM6-162 162 254 0.37 0.09 Good HRM6-164 164 252 0.35 0.09 Good HRM6-166 166 262 0.42 0.09 Good HRM6-168 168 247 0.32 0.09 Good HRM6-170 170 254 0.37 0.09 Good HRM6-172 172 235 0.24 0.09 Good HRM6-174 174 225 0.17 0.09 Good HRM6-176 176 221 0.14 0.09 Good HRM6-178 178 223 0.16 0.09 Good HRM6-180 180 225 0.17 0.09 Good HRM6-182 182 220 0.13 0.09 Good The results of the matching analogue technique (MAT), showing whether the fossil sample possesses a good modern analogue in the training set, are also included. fossil samples possess good modern analogues in the training set. The core commences with a mixed foraminiferal assemblage dominated by Jadammina macrescens, with lesser numbers of Trochammina inflata, Arenoparrella mexicana, Tiphotrocha comprimata and Siphotrochammina lobata. The transfer function estimates SWLIs of 220 230, equating to a marsh surface elevation 14 20 cm above MHW. The predicted SWLI values increase above 1.74 m, reflecting the disappearance of A. mexicana and S. lobata and the increase in abundance of J. macrescens. Between 1.62 and 1.38 m, the foraminiferal assemblage is characterised by J. macrescens and Miliammina fusca. A pronounced peak in M. fusca occurs at 1.54 m and is reflected in the transfer function by an abrupt excursion in predicted SWLI. The upper portion of the core above 1.38 m is characterised by an almost monospecific assemblage of J. macrescens, indicative of deposition toward the upper limit of marine influence. The transfer function produces the highest reconstructed values of SWLI for this interval, equating to marsh elevations around 40 cm above MHW. 3. Assessing transfer function reconstructions The performance of an individual transfer function can be assessed in terms of standard statistical measures such as r 2 and RMSEP, and some sense of reliability determined from MAT. Whilst these parameters provide a good indication of the internal consistency of the resulting model, it is also useful to gauge the reliability of reconstructions by examining the extent to which they are reproduced in different locations and by different methods or independent models. We use the foraminiferal record from core HRM-6.5 presented above, to examine this issue of reproducibility in two contrasting ways. In the first, we assess the pattern of reconstructions against those produced by a second, independent foraminiferal transfer function, compiled from modern foraminiferal distributions recorded in the salt-marshes of Maine. In the second, we compare the reconstructions produced from a replicate core recovered from Hammock River marsh.

246 R.J. Edwards et al. / Marine Micropaleontology 51 (2004) 239 255 Table 4 Estimated standardised water level indices (SWLIs) produced by the Maine-based foraminiferal transfer function (METF03), and corresponding reconstructed sample elevations expressed relative to local mean high water, for fossil foraminiferal assemblages from core HRM-6 Sample Depth in core (m) SWLI Elevation relative to MHW (m) Error (m) Elevation difference (m) Modern analogue HRM6-128 128 271 0.48 0.09 0.06 Good HRM6-130 130 272 0.49 0.09 0.06 Good HRM6-132 132 272 0.49 0.09 0.06 Good HRM6-134 134 273 0.49 0.09 0.06 Good HRM6-136 136 271 0.48 0.09 0.06 Good HRM6-138 138 262 0.42 0.09 0.10 None HRM6-140 140 250 0.34 0.09 0.16 None HRM6-142 142 255 0.37 0.09 0.13 None HRM6-144 144 260 0.41 0.09 0.11 None HRM6-146 146 254 0.37 0.09 0.13 None HRM6-148 148 267 0.46 0.09 0.08 Good HRM6-150 150 269 0.47 0.09 0.08 Good HRM6-152 152 259 0.40 0.09 0.11 None HRM6-154 154 239 0.26 0.09 0.18 None HRM6-156 156 260 0.41 0.09 0.10 None HRM6-158 158 267 0.46 0.09 0.10 Good HRM6-160 160 262 0.42 0.09 0.12 None HRM6-162 162 267 0.45 0.09 0.09 Good HRM6-164 164 263 0.43 0.09 0.07 None HRM6-166 166 269 0.47 0.09 0.05 Good HRM6-168 168 266 0.45 0.09 0.13 Good HRM6-170 170 268 0.46 0.09 0.10 Good HRM6-172 172 263 0.43 0.09 0.19 Good HRM6-174 174 260 0.41 0.09 0.24 Good HRM6-176 176 264 0.43 0.09 0.29 Good HRM6-178 178 261 0.42 0.09 0.26 Good HRM6-180 180 263 0.43 0.09 0.26 None HRM6-182 182 258 0.40 0.09 0.26 Good The results of the matching analogue technique (MAT), showing whether the fossil sample possesses a good modern analogue in the training set, are also included. CTTF03, the Maine training set was compiled using total (live + dead) assemblages, and employed a variant of the SWLI height normalisation procedure that considered highest astronomical tide rather than MHW (see Gehrels, 2000 for details). The modern foraminiferal assemblages used to derive these transfer functions exhibit different distributions or fauna environment relationships (Edwards et al., in press). These differences are manifest as variability in the optimal elevation and height range of individual species which will, in turn, produce differing estimations of SWLI. Consequently, each transfer function is considered as regional in scope, since both reflect the behaviour of salt-marsh foraminifera from different, distinctive functional areas. 3.1. A comparison between two regional transfer functions The foraminiferal transfer function approach has been successfully applied in Maine by Gehrels (2000). Here we use a transfer function derived from this dataset (METF03) to provide an independent assessment of the Connecticut transfer function (CTTF03). Both transfer functions are developed using the same method (WA-PLS), and their statistical performance is similar (Table 2). In contrast to Fig. 4. Palaeomarsh-surface elevation diagrams for core HRM-6.5. The grey curve is produced by the Connecticut-based foraminiferal transfer function for tide level (CTTF03), whilst the black curve is produced by the Maine-based foraminiferal transfer function for tide level (METF03). Each curve plots former marsh-surface elevation relative to mean high water (MHW) at the time of formation.

R.J. Edwards et al. / Marine Micropaleontology 51 (2004) 239 255 247 The calibration procedure (Stage 3) previously employing CTTF03 is repeated on the fossil assemblages from HRM-6.5 using METF03 (Table 4). The resulting estimates of palaeomarsh-surface elevation change are plotted in Fig. 4 along with the original reconstructions derived from CTTF03. Despite employing completely different training sets, the reconstructions of both transfer functions display similar patterns of variation. This similarity reflects the fact that, whilst the precise elevation and vertical range of an individual species varies between sites, the broader distinction between high marsh and low marsh species is more universal. Of greater note is the fact that almost 80% of the reconstructions are comparable within the error terms associated with them. The METF03 estimates consistently plot toward the upper end of the error envelope and are, on average, 0.13 m above those of CTTF03. These discrepancies are smallest (0.05 m) for reconstructions derived from the highest elevation assemblages and largest (up to 0.29 m) from lower elevation environments. The transfer functions estimate different palaeomarsh-surface elevations for six samples from the bottom of HRM-6.5. These samples are distinct from the rest of the fossil assemblages in that they all contain Arenoparrella mexicana and Siphotrochammina lobata. The modern distribution of S. lobata is similar to that of the more abundant Trochammina inflata (Edwards et al., in press), and is well represented in the Maine training set. Conversely, the characteristic lower elevation species A. mexicana, is only recorded in two samples from Maine, in contrast to 33 samples in the Connecticut training set. As a consequence, the METF03 significantly over-estimates the elevation of the former marsh surface. The modern training sets from Connecticut and Maine are compiled from samples collected across similar elevation ranges. The apparent scarcity of A. mexicana in the Maine training set may therefore indicate less favourable marsh conditions in this region, or a preference for lower elevation (un-sampled) environments relative to its preferred ecological optimum and tolerance in Connecticut. In general, the larger discrepancies between reconstructions associated with lower elevation samples reflect the fact that greater inter-site variability is evident in these environments and distributions are more strongly influenced by local conditions within the adjacent estuary (Scott and Medioli, 1980). This is also apparent in the results of MAT (Table 4). Whilst good modern analogues exist for all samples in the local training set used to derive CTTF03, 11 samples (over 50%) have no modern analogue in the Maine training set. These no modern analogue situations are most common in the lower elevation assemblages, but it should be noted that good modern analogues are returned for the A. mexicana zone. This emphasises the fact that transfer functions should not be treated as black boxes and that a firm understanding of the behaviour of foraminifera constituting the training sets underpins interpretation of the reconstructions. 3.2. Examining record replicability The successful use of palaeomarsh-surface elevation indicators in sea-level studies is dependent upon reliably combining local records to infer larger-scale changes. Detailed lithostratigraphic analysis is required to identify sediment sequences that have accumulated in high marsh environments. This is necessary to minimise record distortion caused by processes such as erosion, post-depositional mixing (bioturbation) and infaunal foraminifera. Foraminiferal analysis of a single sediment core produces a local record of palaeomarsh-surface elevation changes. The extent to which this may be regarded Table 5 Summary table of the sediment cores referred to in the text, including source publications Site Core code Chronology Publication Comments Hammock River marsh Hammock River marsh Hammock River marsh Hammock River marsh F Thomas and Varekamp (1991) F6.5 van de Plassche (2000) 4 dates (AD 500-1500) 7 dates (AD 400-1000) Chronology in Varekamp et al. (1992) Located 6.5 m north of core F HRM-F This paper As core F Located adjacent to core F HRM-6.5 This paper As core 6.5 Located adjacent to core F6.5

248 R.J. Edwards et al. / Marine Micropaleontology 51 (2004) 239 255 as representative of a study marsh will depend upon the complexity of the lithostratigraphy, but the replication of a record from another sampling location or marsh is generally considered a good indicator that larger-scale changes are being captured. In the context of foraminiferal transfer functions and sea-level change, the notion that replicability and representativity are linked is dependent upon the idea that biostratigraphic changes can be correlated across cores in a similar manner to variations in lithostratigraphy. This means that whilst the precise composition of foraminiferal assemblages varies across a marsh surface in response to a range of parameters (e.g. elevation and salinity), they must still retain a coherent elevation signal that can be reliably extracted by the transfer function approach. We explore the issue of record replicability by sampling a second core (HRM-F) collected 6.5 m away from HRM-6.5 and covering a similar depth interval (Table 5). Foraminiferal sampling and analysis was conducted as outlined for HRM-6.5. The transfer function CTTF03 was used to calibrate the assemblages and produce associated estimates of SWLI (Fig. 5). MAT indicates that all fossil samples possess good modern analogues in the training set. Comparison of Figs. 3 and 5 reveals that, whilst slight differences in the lower abundance species occurs between the two cores, the principal patterns of change apparent in HRM-6.5 are also evident in Fig. 5. Summary diagram showing the composition of the foraminiferal assemblages from core HRM-F and the predicted values of SWLI produced by the Connecticut-based foraminiferal transfer function for tide level (CTTF03).

R.J. Edwards et al. / Marine Micropaleontology 51 (2004) 239 255 249 Fig. 6. Palaeomarsh-surface elevation diagrams for core HRM-6.5 and HRM-F produced by the reconstructions of the Connecticut-based foraminiferal transfer function for tide level (CTTF03). The curve plots former marsh-surface elevation relative to mean high water (MHW) at the time of formation. HRM-F. Most notable is the replication of a peak in Miliammina fusca at 1.54 m and the subsequent transition to an almost monospecific assemblage of Jadammina macrescens. This example demonstrates that even subtle variability, such as an abrupt faunal change evident in only one or two samples, can be traced laterally between cores. Equally important is the ability of the transfer function to replicate palaeomarsh-surface elevation reconstructions between cores. Despite the general similarity of the two cores, the potential influence of low abundance taxa has already been demonstrated by the A. mexicana zone in Section 3.1. To this end, the estimates of SWLI for HRM-6.5 and HRM-F are converted into estimates of palaeomarsh-surface elevation (Fig. 6). The records from both cores are virtually identical in terms of pattern, sequence and magnitude of change exhibited. Minor differences in the records may reflect slight variations in marsh topography or the local accumulation history of the sediment cores, producing differing age depth relationships. This example demonstrates that, in accordance with previous high-resolution stratigraphic and faunal investigations, it is possible to trace and replicate decimetre-scale changes in palaeomarsh-surface elevation from salt-marsh environments (e.g. van de Plassche, 1991; Varekamp et al., 1992). 4. Testing records of relative sea-level change from Hammock River marsh The original foraminifera-based MPE curve (MPE-1) presented in Thomas and Varekamp (1991) and Varekamp et al. (1992) is derived from core F at Hammock River marsh. These data were re-used in the relative marsh elevation (RME) diagrams of van de Plassche and co-workers, who employed them in concert with revised chronologies to investigate the relationship between sea level and climate in Long Island Sound (van de Plassche, 2000; van de Plassche et al., 1998a). The centrality of these palaeomarsh-surface elevation reconstructions means that any assessment of relative sea-level changes from Hammock River marsh must include validation of this important record. The foraminiferal transfer function (CTTF03) developed here provides an ideal means of assessing the RME (MPE) approach in general, and the core F record (MPE-1) in particular. 4.1. Recalibrating core F We re-calibrate the foraminiferal assemblages presented in Thomas and Varekamp (1991) using CTTF03 to produce a new record of palaeomarsh-

250 R.J. Edwards et al. / Marine Micropaleontology 51 (2004) 239 255 Fig. 7. Palaeomarsh-surface elevation diagram for core F produced by the reconstructions of the Connecticut-based foraminiferal transfer function (grey curve with shaded error band) plotted against the original marsh palaeoenvironmental curve MPE-1 of Varekamp et al. (1992) (black curve). The curves plot former marsh-surface elevation relative to mean high water (MHW) at the time of formation. surface elevation change derived from core F. The resulting curve (RME-1) is plotted in Fig. 7 alongside the original record (MPE-1) of Varekamp et al. (1992). The similarity of the two records is striking, both in terms of the pattern and magnitude of inferred changes, and provides strong support for the interpretations of Varekamp et al. (1992). The reconstructions from marsh sub-environments above and below MHW are comparable for large portions of the sequence. This is particularly the case when the uncertainties associated with both techniques are considered. At Hammock River marsh, elevation estimates derived from the transfer function have quantified error terms of F 0.09 m. Complications inherent in the assemblage zone approach mean that no measure of uncertainty was attached to the original estimates, but a similar magnitude error term can be considered a least estimate. Discrepancies between the two reconstructions are apparent at the top of the core and centred around 0.90 m depth, where the transfer function predicts slightly lower palaeomarsh-surface elevations (greater submergence). The greatest difference between the two reconstructions, however, is associated with the samples between 1.70 and 1.40 m depth, indicating marsh emergence. The original record (MPE-1) suggests an abrupt fall in local relative MHW of 40 cm, around 1.50 m depth. A depositional environment above higher high water is inferred for the two samples between 1.49 and 1.39 m depth on the basis of the virtual absence of foraminifera (zero tests for 1.49 1.45 m and three tests for 1.45 1.39 m). The estimates produced by CTTF03 suggest that the sequence of emergence commenced deeper (earlier) in the record and was interrupted by a ca. 15 cm submergence between 1.60 and 1.50 m depth. The transfer function cannot be reliably employed to estimate palaeomarsh-surface elevations from the barren zone and this appears as a gap in the record. This effectively barren interval is significant since it produces a marsh emergence in MPE-1 that is based on negative evidence (foraminiferal absence). An important question, therefore, is whether this emergence can be identified in other cores from Hammock River marsh. This will be most conclusively demonstrated if the cores are recovered from a location that was never fully above marine influence and contains a continuous foraminiferal record of changing palaeomarsh-surface elevation. 4.2. New data from Hammock River marsh The duplicate foraminiferal records from HRM- 6.5 and HRM-F shown in Fig. 6 cover the same depth interval in which the barren zone is recorded in core F, and provide an opportunity to examine the nature of marsh elevation change during this critical time period. Core HRM-F was recovered from within 0.5 m of core F, whilst core HRM-6.5 was recovered 6.5 m away. The RME curves produced

R.J. Edwards et al. / Marine Micropaleontology 51 (2004) 239 255 251 Fig. 8. Palaeomarsh-surface elevation diagrams for cores F (RME-1), HRM-F (RME-F), and HRM-6.5 (RME-6.5) produced by the reconstructions of the Connecticut-based foraminiferal transfer function plotted against the original marsh palaeoenvironmental curve MPE-1 of Varekamp et al. (1992). The curves plot former marsh-surface elevation relative to mean high water (MHW) at the time of formation. Numbers refer to marsh emergence or submergence (see text). by CTTF03 for each core are plotted in Fig. 8, in combination with the original curve MPE-1 of Varekamp et al. (1992). The three RME curves produced by CTTF03 show good agreement, particularly given the differences in sampling interval and sample thickness used in the analysis of core F. The curves of RME-6.5 and RME-1 both commence with sediments that accumulated at, or slightly below, MHW (1). Both records then show marsh emergence between 1.75 and 1.65 m depth, and reconstruct similar palaeomarsh-surface elevations to RME-F when this record commences at 1.65 m depth (2). The pronounced submergence centred on 1.55 m depth in RME-F and RME-6.5 (3) is also evident, but less dramatic, in RME-1. This attenuated submergence signal is to be expected given the thicker sampling interval used in core F. Above this abrupt submergence, a fall in local relative MHW resulting in marsh emergence (4) is apparent from HRM-F and HRM-6.5. This appears to directly precede the barren zone in core F where, once again, a possibly attenuated change in palaeomarsh-surface elevation is intimated. The data gap in RME-1, relating to the barren interval in core F, corresponds to a new phase of submergence (5) apparent in RME-F and RME-6.5. Finally, the upper sections of the records, above 1.40 m depth, also show good agreement in reconstructed palaeomarsh-surface elevation with a final phase of marsh emergence (6). Whilst the similarity between the three records produced by CTTF03 is clear, the pattern of change exhibited in MPE-1 is more equivocal. Fig. 7 indicates the overall comparability of the MPE-1 and RME-1 curves, but suggests that the variability in marsh-surface elevation is poorly resolved in the interval surrounding the emergence (1.7 1.4 m depth). This is almost certainly a consequence of the difficulties inherent in distinguishing subtle variations in foraminiferal assemblage within vertical assemblage zones. The result is a very attenuated sequence of change shown in MPE-1 below 1.50 m depth. In fact, if the samples between 1.70 and 1.60 m depth are elevated by around 20 cm, the same pattern of change emerges. Whilst the pattern of relative marsh-surface elevation change is replicated through much of the cores, the records clearly diverge around the inferred emergence between 1.49 and 1.39 m depth in MPE-1. In fact, RME-F and RME-6.5 indicate an interval of marsh submergence at this depth (5). This corresponds to a zone of Miliammina fusca above the pronounced peak in the abundance of this species that is responsible for producing the earlier, abrupt submergence (3) (Figs. 3 and 5). Whilst this interval, characterised by

252 R.J. Edwards et al. / Marine Micropaleontology 51 (2004) 239 255 M. fusca, is present in HRM-F and HRM-6.5, it is not observed in core F. Given this difference, and in the light of the similarities evident in Fig. 7, it is likely that the discrepancies between the reconstructions are primarily a consequence of the assemblages used to derive them rather than the contrasting approaches employed. The differences in foraminiferal assemblages are either real, reflecting local-scale variability in environment; or apparent, arising from differences in sampling strategy, preparation or analysis. Local differences in marsh accretion and topography could account for the faunal discrepancies, although this seems unlikely given the proximity of cores F and HRM-F, and the similarity between HRM-F and HRM-6.5. On the basis of the RME data, the difference in marsh surface elevation between core F and HRM-F must have been at least 0.35 0.40 m. This magnitude of difference should be reflected in the lithostratigraphy but there is no evidence of this kind. van de Plassche (2000) notes no lithostratigraphic evidence for marsh emergence as indicated by the absence of foraminifera in core F. Differences in sampling interval and sample thickness have already been noted, but given that the presence of Miliammina fusca is noted in a series of samples from HRM- F and HRM-6.5, this cannot satisfactorily account for the observed discrepancies. Alternatively, it is possible that taphonomic processes, in part relating to sample processing, are involved in altering the composition of the fossil assemblage. Samples from HRM-F and HRM-6.5 were processed wet and counted in suspension, retaining the water content of the sediment throughout. In contrast, samples from core F were dried, suspended in alcohol and then counted after the alcohol had evaporated (Thomas and Varekamp, 1991). This process may selectively remove smaller specimens and those with weaker tests. Previous studies have noted that M. fusca is more susceptible to degradation than other common agglutinated taxa such as Jadammina macrescens and Trochammina inflata (De Rijk and Troelstra, 1999; Goldstein and Watkins, 1999). The disintegration of dried, small M. fusca tests on addition of water has also been observed. The removal of M. fusca from HRM-F and HRM-6.5 produces a virtually monospecific assemblage of J. macrescens. The effect of this is to remove the submergence (5) and produce a prolonged phase of emergence comparable to that shown in MPE-1. 5. Conclusions The distillation of high-resolution records of relative sea-level change from the salt-marshes of New England has proceeded in concert with the development of new, high-precision methodologies in chronology construction and marsh elevation determination. The pioneering MPE (RME) curve approach of Thomas and Varekamp (1991), extrapolated from the seminal work on salt-marsh foraminiferal zonation (Scott and Medioli, 1978, 1980), has permitted detailed interrogation of the coastal sedimentary record of environmental change. The development and application of the foraminiferal transfer function for tide level presented here builds upon and supports this work, and can be viewed as the latest evolution of a technique with a proven and established pedigree. These results demonstrate that the transfer function reconstructions show good general agreement with previous records of palaeomarsh-surface elevation, but with the advantages of quantified error terms and a transparent, reproducible methodology. The similarity of reconstructions suggests that existing records of MHW change developed from MPE curves can be meaningfully compared with new data derived from foraminiferal transfer functions. The collection of data from critical areas and time periods can be used to provide important new information concerning MHW changes in the region. As the number and length of records increase, valuable new insights into the nature of the ocean climate relationship will be gained, coupled with an improved understanding of the ways in which larger scale environmental changes are expressed locally in natural coastal archives. Acknowledgements The manuscript was prepared at Trinity College Dublin and facilitated by a Trinity award. This research is a contribution to the project Coastal Records, currently funded by the Vrije Universiteit Amsterdam. WRG received support from NERC (small grant GR9/04647).

Appendix A. Fossil foraminiferal data Table A1 Fossil foraminiferal assemblages from core HRM-6.5. Unidentified specimens refer to deformed or small (juvenile?) tests Sample Depth in core (m) J. macrescens M. fusca T. inflata H. manilaensis T. comprimata B. pseudomacrescens A. mexicana S. lobata T. earlandi. T. ochracea Not ID No. of tests counted HRM6-128 1.28 360 8 0 0 0 0 0 0 0 368 294 HRM6-130 1.30 144 1 0 0 0 0 0 0 0 0 0 145 116 HRM6-132 1.32 215 2 0 0 2 0 0 0 0 0 1 220 176 HRM6-134 1.34 101 0 0 0 0 0 0 0 0 0 0 101 81 HRM6-136 1.36 99 2 0 0 0 0 0 0 0 0 0 101 81 HRM6-138 1.38 518 87 0 0 2 1 0 0 0 0 0 608 486 HRM6-140 1.40 451 182 2 0 1 0 8 0 0 0 1 645 516 HRM6-142 1.42 226 68 0 0 2 0 0 0 0 0 0 296 237 HRM6-144 1.44 390 77 0 0 6 0 0 0 0 0 1 474 379 HRM6-146 1.46 227 74 1 0 2 0 0 0 0 0 0 304 243 HRM6-148 1.48 132 10 0 0 1 0 0 0 0 0 0 143 114 HRM6-150 1.50 215 11 0 0 2 0 0 0 0 0 1 229 183 HRM6-152 1.52 147 33 0 1 0 0 0 0 0 0 1 182 146 HRM6-154 1.54 196 157 0 2 0 0 0 0 0 0 0 355 284 HRM6-156 1.56 86 17 0 0 0 0 0 0 0 0 1 104 83 HRM6-158 1.58 186 13 0 1 7 0 0 1 0 0 0 208 166 HRM6-160 1.60 148 23 1 1 4 0 0 2 0 0 3 182 146 HRM6-162 1.62 289 22 5 2 5 0 0 1 0 0 0 324 260 HRM6-164 1.64 188 16 36 1 3 0 0 0 0 0 0 244 199 HRM6-166 1.66 16 0 2 0 0 0 0 0 0 0 0 18 14 HRM6-168 1.68 595 4 107 2 77 0 3 14 0 0 6 808 648 HRM6-170 1.70 202 0 28 1 16 0 0 2 0 0 6 255 204 HRM6-172 1.72 249 1 76 1 17 0 18 12 0 0 8 382 307 HRM6-174 1.74 133 0 61 3 13 0 16 11 0 2 2 241 193 HRM6-176 1.76 274 1 32 0 34 0 35 11 0 7 6 400 320 HRM6-178 1.78 151 1 56 0 20 0 16 15 1 1 4 265 212 HRM6-180 1.80 165 0 32 0 9 1 22 7 0 0 1 237 190 HRM6-182 1.82 96 6 36 0 8 0 13 7 0 0 4 170 136 Abundance (no./cm 3 ) R.J. Edwards et al. / Marine Micropaleontology 51 (2004) 239 255 253

254 R.J. Edwards et al. / Marine Micropaleontology 51 (2004) 239 255 Table A2 Fossil foraminiferal assemblages from core HRM-F. Unidentified specimens refer to deformed or small (juvenile?) tests Sample Depth in core (m) J. macrescens M. fusca T. inflata H. manilaensis T. comprimata B. pseudomacrescens S. lobata Not ID No. of tests counted HRMF-128 1.28 86 0 0 0 8 0 0 0 94 75 HRMF-130 1.30 120 0 2 0 3 0 0 2 127 102 HRMF-132 1.32 155 5 0 0 2 0 0 0 162 130 HRMF-134 1.34 191 0 2 0 0 0 0 0 193 154 HRMF-136 1.36 129 0 0 0 0 0 0 0 129 103 HRMF-138 1.38 223 4 0 0 6 0 0 1 234 187 HRMF-140 1.40 181 22 0 0 11 0 0 1 215 172 HRMF-142 1.42 346 78 0 0 4 1 1 0 430 344 HRMF-144 1.44 251 68 1 2 11 0 18 2 353 282 HRMF-146 1.46 214 27 0 1 4 0 8 1 255 204 HRMF-148 1.48 408 87 10 0 21 0 38 3 567 454 HRMF-150 1.50 209 32 0 0 0 0 1 0 242 194 HRMF-152 1.52 504 27 0 0 13 0 0 2 546 437 HRMF-154 1.54 112 105 0 0 4 0 0 0 221 177 HRMF-156 1.56 141 13 0 0 1 0 0 0 155 124 HRMF-158 1.58 140 2 0 0 0 0 0 0 142 114 HRMF-160 1.60 122 0 0 0 0 0 0 0 122 98 HRMF-162 1.62 118 1 3 0 1 0 0 1 124 99 HRMF-164 1.64 34 1 0 0 0 0 0 1 36 29 HRMF-166 1.66 91 6 3 0 0 0 0 0 100 80 Abundance (no./cm 3 ) References Birks, H.J.B., Line, J.M., Juggins, S., Stevenson, A.C., Ter Braak, C.J.F., 1990. Diatom and ph reconstruction. Philosophical Transactions of the Royal Society of London 327, 263 278. De Rijk, S., Troelstra, S., 1999. The application of a foraminiferal actuo-facies model to salt-marsh cores. Palaeogeography, Palaeoclimatology, Palaeoecology 149, 59 66. Edwards, R.J., 2001. Mid to late Holocene relative sea-level change in the Hampshire Basin, UK: New data from Poole Harbour. Journal of Quaternary Science 16 (3), 221 235. Edwards, R.J., Wright, A.J., van de Plassche, O., in press. Surface distributions of salt-marsh foraminifera from Connecticut, USA: modern analogues for high resolution sea-level studies. Marine Micropaleontology. Gasse, F., Juggins, S., Khelifa, L.B., 1995. Diatom-based transfer functions for inferring past hydrochemical characteristics of African lakes. Palaeogeography, Palaeoclimatology, Palaeoecology 117, 31 54. Gehrels, W.R., 1999. Middle and late Holocene sea-level changes in eastern Maine reconstructed from foraminiferal saltmarsh stratigraphy and AMS 14 C dates on basal peat. Quaternary Research 52, 350 359. Gehrels, W.R., 2000. Using foraminiferal transfer functions to produce high-resolution sea-level records from salt-marsh deposits, Maine, USA. Holocene 10 (3), 367 376. Gehrels, W.R., van de Plassche, O., 1999. The use of Jadammina macrescens (Brady) and Balticammina pseudomacrescens Brönnimann, Lutze and Whittaker (Protozoa: Foraminiferida) as sealevel indicators. Palaeogeography, Palaeoclimatology, Palaeoecology 149, 89 101. Gehrels, W.R., Belknap, D.F., Black, S., Newham, R.M., 2002. Rapid sea-level rise in the Gulf of Maine, USA, since AD 1800. Holocene 12 (4), 383 389. Goldstein, S.T., Watkins, G.T., 1999. Taphonomy of salt marsh foraminifera: an example from coastal Georgia. Palaeogeography, Palaeoclimatology, Palaeoecology 149, 103 111. Guilbault, J., Clague, J.J., Lapointe, M., 1996. Foraminiferal evidence for the amount of coseismic subsidence during a late Holocene earthquake on Vancouver Island, west coast of Canada. Quaternary Science Reviews 15, 913 937. Horton, B.P., 1997. Quantification of the indicative meaning of a range of Holocene sea-level index points from the western North Sea. Unpublished PhD Thesis. Department of Geography, University of Durham, UK. Horton, B.P., Edwards, R.J., 2003. Seasonal distributions of Foraminifera and their implications for sea-level studies. SEPM Special Publication 75, 21 30. Horton, B.P., Edwards, R.J., in press. The application of local and regional transfer functions to reconstruct former sea levels, North Norfolk, England. The Holocene. Horton, B.P., Edwards, R.J., Lloyd, J.M., 1999. Reconstruction of former sea-levels using a foraminiferal-based transfer function. Journal of Foraminiferal Research 29 (2), 117 129. Jones, V.J., Juggins, S., 1995. The construction of diatom-based chlorophyll transfer function and its application at three lakes on Signy Island (maritime Antarctic) subject to differing degrees of nutrient enrichment. Freshwater Biology 34, 433 445.

R.J. Edwards et al. / Marine Micropaleontology 51 (2004) 239 255 255 Juggins, S., 1995. Modern Analog Technique program, version 1, Department of Geography, University of Newcastle. Juggins, S., 2003. C2 Data Analysis program, version 1.3, Department of Geography, University of Newcastle. Lotter, A.F., Birks, H.J.B., Hofmann, W., Marchetto, A., 1997. Modern diatom, cladocera, chironomid, and chrysophyte cyst assemblages as quantitative indicators for the reconstruction of past environmental conditions in the Alps 1 climate. Journal of Paleolimnology 18, 396 420. Murray, J.W., 2000. The enigma of the continued use of total assemblages in ecological studies of benthic foraminifera. Journal of Foraminiferal Research 30 (3), 244 245. Nydick, K.R., Bidwell, A.B., Thomas, E., Varekamp, J.C., 1995. A sea-level rise curve from Guilford, Connecticut, USA. Marine Geology 124, 137 159. Roberts, D., McMinn, A., 1998. A weighted-averaging regression and calibration model for inferring lakewater salinity from fossil diatom assemblages in saline lakes of the Vestfold hills: a new tool for interpreting Holocene lake histories in Antarctica. Journal of Paleolimnology 19, 99 113. Saffert, H., Thomas, E., 1998. Living foraminifera and total populations in salt marsh peat cores: Kelsey Marsh (Clinton, CT) and the Great Marshes (Barnstable MA). Marine Micropaleontology 33, 175 202. Scott, D.B., Medioli, F.S., 1978. Vertical zonations of marsh foraminifera as accurate indicators of former sea-levels. Nature 272, 528 531. Scott, D.B., Medioli, F.S., 1980. Quantitative studies of marsh foraminiferal distributions in Nova Scotia: implications for sea level studies. Cushman Foundation for Foraminiferal Research Special Publication (of the Journal of Foraminiferal Research), 17. Ter Braak, C.J.F., Juggins, S., 1993. Weighted averaging partial least squares regression (WA-PLS): an improved method for reconstructing environmental variables from species assemblages. Hydrobiologia 269/270, 485 502. Thomas, E., Varekamp, J.C., 1991. Paleo-environmental analyses of marsh sequences (Clinton, Connecticut): evidence for punctuated rise in relative sealevel during the latest Holocene. Journal of Coastal Research 11, 125 158. Thompson, W.G., Varekamp, J.C., Thomas, E., de Boer, J.Z., 1999. Neotectonics on coastal Connecticut: evidence from the Farm River marsh, Brandford. EOS, Transactions of the American Geophysical Union 1999 Spring Meeting 80 (18), S136 (Supplement). Thompson, W.G., Varekamp, J.C., Thomas, E., 2000. Fault motions along the eastern border fault, Hartford Basin, CT, over the past 2800 years. EOS, Transactions of the American Geophysical Union 2000 Spring Meeting 81 (19), S311 (Supplement). van de Plassche, O., 1991. Late Holocene sea-level fluctuations on the shore of Connecticut inferred from transgressive and regressive overlap boundaries in salt-marsh deposits. Journal of Coastal Research Special Issue 11, 159 179. van de Plassche, O., 2000. North Atlantic Climate Ocean Variations and Sea Level in Long Island Sound, Connecticut Since 500 cal yr AD. Quaternary Research 53, 89 97. van de Plassche, O., van der Borg, K., De Jong, A.F.M., 1998a. Sea level climate correlation during the past 1400 yr. Geology 26, 319 322. van de Plassche, O., van der Borg, K., De Jong, A.F.M., 1998b. Sea level climate correlation during the past 1400 yr: reply. Geology 27, 190. van de Plassche, O., Edwards, R.J., van der Borg, K., de Jong, A.F.M., 2001. 14 C wiggle-match dating in high-resolution sealevel research. Radiocarbon 43 (2A), 391 402. Varekamp, J.C., Thomas, E., van de Plassche, O., 1992. Relative sea-level rise and climate change over the last 1500 years. Terra Nova 4, 293 304. Varekamp, J.C., Thomas, E., Thompson, W.G., 1998. Sea level climate correlation during the past 1400 yr: comment. Geology 27, 189 190.