A late Holocene sea-level curve for the east coast of South Africa

South Africa’s extensive and topographically diverse coastline lends itself to interpreting and understanding sea-level fluctuations through a range of geomorphological and biological proxies. In this paper, we present a high-resolution record of sea-level change for the past ~1200 years derived from foraminiferal analysis of a salt-marsh peat sequence at Kariega Estuary, South Africa. A 0.94-m salt-marsh peat core was extracted using a gouge auger, and chronologically constrained using five radiocarbon age determinations by accelerator mass spectrometry, which places the record within the late Holocene period. Fossil foraminifera were analysed at a high downcore resolution, and a transfer function was applied to produce a relative sealevel reconstruction. The reconstructed sea-level curve depicts a transgression prior to 1100 cal years BP which correlates with existing palaeoenvironmental literature from southern Africa. From ~1100 to ~300 cal years BP, sea levels oscillated (~0.5-m amplitudes) but remained consistently lower than present-day mean sea level. The lowest recorded sea level of −1±0.2 m was reached between 800 and 600 cal years BP. After 300 cal years BP, relative sea level has remained relatively stable. Based on the outcomes of this research, we suggest that intertidal salt-marsh foraminifera demonstrate potential for the high-resolution reconstruction of relative sea-level change along the southern African coastline.


Introduction
Understanding past patterns of sea-level change is important on local (e.g. for coastal management and engineering), regional to national (e.g.national future sea-level predictions) and global scales (e.g. for understanding polar ice sheet history).Evidence of recent sea-level change can be derived from instrumental data such as tide gauges (for example see Douglas 1 ) and satellite altimetry (for example see Nerem 2 ).These records can be extended back into the Holocene by means of proxy data from archaeological sites, geomorphological features, isolation basin contacts and salt-marsh sediments. 3,4Relative sea-level (RSL) curves have been constructed for a substantial portion of northern hemisphere coastlines 3,[5][6][7][8] ; however, to date, few curves have been presented for the southern hemisphere. 9,10Recently, late Holocene RSL curves were produced for New Zealand 11 and Tasmania 12 , yet no such curve exists for South Africa.
The southern African coastline has been tectonically stable throughout the late Quaternary, which means that sea-level change in this region would have been marginally affected by postglacial eustatic rise during the end of the Pleistocene and the early parts of the Holocene. 13Holocene RSL records for the eastern and western coastlines are incomplete in extent and coarse in resolution. 13South African sea-level research has therefore relied largely on global records as a benchmark. 14During the last 7000 years, southern African sea levels have fluctuated by no more than ±3 m. 13,[15][16][17] Sea-level curves based on observational data for southern Africa indicate that Holocene highstands occurred at 6000 and again at 4000 cal years BP, followed by a lowstand from 3000 to 2000 cal years BP. 13,176][17] Evidence from the west coast suggests that there was a highstand of 0.5 m above MSL from 1500 to 1300 cal years BP, or possibly earlier (1800 cal years BP 13 ), followed by a lowstand (-0.5 m above MSL) from 700 to 400 cal years BP. 17 A lowstand along the southern coast, dated to 700 cal years BP, is evident from in-situ tree stumps exposed at low tide. 19The majority of proxy sea-level data from South Africa derive from sites on the western and southwestern coastlines (e.g.Langebaan, Knysna, Verlorenvlei and Bogenfel Pan).
High-resolution sea-level reconstructions can be achieved through the analysis of salt-marsh foraminifera, which are accurate and precise sea-level indicators as a result of their vertically zoned nature relative to tidal levels and elevation above MSL. 20,21Salt marshes experience daily and seasonal variations in salinity, flooding frequency and suspended sediment delivery, directly linked to tidal overflow. 22Surface elevations are predominantly controlled by tidal inundation (sediment delivery) and mean salinities (plant productivity). 23Foraminiferal assemblages are vertically zoned along salt-marsh surface elevational gradients in relation to inundation rates and tidal elevations; thus, in temperate regions, foraminiferal assemblages are considered accurate proxies of past sea-level change. 20,24,25By assigning indicative meanings (environmental ranges with reference to water level) to modern foraminiferal zonations, and applying these to downcore fossilised assemblages, predictions of past sea levels can be made with precisions of between ±0.05 m and ±0.2 m. 8 In the southern African context, the application of foraminifera as biological indicators has been restricted to studies of stratigraphy 26 , temperature change 27 , sedimentology [28][29][30][31] and marine records 32 .The use of salt-marsh foraminifera to reconstruct relative sea-level change has been limited to a single published study at Langebaan. 33ere we introduce an established sea-level proxy to determine proof of concept for South African sea-level research.This technique has the potential to contribute to our incomplete understanding of past sea-level change along the southern African coastline.
In this study we present a high-resolution foraminiferal analysis of a sedimentary sequence derived from the Kariega Estuary, Eastern Cape, South Africa (Figure 1).Modern foraminiferal assemblages analysed at this site exhibit clear vertical zonation across the marsh surface.Modern assemblages were used to construct a training set of foraminifera suitable for transfer function purposes which could then be applied to downcore fossilised assemblages.A late Holocene sea-level reconstruction chronologically controlled by accelerator mass spectrometry (AMS) radiocarbon dating is presented, representing the first high-resolution, continuous record of sea-level history for eastern South Africa.

Study site
The Kariega River is elongated and sinuous, with the estuary stretching approximately 18 km inland from the mouth (Figure 1).The estuary is surrounded by salt marshes, sand flats and steep slopes in the upper reaches. 34The system is hypersaline owing to a lack of freshwater input 34 ; however, scouring by tidal currents maintains a permanent connection with the sea. 35As with many estuarine systems along the Eastern Cape shoreline, Kariega has a small tidal prism, with water levels fluctuating in response to semi-diurnal and spring/neap tidal cycles. 36This system has a low turbidity, with little salinity and thermal stratification of the water column during any stage of the tidal cycle. 36The system receives variable rainfall, has a small catchment area (686 km 2 ) and is regulated by three dams.
There are three major intertidal salt-marsh systems at Kariega: Taylors, Grants and Galpins.The intertidal creeks of Taylors and Grants marshes are relatively shallow with a narrow intertidal area, while Galpins marsh is wider and more extensive. 37,40Several characteristics of Galpins salt marsh make it an ideal location for foraminiferal analysis: (1) the marsh is attached to a small estuarine embayment, and is thus protected from erosional dynamics associated with the main channel; (2) the intertidal zone exhibits clear vegetation zonation, dominated by Spartina maritima and Sarcocornia perennis; and (3) analysis of surface samples across the intertidal zone indicate clear vertical zonation in modern foraminiferal assemblages.

Methods
An extensive programme of coring using a 20-mm diameter gouge auger was conducted to establish the stratigraphy of the salt marsh.Core lithologies were described using notation developed by Long et al. 41 , based on original classification techniques by Troels-Smith 43 .A master core was taken from the high marsh (near the level of mean high water spring tides) for palaeoenvironmental analyses.Five neighbouring, stratigraphically consistent 1-m continuous sediment cores (KAR1-5) were extracted from this point in the salt marsh (33°39'04"S; 26°39'74"E) using a 50-mm diameter gouge auger.Cores were placed into polyvinyl chloride piping, wrapped in polythene sheeting and heavy duty plastic and transported to the laboratory for cold storage and analysis.Of the five cores, one was selected for foraminiferal analysis (KAR2) and another for developing a chronology (KAR4).Remaining cores are in cold storage for potential future studies.
Stratigraphic boundaries were targeted for dating to produce a late Holocene chronology.Five samples were extracted from core KAR4 for AMS radiocarbon analysis at Beta Analytic (USA).The organic fraction South African Journal of Science http://www.sajs.co.za was dated for four bulk sediment samples and the carbonate fraction for a single shell sample at the base of the core.AMS ages for bulk sediment samples were calibrated using the calibration curve SHCal04. 43e marine shell-derived age was corrected using a ∆R value of 146±85. 44This correction provided a spuriously young age, possibly as a result of shell recrystallisation.Several linear age modelling scenarios were considered, including the less parsimonious possibility that the shell age is correct and that overlying bulk sediment ages are all too old.In accordance with the principle of parsimony, the shell age was ultimately excluded and an objective Bayesian age-modelling exercise was applied to the remaining dates.An alternative explanation is that older sediments could have been washed downstream and deposited in the salt marsh.However, the consistently organic stratigraphy does not provide supporting evidence of sedimentary inwash events, as it lacks coarse sand layers.The majority of foraminiferal tests encountered were intact without etching, which does not support sediment reworking.
'Bacon' source code, 45 within the R open-source statistical environment, 46 was used to develop an age-depth model.Bacon utilises a Bayesian approach towards chronology building, which is reliant on a priori information.A mean accumulation rate of 16 cm/year was used for the model (acc.mean=16,acc.shape=2).However, as salt marsh sedimentation rates are highly variable, a low memory component was used (mem.mean=0.3,mem.strength=10).The resultant agedepth model provided age ranges at a resolution of 10-mm within 95% confidence limits.It is understood that, through using an agedepth modelling approach rather than individually dated sea-level index points, sea-level reconstruction interpretation can be compromised. 8n this instance, however, the use of Bacon facilitates the construction of a chronologically constrained sea-level history which would not be possible without age-depth modelling.
Core KAR2 was subsampled at a 20-mm downcore resolution yielding 48 samples for foraminiferal analysis following Scott and Medioli 21 and Gehrels 47 .Samples were stored in ethanol, rinsed in distilled water and washed through 63-µm and 500-µm sieves.The larger sieve was used to remove organic matter and coarse detritus; remaining sediment on the smaller sieve was subdivided into eight aliquots using a volumetric wet splitter and retained for foraminiferal analysis.Samples were kept wet throughout the counting process to prevent aggregation of organic particles and to minimise degradation.
A Leica M205C stereomicroscope with an attached DFC295 digital camera (SMM Instruments, Durban, South Africa) was used for counting and identifying foraminiferal assemblages under 60X, 80X and 100X magnification.At least 250 individuals were counted per sample 48 ; this limit was most necessary where indicator species were present in low numbers. 47Downcore fluctuations in foraminiferal assemblages were plotted using Psimpoll Version 4.263 49 ; the constrained incremental sum of squares (CONISS) 50 stratigraphically constrained ordination technique assisted in defining zonation patterns of downcore fossilised assemblages.
Detrended canonical correspondence analysis was used to determine the environmental (in this case elevation) gradient length of the modern foraminiferal data set in standard deviation (SD) units.Where gradient lengths are < 2 SD units, it is assumed species are responding linearly to the environmental variable. 51The data set demonstrated a short gradient length of 1.4 SD units and therefore a linear-based regression model was suitable for developing a transfer function. 52The transfer function was built using a partial least squares regression model in C 2

53
and the modern foraminifera counts of both live and dead assemblages were combined.Training sets composed of total assemblages are still occasionally used based on the assumption that the live assemblages will in time contribute to the fossil record 54 ; however, it is acknowledged that there are strong arguments suggesting that the dead assemblage alone is most suitable for transfer function purposes (for example, Murray 55 ).Sample-specific elevation prediction errors were calculated using bootstrapped cross validation which provided root-mean-square error of prediction values for each fossil sample.Fossil samples that were similar in composition to the training set samples were included in the reconstruction.The modern analogue technique was used to assign a minimum dissimilarity coefficient value to all fossil samples.Similarity cut-off values follow Watcham et al. 56 Samples below 0.66 m had poor fits with their closest modern analogues.These samples were manually assigned an indicative meaning associated with minimum sealevel heights only.Reconstructed elevation values for all samples were used to calculate past sea-surface elevations in relation to present MSL following Gehrels 57 .

Stratigraphy and chronology
The basal unit for this core (0.97-0.57m) comprises clay, silt and sand in approximately equal quantities.At around 0.92 m, a few unidentifiable shell fragments were recovered.A clay unit containing some silt is present from 0.57 m to 0.37 m.Above this unit is a clayey, silty sand lens, which is 0.70 m thick.From 0.30 m up, the lithostratigraphy is characterised by decreasing grain size (increasing clay content) and becomes increasingly organic.Between 0.70 m and the surface, there is abundant in-situ decaying organic material.Iron oxide staining is present between 0.26 m and 0.70 m.
Accelerator mass spectrometry radiocarbon age determinations and their calibrated ages (with standard deviations) are presented in Table 1.
The age-depth model created using Bacon is presented in Figure 2, and places the record within the late Holocene period.

Foraminiferal data
Foraminifera are present and well preserved throughout the length of the core.The majority of tests identified and counted were intact, with little evidence of corrosion.A total of 16 species were identified; the main species, alongside their common environmental niches, are presented in Table 2. Low concentrations of foraminiferal tests (<100) are present between 0.38 m and 0.26 m.Average foraminiferal test concentration for the core was 213 tests.cm 3 .CONISS cluster analysis identified five assemblage zones:, K-1 (0.94−0.65 m), K-2 (0.65−0.46 m), K-3 (0.46−0.39 m), K-4 (0.39−0.31 m) and K-5 (0.31−0 m).These zones differentiated major changes in foraminiferal populations through the core.Agglutinated foraminifera are most abundant in the upper reaches of the core and gradually decrease down the core, whereas the base of the core is dominated by calcareous foraminifera (Figure 3).Zone K-1 at the base of the core is dominated by the calcareous species Ammonia aberdoveyensis and Haynesina germanica, which commonly comprise up to 90% of the assemblage.At 0.66 m there is a shift away from these species towards Ammonia batavus and the agglutinated saltmarsh species Jadammina macrescens and Miliammina fusca (Zone K-2).J. macrescens is the most abundant species in Zone K-3 which includes low numbers of M. fusca and Trochammina inflata.Zones K-4 and K-5 are dominated by J. macrescens.From Zone K-4 upwards there is an increasing diversity of agglutinated foraminiferal species.T. inflata and Trochammina ochracea represent up to 40% of the assemblage in Zone K-5.

Quantitative reconstruction
Indicative-meaning based predictions of palaeomarsh-surface elevations were converted into sea-level estimates following Gehrels 57 .Samples with good or close modern analogue fits were reconstructed using the transfer function (Table 3).Those samples with poor fit represent foraminiferal assemblages which had no close modern analogues from the training set (Table 4).The presence of H. germanica in these samples was used to provide a sea-level estimate.This species most commonly occupies the intertidal zone 61 and therefore an indicative range from lowest astronomical tide (-1.04 m above MSL) to the lowermost sample from the training set (0.01 m above MSL) was applied to these samples.Each sample was constrained chronologically using interpolated values from the age-depth model.The reconstructed sea-level curve for Kariega Estuary (Figure 4) identifies rapid RSL fall at around 1100 cal years BP to levels down to a minimum of ~1 m below present MSL.Between 1100 and 300 cal years BP, RSL appeared to fluctuate at approximately 150-year periods, with amplitudes of around 0.5 m.After 300 cal years BP, RSL remained relatively stable.

Discussion
The Kariega Estuary appears to have been connected to the open ocean throughout the late Holocene, based on continuous foraminiferal presence throughout the core.The fossil microfauna suggest that prior to 1100 cal years BP, RSL was higher than it is today, in accordance with existing literature. 13,16,17The calcareous foraminiferal assemblages from the lowest ~0.35 m of the core imply the existence of a low intertidal, or possibly shallow subtidal, environment before 1100 cal years BP.Because of the broad habitual range of calcareous foraminifera, it is difficult to determine precise indicative meanings from these assemblages. 25,48espite this difficulty, it is clear that the reconstruction encompasses the falling limb of rapid RSL decline out of the late Holocene highstand which is documented across South African coastlines. 17,19Existing late Holocene sea-level data from South Africa is summarised in Table 5 and presented alongside the modelled reconstruction (Figure 5).There is strong agreement between the existing data and the RSL curve from the Kariega Estuary which provides a well vertically constrained estimate of continuous late Holocene RSL for this part of Africa.

Pre-1100 cal years BP highstand
A late Holocene sea-level highstand has been reported at around 1500-1200 cal years BP 13,15,17 (Figure 5).A sea-level curve from the southwestern coast, 18 and to some extent beachrock evidence from the KwaZulu-Natal coast, 14 indicate highstands between 3000 and 1200 cal years BP.The Zostera facies recovered from Langebaan, 17 a pollen record from Verlorenvlei 16 and oyster shells from Langebaan 61 present further evidence that RSL was 0.5-0.7 m higher than present at around 1300 cal years BP.This highstand corresponds with the Little Climatic Optimum, a period of temperate warming which is associated with higher sea levels around this time. 62The RSL decline out of this highstand is identified by significant changes in foraminiferal assemblages from Kariega.As RSL fell below that of modern day, the abrupt change in foraminifera populations at 0.66 m might be a result of sedimentary erosion as sea For standardisation, ages were calibrated using the same methodology employed in this paper.

1100-300 cal years BP oscillations
The Kariega record provides evidence of sea levels remaining lower (by approximately 0.5 m) than present MSL from ~1100 to ~300 cal years BP.Siesser's 39 study at Vilanculos, using beachrock cement, suggests that sea levels were equivalent to the present day from 842±215 cal years BP (Figure 5), contradicting other studies which indicate a lower RSL during this period.According to Compton 17 , two lowstands of −0.5 m to −1 m took place from 700 to 400 cal years BP (Figure 5).This lowstand is evident on the south coast at 682 ± 54 cal years BP from in-situ tree stumps exposed at Knysna Estuary during low tide 19 (Figure 5).This lowstand corresponds with lower sea-surface temperatures between 500 and 400 cal years BP; therefore steric effects associated with cool water temperatures may have been a contributing factor. 17

cal years BP to present
The RSL oscillations shown by this study and by Compton 17 terminate at around 300 cal years BP.Following 300 cal years BP, RSL shows a gradual rise to present MSL.Evidence from Langebaan Lagoon 17 (Figure 5) suggests that sea level has been rising to present levels since 400 cal years BP.The Kariega reconstruction shows relatively stable sea levels throughout the past 300 cal years BP.Exogenic subsidence linked with large-scale water extraction would likely result in recent (past 100 years) sea-level rise.Such evidence is absent from the Kariega reconstruction, suggesting that local subsidence has not occurred.
The Macassa Bay study indicates that lower sea levels are followed by a rise from 700 cal years BP until the present 38 (Figure 5).However, this multi-proxy data set implies that relatively low sea levels recorded at Macassa Bay were associated with a freshwater phase with little influence from marine processes. 38A noticeable sea-level acceleration occurs in the RSL reconstruction during the middle of the 20th century following a minor RSL fall during the early 1900s.This pattern of recent sea-level change has also been documented in New Zealand and Tasmania. 12Such similarities with other southern hemispheric RSL curves lend weight to the credibility of this study.Proposed sources for this 20th-century sea-level inflexion include northern hemispheric ice melt from Greenland. 12e Kariega record provides high-resolution sea-level data at a regional scale, offering insight into east coast sea-level fluctuations.The pre-1100 cal years BP highstand is supported by existing sea-level data, which provide good chronological estimation of when RSL declined out of this highstand.Observed fluctuations during the late Holocene have been small (~ 0.5 m amplitude) compared with the mid-Holocene (~2-3.5 m amplitude 15 ).Reconstructed patterns of recent RSL changes are in agreement with other records from the southern hemisphere, reflecting processes contributing to sea-level change occurring on hemispheric, or greater, scales.
Few tidal gauge records for the southern hemisphere extend beyond 50 years. 22This lack of long-term tidal gauge data hampers detailed validation of recent reconstructions.To identify and understand patterns of both recent (i.e.19th and 20th century) and late Holocene sea-level changes, it is necessary to consider proxy data. 1,7Further research sites should be investigated along the east coast to establish a greater modern-day analogue for South Africa and to strengthen future reconstructions.Sea-level reconstructions from neighbouring sites could present a more robust and high-resolution picture of sea-level fluctuations for the east coast of South Africa.

Conclusion
Fossil foraminiferal assemblages analysed from a ~1200-year sedimentary record from the Kariega Estuary, South Africa, identify continuous late Holocene RSL changes that show accordance with published records.The reconstruction was supported chronologically using a Bayesian age-depth model based on AMS radiocarbon age determinations.With higher resolution dating, timings of key RSL changes in South Africa could be established with previously unprecedented accuracy.This study presents the first salt-marsh foraminifera transfer function for the southern African coastline used to produce a high-resolution sea-level curve for the Kariega Estuary.A late Holocene RSL was reconstructed with mean errors of ±0.2 m.A sea-level highstand was dated to pre-1100 cal years BP.We observed fluctuations in MSL between 1100 and 300 cal years BP; yet MSL then was consistently lower than present sea levels.From 300 cal years BP, RSL increased gradually to present-day levels, and a sea-level inflexion was shown to occur at some time during the mid-20th century.The results of this study suggest that intertidal salt-marsh foraminifera demonstrate potential for high-resolution reconstruction of RSL change in South Africa.

4 VolumeFigure 2 :
Figure 2: Age-depth model for the Kariega record based on four accelerator mass spectrometry determined ages.The surface age is assumed to represent the present day.

Figure 3 : 5 Volume
Figure 3: Foraminiferal concentrations down core KAR2.Indicated in the figure is the Troels-Smith 43 stratigraphy and the zonation derived from a constrained incremental sum of squares.

Figure 4 :
Figure 4: Reconstruction of relative sea-level change for Kariega Estuary.

Figure 5 :
Figure 5: Late Holocene sea-level curve from Kariega Estuary com pared with previously published sea-level evidence from southern Africa.

Table 1 :
The basal date provides an age of 1331 to 1424 cal years BP at 0.895 m.The age determination at 0.185 m (970 to 1142 cal years BP) is identified by the model as a possible outlier.Radiocarbon dating results for the Kariega record, indicating calibrated and uncalibrated ages

Table 3 :
Performance of the partial least squares (PLS) transfer function

Table 4 :
Results from the transfer function applied to the fossil data 6 Volume 110 | Number 1/2 January/February 2014 South African Journal of Science http://www.sajs.co.za

Table 5 :
Summary of late Holocene sea-level indicators for southern Africa

Lab code 14 C year BP Cal years BP (2 SD) MSL relative to present (m) Material Locality Reference
began to rise out of the lowstand.High precision dating would be required to determine whether a hiatus exists at this level.