Introduction

Determining the chronological framework of sedimentary deposits is paramount for studying past and modern sedimentary systems. Carbonate sedimentary deposits are diverse assemblages of skeletal fragments that are mixed and altered by a variety of physical and biological processes. The combination of varied origins and taphonomic histories makes understanding reef-associated sedimentary deposits especially challenging. The desire to use reef sediment deposits to quantify anthropogenic impacts over the last few hundred yrs provides clear criteria against which to evaluate the temporal resolution of these fossil records.

Chronology in sedimentary systems comprises two key parameters. The most commonly quantified parameter is the relation between age and burial depth within the sediment. While this parameter is critical, it is a single value, and the variation around this value is rarely quantified beyond the analytical uncertainty of the dating method. The second key parameter, time-averaging, is the amount of time represented in a sedimentary deposit, or stratigraphic unit (Flessa et al. 1993; Kowalewski 1996). While the mean age of a sedimentary assemblage is important, it is the age distribution of the constituents within the sedimentary sample, or time-averaging, which fundamentally limits the temporal resolution of a given deposit. Quantifying time-averaging is time- and resource-intensive, so although it is of utmost importance for understanding the processes underlying nearly all paleontological and sedimentological investigations, few studies date a sufficient number of specimens to quantify the age distribution.

Lack of quantitative estimates of time-averaging was not a problem when paleontologists thought about temporal resolution at a coarse scale. Now, however, paleoecological studies spanning the period of colonial expansion and/or anthropogenic impact require a detailed understanding of the formation of the fossil assemblages being utilized. Although time-averaging in tropical sedimentary deposits is still comparatively understudied (see Kidwell et al. 2005; Krause et al. 2010; Kosnik et al. 2013), previous work has shown that reefal sediment accumulating over the past few hundred yrs is not necessarily stratigraphically ordered (Kosnik et al. 2007, 2009, 2013). One needs look no further than volume 11 issue 2 of Coral Reefs to see the controversy caused by the use of paleontological data derived from reefal sediment to examine the Holocene history of Crown-of-thorns outbreaks on the Great Barrier Reef (GBR). The most detailed critiques can be found in Fabricius and Fabricius (1992), Keesing et al. (1992), and Pandolfi (1992), but the key criticisms of the Walbran et al. (1989) study were the effect of post-death processes on the preservation of the spicules (i.e., taphonomy) and the methods used to estimate the age of the samples (i.e., chronology).

A number of factors determine the temporal resolution of a fossil assemblage. First, the sedimentation rate determines the minimum possible degree of time-averaging, with higher rates of sedimentation typically resulting in less time-averaging (Meldahl et al. 1997). Second, the rate and depth of mixing, either physical or biogenic, determine the maximum possible degree of time-averaging, with deeper and faster sediment mixing leading to more time-averaging (Olszewski 2004). Some bioturbators active in tropical sediments (e.g., callianassid shrimp) can effectively sort sedimentary grains by size, shape, and/or other characteristics (Tudhope and Scoffin 1984). Third, sediment grain durability determines the length of time that a sedimentary particle will remain intact and recognizable. Fragile sedimentary particles are more likely to be destroyed (e.g., fragmented, eroded, or dissolved) during mixing, leading to less time-averaging, whereas durable particles can be thoroughly mixed without breaking, leading to more time-averaging (Cummins et al. 1986; Kosnik et al. 2009).

Here, we examine the sediment accumulating in the One Tree Reef (OTR) lagoon to determine the suitability of this lagoonal sediment for paleoecological analyses on the timescales associated with European colonization and the economic development of the Australian continent. We use two independent methods to establish chronologies: 1. Lead-210 (210Pb) is a naturally occurring isotope associated with fine-grained sediments. It is commonly used to determine the age of sediments less than ~150 yrs. 2. Carbon-14 calibrated Amino Acid Racemization (14C/AAR) directly dates individual mineralized skeletal elements, and it is commonly used to determine the age of Holocene molluscan shells. These two methods have entirely distinct assumptions and potential biases, so they are well suited for cross-validation.

Materials and methods

Study site

One Tree Reef (OTR) is an outermost reef in the Capricorn group at the southern end of the GBR, Australia (Fig. 1). The OTR lagoon is entirely rimmed by reef crest and isolated from the ocean at large for much of the tidal cycle (the lagoon ponds at ~1.5 m above lowest astronomical tide). The main lagoon is shallow, reaching a maximum ponded depth of approximately 6 m near the northern edge, and it contains many small patch reefs that reach the ponded surface. At the shallow southern and eastern edges, the lagoonal sediment is dominated by coarse sand, while at the deeper northern edge the sediment is dominantly fine mud. OTR has been comparatively well studied due to the presence of the One Tree Island Research Station operated by the University of Sydney.

Fig. 1
figure 1

Map of the Capricorn reefs, southern Great Barrier Reef, Australia (derived from Great Barrier Reef Marine Park Authority zoning map 18, 2011)

Full size image

Relative to many shallow tropical lagoons, the OTR lagoon contains very fine sediments and very low callianassid shrimp activity. Very few places in the OTR lagoon have anything approaching the classic “callianassid volcano-field” topography observed in many coral reef lagoons. No site in the main lagoon showed signs of anything more than minor burrowing activity throughout field activities in 2012 and 2013 (Kosnik pers. obs.). The area sampled for this study had a flat surface with only a low density of small burrow holes indicating the presence of a bioturbating fauna.

In September 2012, sediments from the northwestern end of the main OTR lagoon were collected using 80 mm hand cores and a 0.5 m × 0.5 m excavation. The ponded water depth at the collection site (23.49677°S, 152.06587°E) was 4.5 m. Here, we focus on a single 1.6 m deep 80 mm diameter percussive hand core and a single 1.2 m deep 0.25 m2 excavation both collected by divers. The method of excavation differs slightly from previous work at Rib and Bramble Reefs (Kosnik et al. 2007, 2009, 2013) in that the retaining wall system was assembled without mechanized assistance and the layers were collected using an 80 mm diameter water dredge rather than an 80 mm diameter airlift. The water dredge was fitted with a small water jet to facilitate the gentle breakup of the compacted mud. As in previous studies, the excavation was collected in 5 cm layers into ~1 mm mesh bags, so complete grain size analysis of the excavated sample was not possible. The most abundant bivalve, Abranda jeanae, was selected as a representative mollusk from the >4 mm sieve size fraction for dating analyses (Fig. 2). Hand cores were taken ~0.2 m away from the excavation site in order to examine the full grain size distribution and to sample the fine sediment fraction required for 210Pb dating.

Fig. 2
figure 2

A right valve of Abranda jeanae. The scale bar is 10 mm. The photographed specimen is deposited at the Australian Museum, Sydney (C.478298)

Full size image

The depth of each excavation layer was measured relative to a fixed datum and is expressed as excavation depth. The 5 cm collection layers determine the finest possible depth resolution of the 14C/AAR dataset. The hand core was split and sampled in 1 cm thick layers. The penetration and recovery of the hand core were measured during collection, but before processing the hand core compressed 2 cm. There are a variety of quantitative methods for uncompressing the core and returning these 2 cm to the total core length, but since the primary goal of this study is to compare the two chronologies and both the original and compressed hand core depths correlate with the same excavation layer, we use compressed core depths which are 2 cm less than the collected core depths.

Sediment description

The hand core was visually homogeneous with no indication of layering or change in sediment characteristics down core. The hand core was subsampled for grain size analyses every 5 cm (i.e., 0–1, 5–6, 10–11, 15–16 cm, etc.) to align with the 5 cm thick excavation layers. These sediment samples were suspended in 15 ml of dispersant and 15 ml of distilled water by spinning for an hour before being injected into a Malvern Mastersizer 2000 where an average of three measurements was used to determine the grain size distribution. We use the 10th, 50th, and 90th quantile grain sizes provided by the Mastersizer software to characterize the grain size distribution through the ~1.6 m core length.

210Pb age determination

Bulk sediment samples from the hand core were dated at the Australian Nuclear Science and Technology Organisation (ANSTO) Environmental Radioactivity Measurement Centre using the 210Pb method, which has been used over the past 4 decades to date recent aquatic sediments up to 150 yrs old (Goldberg 1963; Appleby and Oldfield 1978; Oldfield and Appleby 1984; Appleby 2001). Dating is based upon the determination of the vertical distribution of unsupported 210Pb, a naturally occurring radionuclide that is present in aquatic sediment as a result of atmospheric fallout. By taking samples at regular intervals over the length of a sediment core and determining the atmospherically derived 210Pb activity, a chronology can be established, and a sedimentation rate calculated. The total 210Pb activity in sediments has two components: supported 210Pb activity (210Pbs) derived from the decay of in situ radium-226 (226Ra), and unsupported 210Pb activity (210Pbu) derived from atmospheric fallout. 210Pbs activity can be determined indirectly via its grandparent radioisotope 226Ra. 210Pbu activity is estimated by subtracting 210Pbs activity from the total activity of 210Pb (210Pbt) or its progeny polonium-210 (210Po) (Hollins et al. 2011).

To measure 210Po and 226Ra activities, approximately 5 g of dried sediment was spiked with artificial 209Po and 133Ba tracers. Each sediment sample was subsequently leached with hot concentrated acids to release polonium and radium. Polonium was auto-plated onto silver disks after adding the reducing agent hydroxylammonium chloride. Radium was co-precipitated with BaSO4 and collected on a fine-resolution filter paper. Each polonium disk and radium filter sources were counted by high-resolution alpha spectrometry. The activity of the artificial tracer 209Po on the polonium disk, determined by alpha spectrometry, was used to calculate the 210Po chemical recovery. The chemical recovery for 226Ra was determined by measuring the artificial tracer 133Ba on a High Purity Germanium (HPGe) gamma detector. The 210Pbu activity for each sample was calculated, and the sediment core chronology was determined using the constant initial concentration (CIC) and constant rate of supply (CRS) 210Pb age models (Appleby 2001).

The CIC age model was fit using the six samples (between 0.15 and 0.41 m), excluding the three samples (between 0 and 0.11 m) at the top of the core and the two deepest layers where 210Pbu activities were not different than zero. CIC ages were estimated for the upper three layers by linear extrapolation. The CRS age model was fit using all nine samples analyzed where 210Pbu activities were greater than zero (0–0.41 m). Of note, the CRS age estimate for the 0.40–0.41 m layer has extremely high uncertainty due to the very low 210Pbu activity measured. All of the 210Pb-derived ages are relative to the year of collection (2012 AD), and the data are summarized in Table 1.

Table 1 Hand core and lead-210 data used to determine the relation between sediment age and depth
Full size table

Abranda jeanae

The most abundant mollusk sampled in the excavation layers and in the living assemblage, Tellina (Abranda) jeanae Healy and Lamprell 1992 (Bivalvia: Tellinidae: Tellininae), is relatively large and thin shelled (Fig. 2). The figured specimen has been deposited in the Malacology Collection at the Australian Museum, Sydney (C.478298). Only complete right valves in good condition were analyzed to exclude the possibility of analyzing the same individual twice. The posterior halves of 80 disarticulated right valves were used for AAR analyses. Based on AAR results, the anterior halves of eleven shells were used for AMS 14C dating. The geometric mean size of dated specimens ranged from 4.9 to 10.8 mm with a median size of 8.4 mm. There was no correlation between dated specimen size and excavation depth (p = 0.64).

Amino acid racemization

Single samples from 80 Abranda right valves were prepared for amino acid analysis at Northern Arizona University according to standard procedure (Wehmiller and Miller 2000). The shells were cleaned by acid leaching, then each fragment was weighed and demineralized in 20 ml of 7 M HCl mg−1 of CaCO3. The total hydrolyzable amino acid (THAA) population of amino acids was recovered by hydrolyzing the sample solutions under N2 at 110 °C for 6 h. This hydrolysis procedure minimizes the amount of induced racemization in the most rapidly racemizing amino acids, especially for young shells, while the recovery of amino acids is near maximum (Kaufman and Manley 1998). Sample solutions were evaporated to dryness in vacuo and then rehydrated in 0.01 M HCl containing an internal standard of l-homo-arginine of known concentration for calculating the concentration of amino acids relative to the mass of shell. Each sample was analyzed once using reverse-phase, high-performance liquid chromatography (HPLC). All chromatograms were of consistently high quality, and none was rejected.

The chromatographic instrumentation and procedure used to separate amino acid enantiomers was described by Kaufman and Manley (1998). Briefly, the analytical method employed pre-column derivatization with o-phthaldialdehyde together with the chiral thiol, N-isobutyryl-l-cysteine, to yield fluorescent diastereomeric derivatives of chiral (d and l) amino acids. The derivatization was performed online prior to each injection using the auto-injector of an integrated Agilent 1100 HPLC. Separation was by a reverse-phase column packed with a C18 stationary phase (Hypersil BDS, 5 mm) using a linear gradient of aqueous sodium acetate, methanol, and acetonitrile. Detection was by fluorescence.

We focused on eight amino acids: aspartic acid (Asp), glutamic acid (Glu), serine (Ser), alanine (Ala), valine (Val), phenylalanine (Phe), alloisoleucine/isoleucine (A/I), and leucine (Leu) that are abundant in molluscan-shell protein, and resolvable by reverse-phase HPLC. The Asp and Glu values may include a small component of asparagine and glutamine, respectively, which were converted to Asp and Glu during laboratory hydrolysis. d/l values for all eight amino acids are reported in the Electronic Supplementary Material (ESM 1).

Radiocarbon dating

From the 80 Abranda analyzed for AAR, eleven shells spanning the range of observed d/l values were selected for AMS 14C dating using the suite of criteria outlined in Kosnik and Kaufman (2008) and Kosnik et al. (2008). Prior to AMS analysis, shell fragments were physically cleaned and ~30 % of the remaining mass was removed with 0.125 M HCl. Pretreated samples were hydrolyzed to CO2 using 85 % H3PO4, then converted to graphite using the Fe/H2 method (Hua et al. 2001). AMS 14C measurements were taken at ANSTO using the STAR facility (Fink et al. 2004). Radiocarbon ages were converted to calendar years using OxCal 4.1 program (Bronk Ramsey 2009), Marine09 data (Reimer et al. 2009), and a regional marine reservoir correction (ΔR) value of 4 ± 40 yrs based on the mean 14C offset between the Marine09 curve and the known-age coral core from Abraham Reef (Druffel and Griffin 2004) for the period 1635–1950 AD. The median of the age probability function was used as the calendar age, and the 2σ age range was used as the age uncertainty. All amino acid inferred ages discussed herein are calendar years relative to 2012 AD, the year of core collection (2012 AD ≡ 0 yr, Table 2). The results are the same for any year zero (i.e., 1950 AD), but the interpretation is clearer when the shell ages are set relative to the time that they were collected. Radiocarbon ages ranged from 28 to 617 yrs before collection. The radiocarbon data are summarized in Table 2.

Table 2 Absolute age data used to fit age-d/l models. The specimen number is the specimens’ unique identification number, UAL ID is the unique identification number for AAR analyses at Northern Arizona University and links these data to the d/l data in ESM 1, and ANSTO ID is the unique identifier for the radiocarbon analyses at the Australian Nuclear Science and Technology Organisation
Full size table

Abranda age determination

A single best age estimate was determined for each Abranda valve using AAR calibration models that were created using the Bayesian model fitting procedures outlined in and analytical scripts provided by Allen et al. (2013). In all, 72 different age models were fit using R 3.02 (R Development Core Team 2013). The lognormal distribution was the preferred uncertainty distribution because using the gamma distribution to model uncertainty led to systematic overestimates of the ages of specimens less than ~32 yrs (ESM 2). While six amino acid d/l values (Asp, Glu, Ala, Val, Phe, Leu) were fit, only Asp and Glu d/l contributed to the averaged gamma uncertainty model and only Asp d/l contributed to the averaged lognormal uncertainty model (ESM 3). Three mathematical functions were fit [time-dependent rate kinetics (TDK), constrained power-law kinetics (CPK), simple power-law kinetics (SPK)] with and without fitting a non-zero initial d/l value (R 0). Only TDK and SPK contributed to the final averaged model (Table 3). Time-dependent rate kinetics with a fit initial d/l value contributed 83 % of the averaged lognormal age model (Table 3). The Bayesian posterior distributions for the parameters of each age model are plotted in ESM 4. The shell ages used for this study are derived from Bayesian model averaging of models fit using the lognormal distribution to estimate uncertainty and only aspartic acid contributed to the final age model (Table 3).

Table 3 Bayesian model averaging summary by uncertainty distribution
Full size table

Time-averaging

Time-averaging, like any estimate of variation, is sensitive to sample size. Despite sampling sediment volumes 50 times larger than a typical hand core (each 5 cm excavation layer sampled 0.0125 vs. the 0.00025 m3 sampled by an equivalent 5 cm layer from the hand core), the number of Abranda dated was fundamentally limited by the number of right valves sampled. As each 5 cm excavation layer is represented by ~10 Abranda valves, an estimate of time-averaging based on the two standard deviation age range (2σ or 95 % of the data as defined by 0.977–0.023 quantile range) is essentially the full range of ages recorded in the sample and thus the difference between the two most extreme ages. The one standard deviation (1σ) age range equivalent to 68 % of the data (0.841–0.159 quantile range) is essentially the average age of the 2nd and 3rd oldest shells minus the average age of the 2nd and 3rd youngest shells. Consequently, we use the quantiles associated with the 1σ age range to characterize the extent of time-averaging to minimize our reliance on the most extreme age estimates in each layer, but both the 1σ and 2σ time-averaging estimates are provided in Table 4. All quantiles were calculated using the quantile() function in R 3.02, using the default method of linear interpolation between continuous points (type 7).

Table 4 Summary of excavation chronology by sampled layers
Full size table

Results

Sediment description

The grain size distribution was consistent throughout the entire 1.6 m length of the hand core (Fig. 3). The median grain size was 85 μm, the 10th quantile was 30 μm, and the 90th quantile was 456 μm. Sieved fractions from the excavated layers were also consistent down core with 31 % by dry weight 1–2 mm grains, 44 % 2–4 mm grains, and 25 % >4 mm grains (Fig. 4). The homogeneity of this deposit is consistent with thorough mixing of the sediment, homogenous sediment supply, or minimally, mixing sufficient to homogenize any variation in sediment supply. Of note, the remains of a live-collected callianassid shrimp were found in the 1.05–1.10 m excavation layer indicating the presence of bioturbators capable of moving sediment through the entire excavated depth.

Fig. 3
figure 3

Grain size analyses for the 80 mm diameter hand core every 5 cm down core. The y-axis indicates compacted core depth in meters. The x-axis indicates grain diameter on a log scale. Black boxes on the left side show the corresponding excavation layers from which shells were sampled for radiocarbon analyses, and white boxes on the left side indicate layers from which 210Pb activity was measured. Vertical dashed lines indicate the 10th, 50th, and 90th grain size quantiles; the gray boxes indicate the 95 % confidence intervals on those quantiles; black dots represent individual measurements; and large black circles indicate median values of the replicate measurements

Full size image
Fig. 4
figure 4

Proportion of layer fraction dry weights from the 0.5 m × 0.5 m excavation retained by the >4 mm sieve (dark gray), >2 mm sieve (medium gray), and >1 mm sieve (light gray). Black boxes on the left side show the corresponding excavation layers from which shells were sampled for radiocarbon analyses, and white boxes on the left side indicate layers from which 210Pb activity was measured. Vertical thick dashed lines indicate the median cumulative proportion in each fraction, and the thin dotted lines indicate the 95 % confidence intervals on those proportions

Full size image

Sediment (210Pb) age

Lead activity was measured from 11 of the same 1 cm thick layers used for the grain size analyses (Table 1). Lead analyses show declining 210Pbu activity below 0.15 m and reaching background levels by 0.50 m (Fig. 5). 210Pbu activities between 0 and 0.15 m do not decline, which suggests that the top 0.15 m of sediment is actively mixed on the decadal scale of 210Pb decay; on the other hand, below this mixed layer, fine sediments preserve a chronologically ordered historical record which indicates that this sediment is not mixed on decadal scales.

Fig. 5
figure 5

Lead activity down core. The x-axis is lead activity on a log scale, and the y-axis is core depth in meters. Lead activity begins to decline from 0.15 m, and by 0.5 m, background levels are reached. The 210Pb data are summarized in Table 1

Full size image

The CIC age model assumes a constant initial 210Pb concentration and a single net sediment accumulation rate. Based on the six analyzed layers between 0.15 and 0.41 m, the CIC age model estimated a sediment mass accumulation rate of 0.544 g cm−2 yr−1 (r 2 = 0.941). The CRS model assumes a constant rate of 210Pb supply but allows the total sedimentation rate to vary. The CRS model estimates the sediment mass accumulation rate between each analyzed layer, and these values decreased down core ranging from 1.4 g cm−2 yr−1 in the 0–0.01 m layer to 0.2 g cm−2 yr−1 in the 0.30–0.31 and 0.35–0.36 m layers.

The two age models (CIC and CRS) yield largely consistent age estimates. Relative to the CRS model, the CIC model infers older ages near the top of the core and younger ages at depth. So, while all of the models converge on ages of ~65 yr for the 0.25–0.26 m layer, ~85 yr for the 0.30–0.31 m layer, and ~100 yr for the 0.35–0.36 m layer, there is greater disagreement at shallower layers (Table 1). The CRS age estimate for the 0.40–0.41 m layer is very uncertain due to the very low 210Pb activity levels. Based on the 210Pb data alone, there is little reason to favor one age model. One significant complication in reef sediment is carbonate chemistry and the various taphonomic processes that can result in carbonate loss and thus lead to post-depositional increases in lead concentration.

Shell (14C/AAR) ages

The amino acid data had no notable outlier, and the correlation between Asp and Glu concentrations was highly significant with r 2 > 0.98. For the 11 jointly dated specimens, Asp and Glu d/l were highly correlated with calibrated carbon ages (r 2 = 0.91 and r 2 = 0.93, respectively). In fitting AAR calibration model, there was a noticeable bias associated with the choice of uncertainty distribution. Specifically, the calibration models fit using the gamma distribution to estimate uncertainty resulted in no shell with an inferred age <6 yrs, whereas using the lognormal distribution to estimate uncertainty resulted in nine shells with ages <6 yrs (ESM 3). The models fit using the lognormal distribution had much better fits (i.e., lower BIC scores) than those fit using the gamma distribution (Table 3). Nearly all of the Bayesian posterior distributions in ESM 4 show strong correlations between model parameters as seen by Allen et al. (2013). Some of the Bayesian posterior distributions show multiple modes indicating imperfect model fits (ESM 4), but the averaged age model contains only TDK and SPK functions indicating that these two models outperform CPK. The very low maximum Asp d/l and inferred age (0.25 and 625 yrs, respectively) make it unsurprising that CPK does not contribute to the final model. The very young age of these specimens and the imperfections in model fit results, both contribute to a relatively high median coefficient of variation of 1.7.

Inferred shell ages range from 1 to 625 yrs, and shell age increases with burial depth (Fig. 6). While the top ~0.15 m yields an essentially modern shell assemblage, every dated layer has an older median age than dated layers above although there is considerable overlap in the ages of individual shells (Fig. 6; Table 4). Individual shell ages and median shell age are both highly correlated with layer depth (Pearson’s r 2 = 0.92 and r 2 = 0.99, respectively). These data indicate a sedimentary shell assemblage spanning the last 600 yrs that is preserved in stratigraphic order.

Fig. 6
figure 6

Inferred shell ages with sediment depth. The x-axis is years before the year of collection (2012 AD), the y-axis is excavation depth in meters, black dots indicate the age of each AAR inferred shell age, white dots indicate the age of each 14C dated shell, horizontal lines encompass 95 % of the data, the horizontal bars encompass 68 % of the data, and the vertical bar is the median layer age. The dashed vertical line is the year of collection (2012 AD). The 14C/AAR data are summarized in Table 4

Full size image

Discussion

Agreement between 210Pb and 14C/AAR chronologies

Despite different assumptions and sampled grain sizes, the 210Pb and 14C/AAR chronologies are remarkably consistent over the range of ages that can be directly compared (Fig. 7). While both 210Pb age models are in general agreement with the 14C/AAR chronology, the CIC age model yields a slope shallower than the unity line shown in Fig. 7. The CRS age model more closely matches the 14C/AAR chronology. In comparing the uncertainties plotted in Fig. 7, it is worth noting that the 210Pb error plotted is a 1σ analytical error, and that the dated layer was 1 cm thick, whereas the 14C/AAR time-averaging is the 1σ age range based on ~9 shells per 5 cm thick layer, thus including a greater sediment layer, time-averaging, and modelling uncertainties in addition to analytical error. Despite these sources of variation, the agreement between the 210Pb and 14C/AAR chronologies is exceptionally good. This suggests consistent sedimentary processes spanning the full range of sedimentary particle sizes.

Fig. 7
figure 7

Agreement between the 210Pb chronology on the y-axis and the 14C/AAR chronology on the x-axis. The dashed 1:1 line represents perfect agreement between the two age estimates. 210Pb ages are single analyses with 1σ analytical error from a 1 cm layer, whereas 14C/AAR ages are median shell age (large circles) and individual shell ages (small dots) from the 5 cm excavation layer with confidence intervals encompassing 68 % of the data based on the sample of dated shells from the layer (see Table 4). For sediment layers with 210Pb ages but with no 14C/AAR ages (e.g., layers 0.05, 0.15, 0.25, 0.35 m), the position on the x-axis is mean age of the adjacent dated shell layers. The CIC 210Pb age model plotted in gray estimates a shallower slope than the 1:1 line suggesting younger ages at depth than the CRS or 14C/AAR ages. The CRS 210Pb age model has very good agreement with 14C/AAR age estimates. As expected, neither 210Pb age model works well in the top 0.15 m of sediment. The 210Pb data are summarized in Table 1 and the 14C/AAR data are summarized in Table 4

Full size image

Time-averaging and stratigraphic resolution

These data indicate a young shell assemblage in stratigraphic order. This is in stark contrast to the shell assemblages spanning similar burial depths from Rib and Bramble reefs that yielded age homogenous samples spanning the last ~3000 yr (Kosnik et al. 2007, 2009, 2013). Consistent with previous work, both 210Pb and 14C/AAR chronologies indicate that the top 0.15–0.2 m of sediment is actively mixed. The 0.20–0.25 m excavation layer yields a median shell age of 55 yrs (Table 4), and the 210Pb analyses of the 0.20–0.21 m hand core layer yield an age of ~50 yrs (Table 1). Both of these lines of evidence could suggest that this sedimentary record might be capable of preserving fossil assemblages with ~50 yrs time-averaging.

A preferable approach to quantifying time-averaging is to determine the age of each shell relative to the layer’s median shell age. By normalizing the data in this way, it is possible to examine changes in time-averaging down core independent of layer age (Fig. 8). In the samples from the top 0.15 m (combined 0–0.05 and 0.1–0.15 m layers), the 1σ age range is 19 yrs and the 2σ range is 125 yrs, whereas in samples below the top 0.2 m the median 1σ age range is 111 yrs and the 2σ age range is 218 yrs (Fig. 8). Despite a single exceptionally old specimen (oldest specimen was 186 yrs, whereas all other specimens were younger than 24 yrs) in the top layer, it is readily apparent that the degree of time-averaging increases dramatically below the top 0.2 m (Fig. 8). Time-averaging in surficial sediment is constrained by the impossibility of averaging future shells into the shallowest samples, but it is notable that above 0.6 m, relative age distributions appear to include more older shells than younger shells, whereas below 0.6 m, the relative age distributions appear more normally distributed (Fig. 8). These data support the general consensus that the age distribution of samples from surface collections are skewed (e.g., Kidwell 2013), and observations made in previous studies that the top 0.2 m of surficial sediment typically sampled by grab samplers do not reflect the properties of deeper sediments (Kosnik et al. 2007, 2009, 2013). Estimates of sedimentation or time-averaging derived from the surficial layer are indicative of internal depositional dynamics and not the aggrading sedimentary column. In these samples, the extent of time-averaging in shell assemblages buried below 0.2 m was 2–6× more than in surficial shell assemblages.

Fig. 8
figure 8

Relative shell age with sediment depth. The x-axis is shell age in years relative to the layer’s median age, the y-axis is excavation depth in meters, individual black dots represent the relative age of each shell, horizontal lines encompassing 95 % of the data, the horizontal bars encompass 68 % of the data, and the vertical bar is the median layer age (see Table 4). The light gray vertical box approximates the time accounted for by the 5 cm collection interval

Full size image

In this study and in previous studies by the same authors, subsurface samples shallower than 20 cm sample the same time interval as surface samples. In the context of detecting recent anthropogenic change, it is important to consider the dilution effect of this surficial layer and the related mixing. Time-averaging on the order of 19 to 125 yrs (1σ and 2σ, respectively) implies that a recent mass die-off will be mixed with the remains from normal annual mortality of ~19 yrs. While time-averaging and bioturbation need to be quantified on a site specific basis, it is illustrative to examine the classic study of Diadema mass mortality (Greenstein 1989) in the context of what we now know about time-averaging in coral reef-associated sediment. Assuming a stable population size and an average life span of 4 yrs (Ogden and Carpenter 1987) implies a population turnover of 25 % per annum. Even a 100 % mass mortality event would yield only a mean mortality over the 1σ period of time-averaging of 29 %, or a 15.8 % more than background (18 yrs of 25 % and 1 yr of 100 %), and 3 yrs after the mass mortality event, the net increase is zero. That is a generous estimate as using the 2σ estimate of time-averaging yields an enrichment of only 2.4 %. These are, of course, only rough estimates as a precise estimate requires knowing the extent of time-averaging in the studied sediment, background mortality rates, and ideally, estimates of the rate of taphonomic loss of Diadema fragments within the sediment. Additionally, these calculations are based on the surface sediment, whereas deeper sediment layers have longer periods of time-averaging. In this case, however, the most likely effect of a mass mortality event such as the 1983 die-off may well be the lack of contribution in years after the mass mortality.

Sedimentation/shell burial rate

Both 210Pb age models directly estimate sedimentation rate as mass flux that can be converted to depth using the measured sediment density (Table 1). While the CIC model estimates a constant sedimentation accumulation rate of 3.9 mm yr−1 [(0.544 g cm−2 yr−1 / 1.4 g cm−3) × 10 mm cm−1], the CRS model allows the mass accumulation rate to vary. The CRS model sediment accumulation rates decline down core, especially through the top 0.15 m. The CRS model estimates a sedimentation rate of 9.0 mm yr−1 for the surface (0.00–0.01 m) layer, declining to a rate of ~1.6 mm yr−1 in the two layers below 0.3 m. Breaking the CRS sedimentation rates into the upper mixed layer and the deeper layers showing clear decline in 210Pb activity (see Fig. 5) yields average sedimentation rates of 6.0 and 2.2 mm yr−1 for 0–0.16 m and 0.2–0.36 m, respectively.

The shell ages can be used to calculate effective shell burial rate in several different ways. A simple linear regression of age and burial depth yields a highly significant relation (r 2 = 0.84) and a burial rate of 1.6 mm yr−1. This is analogous to the single sedimentation rate provided by the 210Pb CIC model, although the shell burial rate is half the 210Pb CIC sedimentation rate. The intercept of this regression is also significant and suggests that accumulation starts at 0.145 m. Fitting the shells collected from layers deeper than 0.2 m separately yields a burial rate of 1.4 mm yr−1 (r 2 = 0.81), a value half of that estimated by the CIC model.

Alternatively, one can examine the difference in shell age between sequential layers. This estimate uses the difference between the two layers as the depth of sediment and all pairwise differences in shell ages between adjacent layers. This results in a distribution of burial rates for each depth (Fig. 9; Table 4). These results are most comparable to the CRS sedimentation rates and indicate a clear change shell burial rate between 0.15 and 0.2 m depth. Interpretation of the shell burial rates in the top 0.15 m is complicated by the certainty that these shells are still being actively mixed, and the likelihood that the total sediment volume is being still being reduced by taphonomic processes. In the top 0.15 m, the shell burial rate averages 5.6 mm yr−1. Deeper than 0.2 m, the shell burial rate ranges between 0.7 and 1.9 mm yr−1 with a mean burial rate of 1.2 mm yr−1 (Table 4). In comparison, CRS sedimentation rates are 6.0 and 2.2 mm yr−1 for the same depth ranges. Given the uncertainties and biases of the two different analyses, these two values are remarkably similar.

Fig. 9
figure 9

Effective shell burial rate. The x-axis is effective burial rate between sequential layers in mm yr−1, the y-axis is excavation depth in meters, black circles represent the pairwise comparisons between each shell in the layer to each shell in the layer above, horizontal lines encompass 95 % of the data, and the horizontal bars encompass 68 % of the data (see Table 4). The vertical dashed line is an effective sedimentation rate of 0 mm yr−1. All layers have a positive effective sedimentation rate (see Table 4)

Full size image

Both datasets indicate higher sedimentation/burial rates in the surficial layer, but the magnitude of this difference is a factor of ~2 (lead sedimentation rates are 2.7× higher, whereas shell burial rates are 4.6× higher). It is likely that all the sedimentation rates from the top layer are inflated relative to deeper layers due to surface sediments being less compacted and/or taphonomically less altered. Relative to research at other locations on the GBR, the OTR lagoon has little evidence of size-selective sedimentary processes at work.

In the early 1980s, Davies Reef lagoon was the subject of numerous sedimentological and bioturbation studies (e.g., Tudhope and Scoffin 1984; Tudhope and Risk 1985; Tudhope 1989). Based on four dates, Tudhope (1989) estimated the sedimentation rate to be 3.4 mm yr−1 over the past 640 yrs, 1.4 mm yr−1 between 640 and 2,400 yrs. The estimate of the last 640 yrs is based on a Callista valve aged 642 yrs sampled at 2.5 m (including a −0.3 m depth correction for estimated living depth = 3.4 mm yr−1 over the past 640 yrs), and bulk carbon ages of sediment collected from 1 m minus the bulk age of surface sediment and accounting for the surface mixed layer (“a few tens of centimeters/100 years” p. 1044). The estimate for the sedimentation rate between 640 and 2,400 yrs is based on a bulk carbon age collected from 4.75 m aged 3,672 yrs (1.3 mm yr−1), but subtracting the estimated time required to transport the sediment to the site (1,200–1,400 yrs) yields estimates of 1.9–2.1 mm yr−1. Despite the limited number of dates, these sedimentation rates are similar to those calculated here.

Dates determined during geochronologic and taphonomic studies of Rib and Bramble Reef can also be used to estimate sedimentation rates. The excavation sites at Rib and Bramble reefs are sedimentary environments very similar to Davies Reef. In the Rib Reef lagoon, the median age of Pinguitellina robusta sampled between 0.3 and 1.1 m was 636 yrs = 1.7 mm yr−1 (Kosnik et al. 2013). In the Bramble Reef lagoon, the median age of Pinguitellina sampled in the top 1.25 m was 516 yrs = 2.4 mm yr−1 (Kosnik et al. 2013). As the Rib and Bramble Reef cores are also largely time homogenous, it would be possible to decrease these rates by ~3× by sampling similarly aged shells at ~0.3 m. Additional taxa with different inherent durability were dated from the same Rib and Bramble Reef cores (Kosnik et al. 2009); sedimentation rates calculated for each taxon differ even for the same core layer. Despite that, the shell burial rates are similar to that seen in the OTR lagoon data.

One explanation for a decline in shell burial rate is increased sediment compaction associated with decreased sediment water content with depth. This would be consistent with the increasing difficulty of collecting deeper excavation layers in the field, but not consistent with the measured sediment bulk densities (Table 1). The top 0.2 m was easily penetrated and quickly collected, but sediment deeper than ~0.6 m was notably more difficult to penetrate and required more time to collect. If dewatering was the sole reason for a decline in burial rate with depth, then the amount of time recorded per 5 cm layer should also increase with depth. However, there is no correlation between layer median burial rate and the degree of time-averaging in layers below 0.2 m [Pearson’s p ≥ 0.69 (0.83 for 1σ age range, 0.69 for 2σ age range)]. While we have insufficient data to definitively test the dewatering hypothesis, we find no support for it.

The surface layer is actively mixed by a variety of physical and biological processes. The exceptionally fine grain size distribution suggests that reworking by wave energy is quite rare, but collections from the surface sediment yielded large numbers of living mollusks, arthropods, and annelids suggesting significant bioturbation. Size-specific sorting has been documented in reef sediments (e.g., Tudhope and Scoffin 1984), and this is anecdotally supported by a slightly smaller >4 mm fraction in the surface layers than in the rest of the excavation (Fig. 4) in conjunction with the observation that much of the larger material in the surface 0.1 m was alive at the time of collection (e.g., 56 % of Abranda sampled from the 0 to 0.05 m layer were alive). However, bioturbation is insufficient to obscure stratigraphic order in this sediment. The most promising hypothesis to explain the observed decrease in sedimentation rate is the loss of sediment through taphonomic processes.

Sediment taphonomy

Carbonate sediment is chemically active and an important part of the global carbon cycle. While the dissolution of carbonate grains under natural conditions is incredibly complex, laboratory experiments have shown that grain microstructure and size can each influence rates of reaction by a factor of 7 at the same saturation state (Walter and Morse 1984). Very high rates of fine sediment loss can be used to explain fine surface layers over coarse deeper sediments such as those documented by Tudhope and Scoffin (1984) and Kosnik et al. (2007, 2009, 2013), but the One Tree lagoon data presented here lack significant down core changes in the complete grain size distributions (Figs. 3, 4). This means that changes in sedimentation rate cannot be driven by preferential loss of a particular size fraction, at least any preferential loss of grains within the dominant <1 mm size fraction.

Another potential explanation for the decline in burial/sedimentation rates with depth is continued taphonomic loss of sediment at depth. If we assume that the two deepest excavation layers are below the taphonomically active zone, then the final shell burial rate is 1.2 mm yr−1. Assuming 1.2 mm yr−1 is the net rate, sediment from the top 0.2 m would have a 21 % chance of survival, sediment from 0.2–0.25 m and 0.3–0.35 m layers would have 93 % chance of survival, and sediment from 0.4–0.45 m and 0.6–0.65 m layers would have 95 % chance of survival. Given the uncertainties in these estimates, it is reasonable to suggest that layers deeper than 0.2 m are within depth of final burial, but that the probability of sediment above 0.2 m, i.e., the taphonomically active zone, being buried below 0.2 m, i.e., the depth of final burial, is ~21 % (sensu Olszewski 2004). This explanation does not preclude changes in time-averaging with depth, especially if per grain sediment preservation rates vary. The time-averaging data are consistent with the condensation implied by a 15 % survival rate of non-mollusk shell sediment (although these data are not independent from the burial depth data above). The 1σ time-averaging below 0.2 m is 111 versus 19 yrs above 0.2 m, indicating deeper deposits can be derived by these surface sediments with 16 % survival of non-mollusk remains.

Another indication of surficial sediment loss is that the surface sample (0.0–0.01 m) had lower 210Pb activity than the other three mixed layers. This suggests that the key assumption of the CIC model is not true in this sediment. The observed increase in 210Pb activity between the surface layer and the 0.05–0.06 m layer is consistent with a 16 % loss of carbonate from the surface layer. A 16 % loss is consistent with the 18–30 % loss estimated in experimental studies of microborers in the Davies Reef lagoon (Tudhope and Risk 1985); as this 16 % estimate is conservative in that, it assumes no taphonomic loss from the 0.05 to 0.06 m layer.

The lead sediment accumulation rate data can also be used to estimate the loss of sediment through digenetic processes. Assuming that the minimum sediment accumulation rate represents the final net sedimentation rate, it is possible to calculate the amount of sediment that must be lost from each layer to reach the minimum sedimentation rate. This approach estimates sediment survival in the top layer (0–0.01 m) at 16 %, and preservation rate has a strong linear relation (r 2 = 0.92) with depth, increasing at 0.25 % per mm core depth. This estimate of sediment preservation probability is consistent and entirely independent of the 21–16 % estimate derived from the shell data.

Implications

While the top 0.2 m is actively mixed, the top meter of lagoonal sediment preserves a stratigraphically ordered deposit spanning the last 600 yrs. Despite different assumptions, 210Pb and 14C/AAR chronologies are remarkably similar indicating consistency in sedimentary processes across sediment grain sizes spanning more than three orders of magnitude (0.1–10 mm). Both 210Pb and 14C/AAR data suggest that only ~16 % surface sediment in the OTR lagoon will be buried, whereas ~84 % of the surface sediment with be recycled through the system. 210Pb ages inferred using the CRS model show the highest level of agreement with 14C/AAR age estimates. Molluscan time-averaging in this sediment is ~110 yrs per 5 cm layer. While not a high-resolution record, at least some reef-associated sediment is suitable for paleoecological studies spanning the period of Western colonization and development. This sedimentary deposit, and others like it, should be useful, albeit not ideal, for quantifying anthropogenic impacts on coral reef systems.