Introduction

Massive reef-forming (hermatypic) corals produce density bandings during skeletal growth, which are valuable elements for sclerochronological studies, such as the “coral ring chronology” of Knutson et al. (1972). Therefore, coral sclerochronology can be a tool to analyse environmental change archived in the carbonate skeletons of corals and reefs (e.g. Hudson et al. 1976). Knutson et al. (1972) identified the origin of skeletal density bandings in scleractinian corals as being of annual origin, based on X-ray imaging calibrated with radioactive strontium isotopes (90Sr) derived from nuclear tests on Eniwetok Atoll, Marshall Islands. Buddemeier et al. (1974) confirmed the results of Knutson et al. (1972) in that observed skeletal density bandings reflect seasonal and annual growth patterns in general.

Since then, the formation of annual high- and low-density bandings (HDBs/LDBs) in coral skeletons has been likened to a variety of environmental factors, including sea surface temperatures (e.g. higher SSTs during warmer summer months), light availability (photosynthetic activity of zooxanthellate symbionts), reproduction (size and number of gonads) or cloud cover and precipitation (e.g. Knutson et al. 1972; Buddemeier 1974; Buddemeier et al. 1974; Weber et al. 1975; Hudson et al. 1976; Highsmith 1979; Lough and Barnes 1990a, b; Mendes 2004; Wórum et al. 2007). By and large, it seems most likely that a combination of these controlling parameters (and others) can be seen as the origin for the formation of density bandings in coral skeletons (e.g. Helmle et al. 2000; Helmle and Dodge 2011 and references therein).

However, the formation of density bandings in scleractinian corals is until now not fully understood (e.g. Lough and Barnes 1990a, 1992; Helmle and Dodge 2011). In general, fan-shaped aragonitic needles (sclerodermites) form the base for accretion of carbonate skeletons in scleractinian corals (Barnes 1970). Based on these, the formation of skeletons is developed as a combination of the following intra-skeletal elements: septa, dissepiments, basal plates and the epitheca (Barnes 1970, 1972). Buddemeier et al. (1974) mentioned two possible origins for HDB/LDB alternations during skeletal growth. These represent, first, variations in the packaging of aragonite needles and, second, variations in spacing of intra-skeletal elements (septa and trabeculae). Barnes and Devereux (1988) distinguished the accretion of coral skeletons into “micro-architecture” (crystallographic arrangement of aragonitic needles within intra-skeletal elements), “meso-architecture” (arrangement of intra-skeletal elements and their formation of corallites) and “macro-architecture” (arrangement of corallites within the coral colony). In case of the Indo-Pacific reef coral Porites, Barnes and Devereux (1988) assumed meso-architecture as the main factor for HDB/LDB formation. Le Tissier et al. (1994) evaluated the use of X-radiography for the visualisation of variations in density bandings of Porites corals and assumed that thickening, thinning and/or spacing of intra-skeletal elements is the source for HDB/LDB formation. Dodge et al. (1992) and Dávalos-Dehullu et al. (2008) studied colonies of massive Orbicella corals from the Caribbean and found evidence for HDB/LDB alternation based on size variations of exothecal skeletal elements (horizontal dissepiments and vertical costae located in the skeleton between the corallites). Helmle et al. (2000) studied colonies of the meandroid brain coral Pseudodiploria and showed that HDB/LDB formation occurs due to variations in thickness of septal and columellar elements. Based on such studies, variations in meso-architecture of coral skeletons are identified as the main driver for the formation of density bandings; however, taxon-specific differences must be considered.

Various techniques exist to quantify skeletal density of coral skeletons following the Archimedes principle on estimations of intra-skeletal pore volume or buoyant weight by weight and volume ratios. These include mercury displacement (Dustan 1975), water displacement (e.g. Hughes 1987) and freezing techniques (Carricart-Ganivet et al. 2000). However, all these applications are representing invasive techniques and therefore lead to partial destruction of coral samples. Furthermore, the resolution of those density values is limited to (annual) bulk data. These downsides were also pointed out by Duprey et al. (2012) and Deveaux et al. (2017). Other than these approaches, non-invasive analytical techniques exist, including (1) optical densitometry by using X-ray images, (2) medical X-ray computerised tomography (CT) and (3) gamma densitometry (radiation beam technique).

The quantification of skeletal density of coral samples with different approaches of optical densitometry using X-radiographs (original films or digitised images) has widely been used in various studies (e.g. Dodge and Thomson 1974; Buddemeier 1974; Buddemeier et al. 1974; Dodge and Brass 1984; Chalker et al. 1985; Helmle et al. 2000; Carricart-Ganivet and Barnes 2007; Duprey et al. 2012; Rico-Esenaro et al. 2019). For these, coral samples are scanned with an X-ray beam and the generated radiographs calibrated with aragonite and aluminium standards for optically estimated density values. However, one must consider radiation effects during X-ray scanning, which have an immediate influence on the calibration of skeletal densities in coral samples, including the heel effect and the inverse square law (Chalker et al. 1985). To counteract such issues of irradiation patterns, different approaches have been developed, including corrections for one-dimensional tracks along the X-ray image (Carricart-Ganivet and Barnes 2007) and two-dimensional corrections for the whole X-ray image (digital detrending technique of Duprey et al. 2012).

The application of medical X-ray computerised tomography (CT) for quantifying skeletal density in corals was first used by Bosscher (1993). A main advantage of CT scanning of corals is the ability of three-dimensional visualisation of intra-skeletal elements and the opportunity of selecting the ideal orientation of the skeletal growth direction for density measurements (e.g. Bosscher 1993; Lough and Cantin 2014). Analysing coral skeletons with CT techniques has been used in various studies for visualisation of density bandings, estimations of skeletal density, analyses on skeletal architecture or as a comparative method (e.g. Helmle et al. 2000; Duprey et al. 2012; DeCarlo and Cohen 2017; Benson et al. 2019). However, this method is not easily applicable due to the need of special equipment and handling issues (Carricart-Ganivet and Barnes 2007; Duprey et al. 2012). Ivankina et al. (2020) have used neutron-computed tomography (NCT) for the investigation of coral skeletons. The application of NCT can penetrate coral samples more intensely in relation to classical X-ray CT and, therefore, enables the visualisation of intra-skeletal structures, which were previously not recognised (Ivankina et al. 2020).

Barnes and Devereux (1988) reported first results on skeletal density of Porites corals by using gamma densitometry, which shortly after was described in detail by Chalker and Barnes (1990). Barnes and Devereux (1988) used 210Pb and 241Am sources for one-dimensional analyses of density bandings and compared their results to those of invasive techniques of weight and volume analyses. The results of Barnes and Devereux (1988) have shown that solely the approach of gamma densitometry was able to identify (intra-)annual density bandings. A limitation in gamma densitometry is the fact that the orientation of skeletal growth direction cannot be adapted for one-dimensional track selection (Lough and Cantin 2014). A first attempt to quantify skeletal density with a two-dimensional grid-scanning approach using gamma densitometry was published by Deveaux et al. (2017), yielding preliminary results with annual bulk densities for Porites corals from the Maldives and Orbicella corals from Belize. The modified measurement protocol by Deveaux et al. (2017) for two-dimensional grid-scanning has overcome this limitation and allows free selection of areas of calibrated density grid points for generating annual bulk density datasets. However, for high-resolution density measurements at intra-annual resolution, it is necessary to use coral samples, in which the orientation of density banding is perpendicular to the scanning direction.

The aim of this study is to systematically test the application of two-dimensional grid-scanning gamma densitometry after Deveaux et al. (2017) to quantify skeletal density in coral skeletons at higher temporal (intra-annual) resolution and to evaluate this methodical approach for palaeoecological and palaeoclimatological studies on coral skeletons. For this purpose, a colony of the massive scleractinian coral Orbicella faveolata (Ellis and Solander 1786) is used from the nearly 300-km-long Belize Barrier Reef (BBR), which runs parallel along the coast of Belize, Central America (see, e.g. Purdy and Gischler 2003). The selected coral sample was previously used by Deveaux et al. (2017) for establishing the modified measurement protocol and to develop first annual bulk skeletal density data. However, with the intention of this study, the same colony is herein re-investigated. Time series analysis and statistical investigations with skeletal growth patterns (linear extension, calcification) and geochemical (δ13C, δ18O) datasets are performed to identify likely connections of biological and environmental controls on the skeletal density formation in O. faveolata from within the central BBR.

Material and methods

Coral samples

The analysed coral core BZE-2 from O. faveolata was sampled on the barrier reef platform within the central area of the BBR (16°54′14″N, 88°06′47″W) on 26 August 1999 (Gischler and Oschmann 2005). The core was taken from a large, isolated coral colony on the barrier reef platform less than half a metre below sea level, by using a pneumatic drill with a core barrel of 4 cm diameter and a length of 30 cm (Gischler and Oschmann 2005). Core pieces were cut lengthwise with a rock saw. The individual core slabs are stored in the Institut für Geowissenschaften at the Goethe-Universität in Frankfurt am Main, Germany, and referred as BZE-2–1, BZE-2–2, BZE-2–3, BZE-2–4 and BZE-2–5.

Gamma densitometry: measurement procedure and calibration

In contrast to previous one-dimensional track-scanning procedures (e.g. Barnes and Devereux 1988; Chalker and Barnes 1990), the newly developed gamma densitometer of Deveaux et al. (2017) allows a two-dimensional grid-scanning approach. For grid-scanning a coral sample, a source of radioactive-decaying americium isotopes (241Am) is used, generating X-ray beams with a (mostly) monochromatic 59.5 keV gamma ray beam. The beam intensity is measured prior to each data collection and remains in practical terms constant during the measurement thanks to the fixed beam optics and the long half-time of 432 years of the source. Emitted X-ray beams can penetrate the coral sample, which is fixed on an electronically moving positioning table. The grid-scanning procedure follows a meandering pattern of grid rows perpendicular to the axis of coral core slabs, with a sample step size of 0.5 mm (Deveaux et al. 2017).

A collimator between coral sample and detection system helps to absorb unwanted wavelength spectra for a precise control of monochromatic photons, which have penetrated the coral skeleton. The detection system consists of a two-stepped radiation sensor. First, the energy-rich X-ray photons are absorbed by a bismuth germanium oxide scintillator, which converts them into several photons in the energy range of visible light. In the next step, a photomultiplier tube (PMT) detects these visible light photons and converts them individually into electrical pulses. The pulse length is short as compared to the elapsed time; the scintillator needs to emit the photons related to one impinging X-ray. In consequence, the energy of the impinging X-ray can be measured within some limits by counting the number of related pulses. With this approach, it is possible to count individual X-rays (Deveaux et al. 2017).

A limiting factor for the spatial resolution of grid points is the 1 mm radius of the collimator hole between the coral sample and the underlying scintillator. The standard deviation and, thus, the resolution of a uniform circular beam of radius r amount to σ = r/2. The spatial resolution is therefore given with 0.5 × 0.5 mm per grid point (Deveaux et al. 2017). Coral samples are grid-scanned on a width of 2 cm, therefore enabling up to 40 individually calibrated density measurements per each grid row.

For the calibration of density measurements, the thickness of each core slab was measured by using a micrometre screw, which has a precision of 10 µm. Double measurements of core slab thickness were taken on the left and right sides for each increment individually. For a precise position of each thickness measurement, available X-ray images of coral core slabs were used to identify areas of HDBs and LDBs. Based on individual measurements, an average thickness with associated error bars was calculated for each growth increment. The average core slab thickness of the investigated coral core amounts to 5.60 ± 0.06 mm. An overview of mean core slab thicknesses is given in Table 1. Variations in thickness of the samples occur due to limitations in the sawing precision.

Table 1 List on average datasets of core samples from BZE-2
Full size table

A graphical user interface (GUI) was programmed by Deveaux et al. (2017) to control settings for the described measurement procedure and to select individually calibrated grid points for quantifying skeletal density of coral samples. For the selection of areas in coral samples, contrast settings can be adjusted for an optimal identification of HDBs and LDBs in the generated scan. For a precise orientation, available X-ray images were used to correlate the observed banding patterns. Before selecting grid point rows, the measured average core slab thickness and uncertainties were added to the GUI, to calibrate individual density measurements within each annual growth increment.

The skeletal density ρ (g cm−3) of coral samples can be quantified by the following equation (Deveaux et al. 2017):

$$\rho=-\left(\frac1{k_0\ast t}\right)\ast\log\left(\frac I{I_0}\right)$$

The intensity I0 (Hz) resembles the number of photons per unit time, which enter the coral sample and I (Hz) the number of penetrating photons. A time frame (t) of 12 s (corresponding to \({I}_{0}=3.5\times {10}^{4}\) counts) per grid point is used to ensure a robust detected signal during grid-scanning. The attenuation coefficient k0 (cm−2 g−1) represents a material constant and is connected to the quantum energy of the X-ray beam. The effective attenuation coefficient was calibrated by Deveaux et al. (2017) with k0, eff = 0.362 ± 0.013 cm−2 g−1 by using an aragonite wedge of a Tridacna gigas bivalve shell as an independent standard with an estimated density of ρ = 2.85 ± 0.03 g cm−3 (for details on the calibration of k0, eff see therein). For the calibration of density measurements, the non-attenuated intensity I0 of the X-ray beam was measured for each coral sample within a time frame of 2 min, before the positioning table moved the coral core slab under the emitting 241Am source.

Each density measurement was taken by selecting individual grid rows parallel to the observed density bandings and perpendicular to the axis of the coral growth direction along the core slab. For this procedure, it is necessary that the orientation of density bandings is perpendicular to the axis of the core slab. Since a perfect geometrical 90° orientation of density bandings against to the core slab axis is impossible, minor deviations and, therefore, subordinate influences of neighbouring areas of skeletal density cannot be excluded completely. This is especially the case for the investigated coral O. faveolata, which has a bulbous skeletal surface with characteristic bumps (e.g. Weil and Knowlton 1994). Nevertheless, the studied colony shows a skeletal density banding, which fits remarkably well for the selection of individual grid rows. Therefore, significant deviations of observed intra-annual density fluctuations are reduced to a negligible proportion and still allow the identification of HDBs and LDBs (Fig. 1a, b). Solely the base of this colony in the core slab BZE-2–5 was not analysed, because banding patterns show a curved orientation to the core slab axis and therefore would limit a precise selection of density areas along the growth axis at the same spatial resolution.

Fig. 1
figure 1

a Detail of grid scan from core slab BZE-2–3 with plotted density values for each grid row along the growth axis. Summer HDBs and identified double-stress bandings (1982 and 1984 CE, red arrows) are figured. b Complete density dataset of the coral core BZE-2, plotted against to the individual scans of each core slab and the assumed growth time. Uncertainties for density values are figured with vertical error bars (grey). c Corrected density data expressed as the variance in amplitudes of mean skeletal density. Below, continuous wavelet transform (CWT) for BZE-2 is plotted with observed periodicities and artificial LACs (local abrupt changes after Hochman et al. 2019) within the cone of influence (boundary area). Statistical significances of periodicities are indicated by a thin black line (p ≤ 0.05). Intervals corresponding to individual core samples are figured with vertical dotted lines (red). d Age versus depth model for BZE-2 (black). A theoretical model (grey) assuming a perfect bi-weekly resolution per grid point (translating into 26 grid points per year) is figured for comparison. The regression line for BZE-2 is slightly steeper, implying an overall higher temporal resolution for individual grid points

Full size image

Age model for Orbicella faveolata

Since the pioneering work of Knutson et al. (1972), it is known that hermatypic corals produce their HDBs during warmer summer months and LDBs during the rest of the year. However, according to interspecies differences within the ecological niche of hermatypic corals, it is herein considered to use a taxon-specific age model for the relative timing of HDBs and LDBs in O. faveolata.

Hudson et al. (1976) proposed for Orbicella species of the former “Montastraea annularis complex” a formation of HDBs during warmer summer months (July to September). Similar suggestions have been made by other authors, e.g. Carricart-Ganivet et al. (2000) (HDB formation during July to September) and Mendes (2004) (HDB formation during late August to early October). Based on a taxonomic revision combining morphological and molecular data (Budd et al. 2012), the genus Orbicella Dana 1846 was established for the three species of the former “Montastraea annularis complex”. These include the three often sympatrically occurring sibling species Orbicella (ex Montastraea) annularis (Ellis and Solander 1786), Orbicella (ex Montastraea) faveolata (Ellis and Solander 1786) and Orbicella (ex Montastraea) franksi (Gregory 1895), which are separated by their skeletal morphology (e.g. phenotypic variations of bumps), some genetic differences (e.g. separation of O. faveolata by mitochondrial genomes) and contact-induced mortality (aggressive behaviour), as well as by their independent reproduction during spawning events (Knowlton et al. 1992, 1997; Weil and Knowlton 1994; Lopez et al. 1999; Fukami et al. 2004; Dawson 2006). Cruz-Piñón et al. (2003) analysed monthly extension rates of living colonies of O. faveolata from the Mexican Caribbean and from the Gulf of Mexico. The authors have shown that HDBs of O. faveolata form during summer months (July to September), whereas the LDBs form during the rest of the year, thereby confirming earlier observations on the Montastraea species complex.

The relative timing of HDB/LDB patterns is herein calibrated with available δ18O isotope data from the same coral colony (Gischler and Oschmann 2005) (Fig. 2a). The 18O/16O ratio in coral carbonate skeletons is inversely related to fluctuations in ambient sea water temperatures, with higher δ18O values relating to colder temperatures and lower δ18O values to warmer ones (e.g. Weber et al. 1975). However, other fractionation effects should be considered as well, which might lead to overprints in the geochemical signature (see discussion). For this calibration, the exact position of each borehole position of previous carbonate sampling for mass spectrometry was used to enable a precise correlation with the obtained density measurements. A total of 4 to 6 individual δ18O values is available per growth increment. This resolution allows the assignment of HDBs and LDBs to corresponding oxygen isotope ratios (Fig. 2a). However, a precise connection of other intra-annual bandings (e.g. “winter stress bandings” according to Hudson et al. (1976) and Dodge et al. (1992); double-stress bands according to Wórum et al. (2007)) is limited, due to the lower spatial resolution comparative to the individual grid row values.

Fig. 2
figure 2

Comparison of geochemical isotope datasets (a δ18O, green; c δ13C, light blue) from Gischler and Oschmann (2005), with intra-annually resolved skeletal densities (b this study, black). Linear regression lines added to each time series. Age model for HDB/LDB alternation is calibrated for all datasets with figured 5-year interval inscriptions. A growth interruption interval with anomalously higher density values at the base of the colony is followed by a time of growth recovery (black arrow) until regular LDB/HDB alternations sets in again

Full size image

Next to the formation of HDBs and LDBs, an overall age versus depth model (Fig. 1d) was generated to quantify and evaluate the overall temporal resolution of a single grid point measurement within the present coral colony on an intra-annual time scale. Hereby, the total number of taken grid point measurements per growth increment was plotted against the estimated growth time.

Time series analysis

For identification of growth periodicities, the software package PAST (Paleontological Statistics) version 4.06b of Hammer et al. (2001) is used to generate a continuous wavelet transform (CWT) after Torrence and Compo (1998) (Fig. 1c). The obtained density dataset represents equidistant values due to the defined spatial resolution of grid rows, therefore enabling an evenly spaced time series. Short gaps occur only between the individual core slabs, which were supported by an interpolation prior to the wavelet generation. Before the wavelet analysis, datasets were first calibrated as the variance of amplitudes against to mean density. To exclude bias, long-term trends in the data were extracted by using the LOESS smoothing function of PAST v4.06b (Fig. 1c). This approach is based on the LOWESS (locally weighted smoothing scatterplots) regression smoothing method after Cleveland (1979, 1981). Prior to the wavelet analysis, anomalous values from the top and the base of the investigated coral skeleton were excluded to reduce bias in form of artificial periodicities. The CWT (Morlet function) allows the visualisation of growth periodicities in the studied O. faveolata coral in form of a power spectrum with computed cycle lengths within the selected grid row dataset. Significant growth periodicities are labelled with thin black lines, marking the significance level of p ≤ 0.05 (Fig. 1c). A cone of influence was added (Fig. 1c) to consider boundary intervals in the observed cycle lengths along the growth time of the coral. An assignment of time-specific periodicities along the y-axis of the CWT was done based on the average number of grid point data per growth increment (respectively, per growth year).

Statistical analysis

Statistical analyses with skeletal and geochemical data were performed with PAST v4.06b. Regression analyses were used for overall long-term trends of skeletal growth patterns. Statistical correlations were done to identify likely connections between the (intra-)annual skeletal density values and other datasets. Analyses of density data were obtained either on annual scale with bulk density values (for skeletal growth patterns) or with intra-annual density data tuned to the same spatiotemporal resolution (for geochemical proxy data).

For statistical analysis with annual growth patterns, annual bulk density measurements have been made for each growth increment with the GUI. Linear extension rates were measured along the growth axis of the coral skeleton by measuring distances between each top of a summer HDB with available X-ray images. Six measurements have been made per growth increment and were averaged to annual means with a standard deviation. Calcification rates were calculated after the following equation, which is commonly used in studies on coral sclerochronology (Helmle and Dodge 2011):

$$\mathrm{Calcification}\left[\mathrm g\;\mathrm{cm}^{-2}\ast\mathrm{yr}^{-1}\right]=\mathrm{Linear}\;\mathrm{extension}\;\mathrm{rate}\;\lbrack\mathrm{cm}\; \mathrm{yr}^{-1}\rbrack\ast\mathrm{skeletal}\;\mathrm{density}\;\lbrack\mathrm g\;\mathrm{cm}^{-3}\rbrack$$

Error bars for calcification rates were calculated by propagation of uncertainty.

Gischler and Oschmann (2005) investigated the herein used colony for carbon (δ13C) and oxygen (δ18O) isotope compositions at intra-annual resolution with respect to the PDB standard (Pee Dee Belemnite) (Fig. 2a, c). In addition to the age model calibration of the high-resolution density datasets with δ18O isotope values, the oxygen and carbon isotope data were used for testing further statistical connections with and implications on skeletal growth in O. faveolata. For statistical analysis, the obtained intra-annual density data were tuned to the same spatiotemporal resolution as the geochemical values. For this procedure, those density values were calculated as mean values, which are equivalent to the positions of individual boreholes from previous geochemical sampling. The reduced resolution still enables statistical analysis at intra-annual resolution.

Results

Intra-annual and bulk density datasets

A total of 38 growth increments (1962–1999 CE) was analysed completely by the selection of individual grid rows using the GUI, allowing the measurement of a dataset yielding a total of 1005 individually calibrated density values. Solely the oldest increment (1962 CE) could not be analysed completely, due to a change in orientation of HDBs and LDBs, limiting spatial resolution of further measurements. An average of 27.6 ± 3.5 grid row measurements per growth increment is available, which would translate into a temporal resolution of approximately 2 weeks. Based on the age versus depth model, this overall temporal resolution of single grid points fits relatively well for the complete time frame of skeletal growth within the present coral (r = 0.9995, p = 0.000, n = 1005) (Fig. 1d). The average skeletal density for this colony is calculated with 0.904 ± 0.036 g cm−3. The mean uncertainty of 3.95% is given for error bars of the calibrated density measurement. An average variance of the mean bulk density amounts to 0.016 g cm−3. Compiled results for each analysed core slab of BZE-2 are figured in Table 1. The highest density values are observed for the HDB during summer 1999 (Top of BZE-2–1 with up to 1.544 ± 0.058 g cm−3) (Fig. 1b) and for the HDB during summer of 1963 (up to 1.597 ± 0.058 g cm−3) (Fig. 1b). The former peak is primarily caused by the presence of organic matter of the holobiont (dry coral tissue and remains of zooxanthellae) and, therefore, represents anomalous values. HDB/LDB alternations are clearly expressed and/or can be identified by correlating the density values with the measured scans and available X-ray images (Fig. 1b). Additionally, intra-annual density bandings occur also, but these are usually less pronounced compared to summer HDBs, which correspond to the available low δ18O peaks (Fig. 2a). The obtained density values show an overall decreasing trend with a statistical significance, which was calculated either for intra-annual density data (r =  − 0.531, p = 0.000, n = 1005, respectively, r =  − 0.538, p = 0.000, n = 966 by excluding outliers from the base and the top) (Fig. 1b, 2b) as well as for annual bulk density values (r =  − 0.641, p = 0.000, n = 37, respectively, r =  − 0.659, p = 0.000, n = 35 by excluding outliers from the base of the coral) (Fig. 3a).

Fig. 3
figure 3

Comparison of annual growth patterns in core slab BZE-2: bulk density (a violet), linear extension rate (b green) and calcification rate (c light blue). Vertical dotted lines indicate HDB/LDB boundaries between each annual growth increment. El Niño years are indicated (very strong El Niño red, strong El Niño light red)

Full size image

Growth periodicities in Orbicella faveolata

The CWT (Fig. 1c) yields several periodicities, which are statistically significant (p ≤ 0.05) and occur mostly continuous through the complete skeletal growth time. Cycle lengths include periodicities with a 1-year cycle, a ~ 2- to 4-year cycle, a ~ 5- to 8-year cycle and a ~ 9- to 10-year cycle. Intra-annual growth fluctuations are just occasionally expressed and show in few cases a cone-like transition to the 1-year cycle. Such cone-like patterns in high-frequency fluctuations in CWTs were described as local abrupt changes (LACs) by Hochman et al. (2019) and represent artefacts affecting lower-frequency patterns. Therefore, partial bias within the expression of the 1-year cycle cannot be ruled out (at least for a few increments). All observed periodicities are within the frame of the cone of influence, suggesting significant periodicities for observed cycle lengths.

Statistical analysis of datasets

Annual bulk density values show statistically significant long-term connections with linear extension rates and calculated calcification rates (Table 2, Figs. 3 and 4c and d). This connection is expressed as an anticorrelation between bulk density values and linear extension rates (r =  − 0.598, p = 0.000, n = 37), whereas a positive correlation is seen with calcification rates (r = 0.523, p = 0.001, n = 37). A statistically significant connection of down-sampled skeletal density values and δ13C isotope datasets is evident (r = 0.402, p = 0.000, n = 170) (Table 2, Fig. 4a). However, an overall significant correlation between skeletal density and δ18O values is not evident (r =  − 0.029, n = 0.710, n = 170) (Table 2, Fig. 4b). Nevertheless, the lower and mostly even lowest intra-annual δ18O isotope values correspond temporally to the highest skeletal density values (Fig. 2), which might support the formation of HDBs during warmer summer months in O. faveolata.

Table 2 Statistical correlations with skeletal density values of BZE-2
Full size table
Fig. 4
figure 4

Statistical correlations of skeletal density obtained from core slab BZE-2 with geochemical (a δ13C, b δ.18O) and skeletal growth data (c linear extension rate, d calcification rate). For correlations with geochemical data, down-sampled density values have been used (a, b green plots). In case for annual growth data, annual means were used (c, d light blue plots)

Full size image

Discussion

Based on previous annual bulk density measurements performed by Deveaux et al. (2017), the same colony of O. faveolata coral was herein re-investigated with the aim of testing the quantification of skeletal density and density banding patterns at higher temporal resolution as well as its use as a palaeoecological and palaeoclimatological archive. The extracted coral dataset contains the first obtained skeletal density values at high intra-annual resolution (approximately bi-weekly resolution per grid row), by using a two-dimensional grid-scanning approach with a 241Am gamma densitometer. Based on the age versus depth model, the overall temporal resolution in the studied O. faveolata colony coincides with an approximately bi-weekly resolution for individual grid points. Nevertheless, the temporal resolution for individual grid points depends directly on the skeletal growth length (respectively linear extension rate) per growth increment, which strongly controls and limits the potential for measuring skeletal density at high intra-annual resolution.

Previous attempts with gamma densitometry focused on one-dimensional track-scanning techniques (see, e.g. Barnes and Devereux 1988; Chalker and Barnes 1990; Lough and Barnes 1990a, b, c, 1992, 2000; Carricart-Ganivet et al. 2012). Lough and Barnes (1990c) discussed the influence of the beam diameter in gamma densitometric studies for the quantification of coral skeletal density along one-dimensional tracks through the coral skeleton. However, with the two-dimensional grid-scanning gamma densitometer developed by Deveaux et al. (2017), in which each single grid point represents a single calibrated density value, the effect of small-scaled local variations in skeletal density of a coral skeleton (e.g. pore spaces) can be reduced to an influence of minor importance. Therefore, by selecting rows of grid points with the GUI, it is possible to overcome the above-mentioned limitation. The local influence of the calyx morphology of each coral polyp, as discussed in Lough and Barnes (1990c), can be reduced in the same way.

The observed long-term decrease in skeletal density is accompanied by a similar decrease in the calculated calcification rates (Fig. 3c). A long-term decline in skeletal density of massive reef corals during the twentieth and twenty-first century has been observed by various authors (e.g. De’ath et al. 2009; Helmle et al. 2011; Mollica et al. 2018; Rippe et al. 2018) and was discussed in the context of the impact of ocean acidification (respectively, reduced aragonite saturation in the surface ocean) due to increasing anthropogenic emission of carbon dioxide during the industrial era. The present colony of O. faveolata follows the “stretching modulation” proposed by Carricart-Ganivet and Merino (2001) for corals growing in environments with more stressed conditions, under which the linear extension rates and skeletal densities are clearly anticorrelated. The reduced calcification rates are positively correlated with the decreasing skeletal density in the present O. faveolata colony, which would support such a modulation type, in that lower skeletal density has been developed by reduced skeletal accretion of CaCO3, during more increased skeletal growth (linear extension). It should be noted that the herein studied O. faveolata colony only shows just a slight increase in linear extension (without statistical significance); nevertheless, the anticorrelation with skeletal density still is evident. Such combined growth patterns of less dense and overall reduced carbonate accretion might indicate the impact of ocean acidification within the studied O. faveolata colony. However, it needs to be verified, if an overall reduction in skeletal density holds true for O. faveolata corals growing within the BBR (independent of different reef zones), so that an impact of ocean acidification can be assumed for this coral species within the study area.

The timing of the formation of HDBs in the studied O. faveolata colony correlates with the lowest values in available oxygen isotopes (Fig. 2) and therefore might reflect the previously suggested connection of HDB formation during warmer summer months in O. faveolata and their sibling species of the former “Montastraea annularis group” (e.g. Hudson et al. 1976; Carricart-Ganivet et al. 2000; Cruz-Piñón et al. 2003; Mendes 2004). However, an overall statistical significance of skeletal density and oxygen isotope ratios has not been verified in the present colony. The influence of salinity variations on oxygen isotope fractionation cannot be excluded. The Orbicella colony was sampled thriving under fully marine conditions within the central BBR; however, annual variation of up to 0.4‰ was observed in δ18Oseawater offshore Belize (Gischler and Storz 2009) (their Table 2). Another limitation might be sampling bias by extracting carbonate samples from different intra-skeletal elements or even individual calices from within the colony. This issue and limitations by sampling resolution have been pointed out by Watanabe et al. (2002).

In contrast to δ18O values, a statistically significant positive connection is evident for δ13C values and skeletal density in the investigated O. faveolata colony. During photosynthesis of zooxanthellae, a fractionation of 13C/12C ratios occurs, leading to a preferred uptake of lighter 12C isotopes in organic matter, whereas the carbonate skeleton of the coral will be enriched in heavier 13C isotopes (e.g. Swart 1983; Linsley et al. 2019). Therefore, carbon isotope values are likely figured to be a primary function of respiration and metabolic effects, expressed by light-dependent photosynthetic activity (carbon uptake from dissolved inorganic carbon/DIC from ambient seawater) as well as coral heterotrophy (carbon uptake from zooplankton feeding) (see, e.g. Swart 1983; Grottoli 2000; Linsley et al. 2019 and references therein). This corresponds well with the observed positive correlation between tuned density data and δ13C values, which implies a likely connection of meso-architectural CaCO3 accretion and coral mixotrophy and, therefore, the influence of zooxanthellate activity. However, other author groups (e.g. Watanabe et al. 2002; Linsley et al. 2019) have reviewed and discussed limitations of δ13C values from coral-bound carbonate skeletons; therefore, other and more complex relationships might explain the observed relationship in the investigated colony as well.

Time series analyses of the studied coral colony of O. faveolata have shown that a 1-year periodicity is the most dominant cyclic length, which is clearly a pattern of annual HDB/LDB alternation during skeletal growth. In addition to annual HDB/LDB alternations, further expressions of intra-annual density bandings occur as well. Similar intra-annual density banding patterns are common in coral colonies within the former “Montastraea annularis group” (and therefore also in O. faveolata) and have been generally assigned as “winter stress bands” formed during periods with colder SSTs (see, e.g. Hudson et al. 1976; Dodge et al. 1992). However, the resolution of the used oxygen isotope data limits a precise identification of such intra-annual density bandings. Future geochemical sampling with a higher temporal resolution and an evenly spaced sampling procedure might overcome this limitation.

The observed periodicities with cycle lengths of approximately 2 to 4 and 5 to 8 years might be evidence for the influence of the El Niño–Southern Oscillation (ENSO) phenomenon (respectively, El Niño years). Wórum et al. (2007) developed a model for the formation of double-stress HDBs in O. faveolata related to thermal stress (> 28.8 °C) even prior to bleaching events. In case of the strongest El Niño event known in the study area, which occurred during 1997–1998 (e.g. Aronson et al. 2000, 2002), one would expect a clear expression as double HDBs. However, in the studied colony of O. faveolata, no such banding pattern can be seen for this El Niño event. During the other very strong El Niño event of 1982–1983 CE, however, double-stress HDBs were developed during the summer months (Fig. 1a), which are likely connected to thermal stress events. The only other identified double-stress band occurs during 1984 CE (see Fig. 1a). This suggests that the approach of Wórum et al. (2007) for the identification of thermal stress intervals (and a likely connection to El Niño years) is not adequate within the present colony for reasons not entirely clear. Within the time intervals of both strong El Niño events (1982–1983 CE, 1997–1998 CE), bulk density decreases (Fig. 3a) and, therefore, might suggest a reaction to thermal stress levels. However, in case of the El Niño event of 1982–1983 CE, an additional increase in linear extension rate is evident, accompanied by a similar increase in calcification, which fits with the stretching modulation model after Carricart-Ganivet and Merino (2001). During the El Niño years 1963–1964 CE, a significant decrease in all skeletal growth patterns is expressed (Fig. 3), following directly on the formation of the striking HDB at the base of the studied colony. The coral might have suffered from a previous environmental disturbance, after which a recovery interval started. This may have been augmented by the aftermath of cat. 5 hurricane Hattie, which crossed the central BBR and made landfall in the study area in late October 1961 with devastating effects on the Belize reefs (Stoddart 1962). The decreasing growth following this time interval might be related to the overall weakened growth regeneration of the colony.

Gischler and Oschmann (2005) observed a decadal cyclicity of approximately 10 to 15 years, based on δ18O values within this colony and suggested a likely connection to the Atlantic SST dipole variation. Chang et al. (1997) modelled the variation of the Atlantic SST dipole with a quasi-decadal periodicity (on average around 12 to 13 years) and proposed a thermodynamic interaction between the Atlantic Ocean and the atmosphere, leading to reciprocal influences of SST variations and wind-induced heat entry. In the present study, time series analysis of intra-annual density datasets of the studied O. faveolata colony reveals a periodicity of approximately 9 to > 10 years, which might be a supporting evidence for an influence of the Atlantic SST dipole. A general limitation might be the fact that the studied time series encompasses 38 years only so that decadal, quasi-decadal or even multidecadal periodicities cannot be deciphered with certainty.

In summary, the presented results show that the application of two-dimensional grid-scanning gamma densitometry is a useful tool to analyse temporal highly resolved skeletal density fluctuations in corals. A basic limitation is the precise connection of individual density bandings with corresponding geochemical data, which can be easily overcome with an equidistant temporal sampling resolution comparative to the obtained grid point selection approach. However, as pointed out by other authors (e.g. Lough and Cantin 2014), it is necessary to analyse not just single coral colonies but to develop a better understanding of the variability in skeletal growth for whole populations within a reef ecosystem and its different intra-reef settings. Therefore, it will be crucial to analyse datasets from several corals within the BBR to develop a better understanding on skeletal growth of O. faveolata and the impact of environmental influences within the study area. Furthermore, future studies should also focus on the use of two-dimensional grid-scanning gamma densitometry for the analyses on taxon-specific skeletal growth strategies in other massive reef-forming corals, in which the “stretching modulation of skeletal growth”, as described by Carricart-Ganivet and Merino (2001), is not inherent. Future studies should furthermore include evaluations on the spatiotemporal resolution of single grid point measurements in other massive corals with lower skeletal growth rates and discussions on the interspecies-specific timing and expression of density bandings.