Rabu, 04 November 2020

Nonlinear landscape and cultural response to sea-level rise - Science Advances

sea.indah.link

Experimental design

We undertook field studies to collect sedimentary records of past sea level and landscape changes. Recovered sediment cores were dated using OSL and radiocarbon techniques and underwent age-depth modeling. Paleoenvironmental reconstructions were based on palynological analyses (including charcoal), phytosociological classifications, and nonmetric multidimensional scaling (nMDS). Sea-level reconstructions were based on foraminiferal analyses and a combined transfer-function/indicative meaning–based approach. Bayesian inference of sea-level trends from observational data was modeled using errors-in-variables integrated Gaussian process. Sea-level trends were extended beyond observational data using glacial-isostatic adjustment (GIA) modeling. Paleogeographies were developed by combining a model of present-day bathymetry and topography with GIA model outputs and a new paleotidal range model. Archeological indices were developed using SPDs of regional archeological radiocarbon dates and by using a database of local, age-inferred archeological evidence.

Surveys and core collections. Surveys and core collections took place across Scilly in 2009 and 2010 during low spring tides. An audit of intertidal peat deposits that were recorded by (i) English Heritage in the Intertidal and Coastal Peat Database, (ii) the Historic Environment Service in the Rapid Coastal Zone Assessment for the Isles of Scilly, and (iii) the Cornwall Archeological Unit in The Early Environment of Scilly (27) was carried out and digitized by Cardiff University (23). The audit guided field surveys of extant deposits, which were surveyed using a Trimble 4700 RTK GPS base and rover system and converted to the British National Grid coordinate system using the Ordnance Survey (OS) National Grid Transformation OSTN02. Elevation was converted to Ordnance Datum St Mary’s using OS National Geoid Model OSGM02. Baseline postprocessing used local reference stations from the OS National Survey GPS Network to reduce GPS position errors. Mean horizontal and vertical positional uncertainties across all surveys were ±0.04 and ±0.10 m, respectively. Sediment cores for paleoenvironmental analyses were collected as hand-cut monoliths from 15 locations exhibiting thick peat sequences (table S1).

The Cornwall and Isles of Scilly Maritime Archeology Society (CISMAS) carried out a submarine geophysical survey on board the vessel Tiburon of DiveScilly to map subtidal peat deposits around the islands using a type C-Max CM2 side-scan sonar with Garmin 76C EGNOS–enabled GPS for positioning (23). An area of 1.35 km2 across seven search areas was surveyed in 2009, revealing subtidal peat deposits in two search areas. The thickness of the deposits was measured with a SyQwest Stratabox subbottom profiler, and search areas were ground-truthed during 29 dives by CISMAS divers. Sediment cores were collected at 13 sites using percussion coring and hand-cut monoliths and geolocated using a Garmin 76C EGNOS–enabled GPS with Chart Depth and then estimated from Admiralty Chart number 0883 (table S1).

Chronologies. Chronologies were developed for the collected cores and monoliths using radiocarbon and OSL dating techniques. Samples of plant and animal macrofossils, wood, charcoal, and bulk organic sediment were sent for radiocarbon dating at either the Oxford Radiocarbon Accelerator Unit (OxA), University of Oxford, or the Scottish Universities Environmental Research Centre (SUERC), University of Glasgow. Sample pretreatment and AMS measurements followed OxA and SUERC laboratory procedures. Where identifiable macrofossils were unavailable for dating, humin and humic fractions of bulk organic sediments were dated following pretreatment and age determinations were calculated using weighted means (23). Of the 70 samples sent for dating, 78 age determinations were returned from the two facilities; replicate determinations were obtained from the humin and humic fractions of organic bulk sediment and 16 samples failed during pretreatment due to low carbon yields. Age determinations were catalogued with relevant sample metadata (stratigraphy, depth, unit thickness, positioning, elevation, and error estimation) in dataset S1.

For OSL dating, samples were taken from sediment monoliths under subdued red-lighting conditions. The light-exposed surface material of the sediment monoliths was removed and used for assessment of the dose rate (Gy/ka), based on measurements of finely ground material made using Daybreak detectors for thick source alpha counting and a Risø GM-25-5 beta counter. The units subsampled for OSL dating were all within 20 cm of the uppermost surface of the 30-cm-long monolith tin and were bracketed by different stratigraphic units. The gamma dose rate to the sample was therefore calculated according to the principles outlined in Appendix H of Aitken (1985) (39), using the multilayer gamma model of I. Bailiff and S. Barnett (University of Durham). The non–light-exposed sediments within the monolith were used to determine the equivalent dose (De, Gy). Coarse-grained (i.e., sand-sized) quartz was prepared for dating, using treatment with a 10% (v/v) dilution of 37% hydrochloric acid (HCl) to remove carbonates, followed by treatment with 20 volumes of hydrogen peroxide (H2O2) to remove organics, before sieving to narrow (~20- to 30-μm interval) grain-size ranges and density separation using sodium polytungstate to separate the quartz from other minerals. The quartz-rich fraction was then treated with hydrofluoric acid (HF; 40% for 45 min) to remove the alpha-irradiated outer surface of the quartz grains and to dissolve any feldspars present, followed by treatment with concentrated (37%) HCl for 45 min and finally resieved on drying as a further purification step.

Measurement sequences for quartz OSL dating were applied to 24 aliquots of each sample, which were screened using signal intensity, background signal levels, recuperation, recycling, and OSL infrared depletion ratio criteria (see English Heritage RRS 2013-2 for details) (39). The minimum number of aliquots passing screening criteria for any given sample was 18/24 (table S2); the few aliquots that failed these screening criteria tended to do so on the basis of recycling ratios exceeding ±10% of unity. Values of De were calculated as weighted means of screened aliquots with root mean square error (RMSE) SEs, with the exception of one sample (161/LPPM1-1), where the simple arithmetic mean of the De values and SD was used to reflect the broad distribution and hence relatively large uncertainty in the De value (table S2). The dose rate, De values, and final ages determined for each sample are shown in table S1, expressed in years relative to a datum of 1950 CE to enable direct comparison with calibrated radiocarbon ages. Analysis of two modern analog intertidal surface samples (161/LPTR1-M and 161/LPT3-M) gave results equivalent to burial for 3 ± 2 years for both samples, suggesting that incomplete bleaching is not a problem for the intertidal samples in this OSL dating study.

Bayesian age-depth modeling was used to provide chronologies for sediment cores and monoliths that contained multiple dates from radiocarbon and OSL dating (23). Age-depth models were built using OxCal v4.1 (40) using the IntCal13 calibration curve (41) with the P-sequence function and the deposition rate prior defined as log_10(k/k_0), where k_0 = 1, allowing k to take any value between 0.01 and 100. Highest posterior density intervals throughout the sediment cores are reported in calibrated years before present (1950 CE) as modeled mean and 2σ ranges (dataset S2).

Paleoenvironmental reconstructions. Paleoenvironmental reconstructions were based on pollen of the sediment cores and monoliths, which were subsampled at 1-cm resolution (23). Sediment subsamples (1 cm3) were combined with Lycopodium clavatum L. as an exotic spike in 10% HCl (4 ml) to provide concentration data. Samples were then digested in 10% NaOH (40 ml) before sieving (15- to 106-μm fraction), treated with 10% HF (4 ml), washed with 10% HCl, and underwent acetolysis before being stored in glycerol for counting. Pollen counts continued until 300 land-derived grains had been identified in each sample. Charcoal counts on two size fractions (>50 and <50 μm) were carried out simultaneously.

Pollen assemblages were assigned to clusters using Ward’s hierarchical agglomerative clustering using the “rioja” software package (42) in R (43). This unsupervised data-driven approach was used to assign pollen samples to cluster groups based on the similarity of their taxa assemblages. After assigning pollen samples to clusters statistically, a phytosociological classification approach was used to identify the frequent and abundant taxa within each group based on the number of occurrences of the taxon (the frequency), the average percentage, median, and interquartile range (26). An individual taxon’s frequency is determined by calculating its number of occurrences divided by the number of samples in the cluster and assigning one of five frequency classes based on cutoff values between each group. If a taxon appears in 81 to 100% of all samples in the cluster group, it is assigned the highest frequency class. nMDS was applied to the data using the “vegan” software package (44) in R (43) as a complementary method to summarize major variation in the dataset (fig. S1). The stress of the nMDS was used as an indicator of the quality of the fit of the ordination.

The results of the phytosociological classification (tables S3 to S6) were used to identify samples that had a greater influence of coastal processes. This was deemed necessary to differentiate pollen assemblages that have a greater forcing through human influence from those driven by coastal change (26). Clusters that included key halophytes in high frequencies (e.g., Chenopodiaceae) were screened from the dataset used to develop the land cover index for Scilly. The remaining samples were first orientated in time (x-axis age values based on age-depth modeling results of the sediment cores) and space (y-axis values based on the primary nMDS ordination axis) before regression analysis to determine a trend through the data. The regression used generalized additive models to estimate smooth, nonparametric link functions between the two variables, fit to the data using penalized regression splines in the “mgcv” software package (45) in R (43) with uncertainties determined as the variance of the SEs of the estimated trend in time.

To examine biomass burning across Scilly during the Holocene, charcoal records were analyzed using the “paleofire” (v. 1.2.3.) software package (46) in R (43). Seventeen charcoal records from the analyzed sediment cores and monoliths (table S1) were included in the analysis to create a composite charcoal curve. To facilitate intersite comparison, the 17 records were pretreated using the established protocol for transforming and standardizing individual records: (i) transforming noninflux data (e.g., particle concentration values in cm−3) to influx values (particle cm−2 year−1), (ii) homogenizing the variance using a Box-Cox transformation, (iii) rescaling the values using a min-max transformation to allow comparisons among sites, and (iv) rescaling the values to z scores using a base period of 200 years. Sites were smoothed with a 300-year half-width smoothing window and a bootstrap of 100 years (46).

Sea level. Sea level was reconstructed using assemblages of intertidal foraminifera from the fossil sediments as precise sea-level indicators, and subsampling was concentrated along sediment cores and monolith sections that exhibited high concentrations of Chenopodiaceae pollen. Known volumes of sediment (1 to 5 cm3) were washed through 63- and 500-μm mesh sieves, wet-split into equal aliquots, and counted until >100 tests were identified or until the entire sample had been counted (23). Paleomarsh surface elevations were estimated from samples with count totals of >50 tests by applying a two-component weighted average partial least squares transfer function model with inverse deshrinking to predict sample elevations from relative abundances of foraminifera (23). Paleomarsh surface elevation predictions from the fossil assemblages and their bootstrapped root mean squared prediction errors (RMSEPs) were corrected for tidal range differences between the training set location (Erme Estuary, South Devon) and Scilly (St Mary’s tide gauge) before being converted to estimates of sea level, following: S = H – I, where the position of former sea level (S) is the elevation of the sample relative to MSL (H) minus the indicative meaning (I) of the sample, in this case, the paleomarsh surface elevation result from the transfer function (dataset S2).

The transfer function approach provided 14 precise estimates of former sea level with uncertainties calculated as 2σ RMSEs of surveying, sampling, and transfer function errors. These samples all paired with precise dating (radiocarbon or OSL) constraints, which provided corresponding 2σ age uncertainties. The remaining age determinations were used to develop a further 56 sea-level index points (SLIPs), 8 of which were precise and the remainder were marine- or terrestrial-limiting. Indicative meanings for these SLIPs were based on lithologies and paleoenvironmental analyses (dataset S1), and the corresponding reference water levels and indicative ranges were based on those used for the sea-level database of Britain and Ireland (20). The 70 new index points developed in this study (dataset S1) were combined with the 20 existing limiting data points for Scilly (26) from the British and Irish sea-level database (20) to produce a new Holocene sea-level database and sea-level curve for Scilly.

Temporal trends and uncertainties in relative sea level were estimated for the Scilly region by applying an error-in-variables integrated Gaussian process (EIV-IGP) model (47) to the combined relative sea-level dataset. This method incorporates sample-specific vertical and temporal uncertainty and the uneven distribution of data points through time. Uncertainties reported in the text from the EIV-IGP model are a mean with 95% credible interval.

Tidal range. Tidal range changes through the Holocene were simulated for Scilly using a paleotidal model of the northwest European shelf seas (48). The Regional Ocean Modeling System was used to develop a three-dimensional tidal model, configured with latitude/longitude grid spacing of 1 to 2 km. Tidal elevation amplitudes were output at 1-ka intervals from 21 ka to present day, with boundary forcing for the paleotidal runs derived from a global paleotidal model (49). Effects of vertical land motion were estimated using 1-ka time slices from a GIA model for the United Kingdom (50) in combination with present-day bathymetry. Simulated tidal range changes were used to correct the indicative ranges of the sea-level indicators used in this study (dataset S1) and to account for tidal amplitude changes when calculating the area of the intertidal zone through time.

Glacial isostatic adjustment. Glacial isostatic adjustment models were used to derive relative sea-level predictions for the study region and extend sea-level histories beyond the extent of the observational data. The glacial isostatic adjustment model used builds upon Bradley et al. (50). The model was run at a spherical harmonic degree of 512° to provide ~35-km resolution across the study region and considers two different global ice-sheet models, Bradley2017 (20) and ICE5G (51), which prescribed the evolution of major ice sheets from ~122 ka to present day. Sensitivities to the choices of Earth rheology were investigated by generating relative sea-level predictions for a range of Earth model parameters with χ2 misfits calculated byχ2=1(N−1)Σi=1N((RSLip−RSLiobs)σi)2where N is total number of precise SLIPs; RSLip and RSLiobsare the predicted and observed relative sea-level data defined at given latitude, longitude, and time; and σi is 2 sigma errors on the observed SLIP (dataset S1). For a lithosphere thickness of 71 km, the minimum and maximum ranges of upper and lower mantle viscosities are 0.1 × 1021 to 1 × 1021 Pa and 1 × 1021 to 10 × 1021 Pa, respectively (54 models in total). For a lithosphere thickness of 96 km, the ranges are the same, although only 48 models were used to produce χ2 misfit contour plots (fig. S2). Earth models with the lowest χ2 misfit for each global ice-sheet reconstruction were selected for use in the study. This produced models with a lithosphere thickness of 71 km and upper and lower mantle viscosities of 0.3 × 1021 Pa and 50 × 1021 Pa, respectively, for both Bradley2017 [Bradley(71p350)] and ICE5G [ICE5G(71p350)] ice-sheet histories. An additional high-performing model with a 96-km lithosphere thickness and 0.5 × 1021 Pa and 30 × 1021 Pa upper and lower mantle viscosities using the Bradley2017 ice-sheet history was also selected for comparison.

Paleogeographies. Paleogeographies for Scilly were developed using three sea-level histories: (i) the newly developed probabilistic relative sea-level curve (the 50% confidence interval) for 7.5 to 0 ka, (ii) the relative sea-level output for the Bradley(71p350) glacial isostatic adjustment model, and (iii) the relative sea-level output for the ICE5G(71p350) glacial isostatic adjustment model. Bathymetry data for the Isles of Scilly and surrounding waters were downloaded from the Channel Coastal Observatory (www.channelcoast.org; accessed June 2018). Separate ascii files of Lidar data, collected between 18 March 2014 and 18 June 2015, were collated and gridded in ArcGIS, with a grid resolution of 30-m offshore, increasing in resolution to 1.5 m in the intertidal zone and on land. Bathymetry and Lidar data were both referenced to present-day MSL. The relative sea-level histories were combined with the present-day bathymetric grid and corrected for paleotidal range changes by extracting the amplitudes of the M2 and S2 tidal constituents at 1-ka intervals using the paleotidal model outputs developed in this study. The paleogeographies and paleotidal ranges were then used to calculate land- and intertidal-area changes for the three relative sea-level histories through the Holocene at 1 ka intervals.

Archeological indices. Archeological indices were developed using two different methods. Time series of population demographic change were inferred for southwest Britain and northwest France using regional compilations of archeological radiocarbon dates to produce SPDs (52). This use of radiocarbon dates as a proxy for population has been extensively discussed, with considerable methodological response to concerns that the preserved signal might more often be dominated by biases in modern archaeological investigation intensity rather than by changing intensity of past human activity. Equally reassuringly, the SPD for southwest England used here preserves a similar signal to other parts of southern England and exhibits good congruence with evidence for likely anthropogenic landscape impacts provided by aggregated pollen records (52). The data compilations comprise 1410 archeological radiocarbon dates from Devon and Cornwall and 1424 dates from Brittany and Normandy, collated from over 150 databases and published resources (dataset S3; see “Data and materials availability” for database citations). Dates with analytical (e.g., low carbon yields), sampling (e.g., evidence of contamination), and provenance (e.g., environmental associations rather than archeological) issues were screened from both compilations, which resulted in final dataset sizes of n = 920 (Devon and Cornwall) and n = 1214 (Brittany and Normandy). The “rcarbon” software package (53) in R (43) was used to calibrate the remaining dates with the IntCal13 calibration curve (40) and then cluster dates from same sites that occurred within 50 year bins (i.e., similar aged dates) to avoid overrepresenting disproportionately sampled phases or archeological sites in the resulting radiocarbon SPD curves. Radiocarbon distributions were summed separately for the two regions without normalizing postcalibrated dates (52). A significance test of conditional (fixed dates but with randomized attribution of each date to one of the two regions), random permutations (n = 1000) was applied across the combined regions to develop a 95% confidence interval of “expected” radiocarbon SPDs. Local departures from this confidence interval were considered significant when the observed regional SPD curves occurred outside of this distribution of global permutations.

A further secondary archeological index was developed to provide an approximation of population variability on Scilly based on time-conditional densities of archeological evidence available from the landscape (fig. S3). A dataset (n = 2411) of recorded archeological monuments from Scilly (dataset S3) was obtained from the Cornwall and Scilly Historic Environment Record (heritagegateway.org.uk). Dataset metrics included classifiers (e.g., identifiers and locators), monument types (e.g., structures, settlements, cairns, entrance graves, boundary and field markers, findspots, middens, and wrecks), and inferred relative ages. The relative ages associated monuments with known (e.g., Bronze Age), or “windows” (e.g., Medieval to Modern) of known, archeological epochs, which were subsequently used to assign calendar dates to the monuments, based on established archeological period timings for southwest Britain (22). The index was developed using a probabilistic approach by assigning the likelihood of a monument occurring within its temporal range to 1 and summing probabilities of occurrence for all monuments in 200-year bins through the Holocene (fig. S3). This approach produces an aoristic sum, which is based on the age-certainty of each monument and downweights monuments with temporal uncertainty (i.e., occurring within large windows of age-inferred archeological periods). Data points of unknown relative ages were screened from the dataset before analysis.

The Link Lonk


November 05, 2020 at 02:20AM
https://ift.tt/3kXrjte

Nonlinear landscape and cultural response to sea-level rise - Science Advances

https://ift.tt/2CoSmg4
Sea
Share:

0 Comments:

Posting Komentar