The skeleton of zooxanthellate corals (z corals) is a highly organised, porous hard tissue formed of mineral CaCO3 (aragonite). In X-ray images of slices parallel to the axes of the corallites (axes of maximum growth), massive z coral skeletons typically display alternations of light and dark bands. One pair of these “density bands” usually represents 1 year of growth (Knutson et al., 1972) and forms the basis for the calibration of internal age models and for estimates of the extension rates, i.e. the rate of upward and outward growth of the colony surface (Lough and Cooper, 2011). Skeletal bulk density is a measure of the pore volumes within the skeleton; the less porosity, the closer the density will be to that of mineral aragonite (2.93 g cm-3). Extension rate and density combine for estimates of calcification rates according to Eq. (1) (Lough and Cooper, 2011): calcification rate(gcm-2yr-1)=annual extension rate(cmyr-1)×density(g cm-3). Alternative concepts of quantifying coral skeletal growth have been reviewed by Pratchett an co-workers (Pratchett et al., 2015). In addition to the basic calcification parameters described above, serial chemical and isotope proxy data retrieved along the direction of maximum skeletal extension provide independent quantitative measures of the environment. Stable isotope ratios of the oxygen (δ18O) are sensitive to sea surface water temperature (SST) and serial samples over the growth bands allow for the documentation of seasonal or interannual SST variability on multi-annual timescales (decade and century scale; Felis and Pätzold, 2004; Leder et al., 1996; Swart, 1983). Limitations of the method pertain to the influence of seawater δ18O which is subject to changes due to precipitation and/or evaporation (i.e. salinity), river discharge and global ice volume. To overcome the problem of variable seawater δ18O for SST estimates, chemical element proxies of SST rather insensitive to evaporation and/or precipitation (Sr / Ca, U / Ca, Mg / Ca) are in use in combination with skeletal δ18O (Felis et al., 2004; Shen and Dunbar, 1995; Swart, 1981). Other chemical elements (Ba / Ca, Y / Ca, B / Ca) and carbon stable isotope ratios (δ13C) have been shown to be recording sensitively productivity, river discharge, pH, or also subtle diagenetic alterations (Allison et al., 2007; McCulloch et al., 2003; Sinclair et al., 1998; Swart et al., 2010).

In the geological record, the skeletons of scleractinian corals and other sedimentary grains composed originally of metastable aragonite (CaCO3) usually form moldic porosity, or are more or less completely replaced by mosaics of blocky calcite spar (Schroeder and Purser, 1986). Although these secondary alterations generally pose no problem for classical approaches in palaeoecology and taxonomy, all information stored as isotope and geochemical proxy data have been reset and makes the corals no longer available as environmental or geochronological archives. The first diagenetic alterations of the skeletons still happen at the sea floor, in deeper parts of the skeleton where the living organic tissues were previously withdrawn. These alterations represent growths of inorganic aragonite fiber crystals and subtle dissolution phenomena within the centres of calcification (COC; Perrin, 2004). Differential diagenetic processes on crystalline phases and organic matrices also exist and include aragonite–aragonite recrystallisations associated with a loss of micron-sized growth information (McGregor and Gagan, 2003; Nothdurft and Webb, 2009; Perrin, 2004). In contrast, in the classical freshwater diagenetic environment, the primary surface area of the skeleton controls diagenetic susceptibility and rates of alteration (Constantz, 1986; Dullo, 1984). The freshwater effects are dominated by dissolution via moldic porosity and subsequent reduction of pore spaces by cementation, or dissolution and associated crystallisation of blocky calcite without developing a significant moldic stage (Bathurst, 1975). In the latter process, ghost structures reflecting original microstructures will be preserved (Flügel, 2004). More often, ghost structures of the growth bands form by subtle, diffusion-controlled dissolution which preferentially starts at the COCs and continues to form increasingly hollow skeletal structures (Reuter et al., 2005). The rate of skeleton-internal dissolution via diffusion differs among growth bands within a specimen and responds to bands of higher and lower density (Reuter et al., 2005). Given the situation where no secondary addition of carbonate material has taken place, however, the hollow structures may still be suitable for isotope and geochemical proxy analysis (Mertz-Kraus et al., 2008, 2009a, b). Following infilling by late diagenetic calcite spar, this differential dissolution process leaves records of growth bands from which skeletal extension can be retrieved (Brachert et al., 2006b; Johnson and Pérez, 2006; Shinn, 1966). But, this process of dissolution and subsequent cementation of moldic and intra-particle porosity tends to destroy all information pertaining to skeletal density. Alteration of the primary skeleton along this diagenetic pathway is obvious by the presence of calcite, either replacing skeletal structures or infilling skeletal porosity. While the petrographic aspect of the calcite documents the type of freshwater or burial alteration environment, cathodoluminescence analysis and geochemical data may provide further information as to the redox character of the diagenetic fluids (Flügel, 2004). Alteration of aragonite is commonly a rapid process, but in the rare event of low pore-water circulation rates, corals do escape diagenetic alteration (Anagnostou et al., 2011; Brachert et al., 2006a, 2016; Denniston et al., 2008a; Gothmann et al., 2015; Griffiths et al., 2013; Mertz-Kraus et al., 2008).

In this study we present calcification data from z corals with rather intact primary skeletal density from Plio-Pleistocene interglacial deposits on the Florida Platform (USA; Fig. 1). We show that corresponding calcification rates were 50 % lower than they are in the present-day Western Atlantic (WA). For an understanding of the possible mechanisms behind these low calcification rates, we use modern analogue data compiled from the literature on recent z corals of the WA and Indo-Pacific (IP). According to this database, temperature generally boosts calcification rates in modern z corals, but field studies on single species of z coral suggest the rates to decline beyond optimum values (Carricart-Ganivet et al., 2012; Cooper et al., 2008). The non-linearity of calcification rates (g cm-2 yr-1) derives from inputs of two independent variables: skeletal growth rate (extension rate, cm yr-1) and skeletal density (g cm-3; Lough, 2008). The temperature effects on extension rates of Porites from the IP are well documented over a large temperature window and display slow increases with temperature below but sharp decreases above optimum (Cantin et al., 2010; Carricart-Ganivet et al., 2012; Lough and Barnes, 2000). The temperature responses of extension rate and density, however, are generally believed to markedly differ according to taxon and/or ocean region (Highsmith, 1979) and are further complicated by proximality trends reflecting temperature and seasonality gradients, exposure, effluxes of “inimical” bank waters, or nutrient supplies (Lough and Cooper, 2011; Manzello et al., 2015b). We discuss whether the patterns of z coral calcification found in the fossils from the Florida Platform is a local or global signature corresponding with temperature stress or low supersaturation of the sea water with respect to aragonite (Ωaragonite) during the Plio-Pleistocene interglacials. This study complements two previous papers using sclerochronology of bivalves and z corals for reconstructions of the paleoenvironments and long-term changes of seasonality in southern Florida (Brachert et al., 2014, 2016) and provides a discussion of the quantitative data in the context of recent global z coral calcification patterns.

During the Plio-Pleistocene interglacials, global sea levels were up to 22 m (Miller et al., 2012) or even 35 m higher (Dowsett and Cronin, 1990) and global mean temperatures 2 to 4 ∘C warmer than present, whereas SSTs of the warm pools at low latitudes were ∼ 2 ∘C higher than present (Fedorov et al., 2013; O'Brien et al., 2014). Although dramatic cooling occurred in the high latitudes, long-term atmospheric pCO2 appears to have remained rather constant after the mid-Pliocene climatic optimum (∼ 3 Ma) until the present (Seki et al., 2010). During and before the optimum, however, pCO2 reached values expected for the end of this century through the burning of fossil fuels (IPCC, 2013; Seki et al., 2010). Modelling of the oceanic carbonate systems suggest the long-term pCO2 changes to have had no effect on the saturation state of seawater with regard to Ωaragonite (Hönisch et al., 2012), but evidence exists that rates of microbial carbonate precipitation and skeletal accretion of planktic foraminifera differed over the last glacial–interglacial cycle (Barker, 1986; Beaufort et al., 2011; Riding et al., 2014).

The Plio-Pleistocene Florida carbonate platform represents a stack of shallow marine carbonate sequences formed during sea level highstands which are separated by paleosols or thin freshwater units formed during lowstands. A pronounced reef system existed along the south-western margin of the peninsula (Meeder, 1979). The single unlithified marine units contain a diverse mollusk and coral fauna comparable to that of the present reef tracts and back-reef systems (Meeder, 1979; Petuch and Roberts, 2007). Combined oxygen and carbon stable isotope data (δ18O, δ13C) of diagenetically pristine mollusks and z corals from the platform sediments reflect the complexity of the depositional setting including brackish to hypersaline and well-mixed, open marine environments (Brachert et al., 2014; Lloyd, 1969; Tao and Grossman, 2010). The reasons for high benthic carbonate productivity by mollusks during the Plio-Pleistocene is controversial, and has been suggested to be due to high nutrient concentrations resulting from freshwater input (Tao and Grossman, 2010) or upwelling (Allmon, 2001; Allmon et al., 1995; Brachert et al., 2016; Emslie and Morgan, 1994; Jones and Allmon, 1995). Recently, SST estimates for the Pliocene and Pleistocene interglacial units have been retrieved from δ18O values from the reef corals Solenastrea and Orbicella and assuming a modern seawater value for δ18O (δ18Owater) at the Florida Reef Tract (FRT). Apart from assumptions for δ18Owater low SSTs are believed to be essentially the effect of upwelling. The large range of SST values is also likely in part an artifact of the uniform value used for the calculations, irrespective of sampling locality and stratigraphic unit (Brachert et al., 2016). In contrast, seasonal SST variability (∼ 7 ∘C) inferred from cyclic δ18O variations of the fossils is more independent of assumptions of δ18Owater. Reconstructed seasonality is not only remarkably constant within specimens and over the last 3.2 Ma, but also fits modern surface seasonality along the reef tract (Brachert et al., 2014, 2016). Large seasonality as prevailing off North Carolina (Macintyre and Pilkey, 1969) or in inner coastal waters of Florida Bay (FB; Swart et al., 1996) has not been encountered in the data from the reef corals and has also been taken for inferring a normal shallow-marine environment without unusual stress from cool waters or evaporation and freshwater influxes (Brachert et al., 2014).

In southern Florida, the most extensive growth of reef corals occurs at present along the FRT on the Atlantic side of the peninsula, whereas only limited z coral growth occurs along the Gulf side in the west and the shallow FB in the south-east. On the Atlantic side, coral communities are characterized by diverse stands comprising abundant Orbicella (Lidz, 2011), whereas on the Gulf side and in FB, coral growth is restricted to the two eurytopic taxa Siderastrea and Solenastrea (Okazaki et al., 2013; Swart et al., 1999). Published extension rates for recent Solenastrea inhabiting the most marine segments of FB range from 0.51 to 0.9 cm yr-1 (Hudson et al., 1989; Swart et al., 1996). Recent Solenastrea has also been recorded to grow under rather cold water conditions along the US south-eastern Atlantic coast off North Carolina (Macintyre and Pilkey, 1969), but quantitative calcification data from that setting are not available, leaving the question unanswered regarding the effects of low SST on extension and density. Colony sizes at the northern sites similar to those of the lower latitudes have been suggested to indicate similar extension and calcification rates, however (Macintyre and Pilkey, 1969).

Z corals were sampled from four distinct stratigraphic units of the Florida carbonate platform (USA) representing interglacial highstands of sea level subsequent to the Pliocene warm period. They were dated 3.2, 2.9, 1.8 and 1.2 million years (Ma) of the mid-Pliocene and early Pleistocene (Fig. 1, Table 1; Brachert et al., 2014). Our own sampling focused on Solenastrea (n= 11) which is a common taxon in the Plio-Pleistocene shallow water carbonates of south-western Florida. This data set was complemented by specimens of Orbicella (n= 2) and Porites (n= 1) and one data set of a Solenastrea taken from the literature comprising serial δ18O and δ13C values and annual extension rates (Roulier and Quinn, 1995; Table 1).

Fossil corals selected for this study were cut into < 1cm thick slabs along the plane of maximum growth using a conventional rock saw equipped with a water-cooled diamond blade. All corals were screened for diagenetic alteration using a binocular microscope and scanning electron microscope (SEM). In order to detect minimal contaminations by secondary calcite, powder samples taken at random were prepared for X-ray diffraction (XRD) and analysed using a Rigaku Miniflex diffractometer at angles between 20 to 60∘ 2θ. Only skeletal areas that retained their original aragonite mineralogy (XRD), skeletal porosity and microstructure without evidence for significant secondary crystal growth or dissolution (microscopic and SEM observation) were accepted for further sample preparation. Coral slabs of equal thickness were X-rayed using a digital X-ray cabinet (SHR 50 V) to identify potential zones of diagenetic alteration (McGregor and Gagan, 2003; Reuter et al., 2005), bioerosion, and to document the density bands (Knutson et al., 1972). One coral specimen (452K1) was analysed geochemically using LA-ICP-MS (Böcker, 2014) with regard to concentrations of environmentally sensitive elements (e.g. Sr / Ca, U / Ca, B / Ca) and following recommendations for evaluating the diagenetic status of corals from strongly lithified and altered limestone (Anagnostou et al., 2011; Gothmann et al., 2015). LA-ICP-MS analyses were performed at the Max Planck Institut für Chemie (Mainz, Germany) using a NewWave UP 213 laser ablation system coupled to a ThermoFisher Element 2 ICP-MS.

Quantitative density measurements were made using the software CoralXDS (freeware) according to Helmle and co-workers (Helmle et al., 2002). In this approach, the CoralXDS software compares the gray values recorded in X-radiographs from corals with those from aluminum plates having the same thickness as a background picture and an aluminum wedge for density calculations. Measurements were done along transects parallel to the corallites and parallel to the sampling transects for stable isotope analyses (Brachert et al., 2016). Bulk skeletal density was calculated as the mean of all individual measurements taken along a given transect. Calibration of the measurements was tested by measurements of standards for zero density (air) and massive aragonite (slice of a Glycimeris bivalve shell having a thickness equaling that of the coral slice). External analytical precision of the routine measurements was tested by double blind measurements, and mean deviation from regression (R2= 0.91, p < 0.05) was found to be 0.04 ± 0.01 g cm-3 (range = 0.02 to 0.05 g cm-3; n= 18).

As a baseline for the description and interpretation of the data from the fossils, we use calcification data from recent corals reported in the literature deriving equally from tropical and high latitudinal localities within the shallow-water reef belt (Baker and Weber, 1975; Bessat and Buigues, 2001; Carricart-Ganivet et al., 2000; Carricart-Ganivet and Merino, 2001; Dodge and Brass, 1984; Dustan, 1975; Elizalde-Rendon et al., 2010; Fabricius et al., 2011; Goodkin et al., 2011; Graus and Macintyre, 1982; Helmle et al., 2011; Highsmith et al., 1983; Hudson et al., 1989; Lough, 2008; Mallela and Perry, 2007; Tanzil et al., 2009), and one unpublished record of Solenastrea from FB (FB-6). We present a set of three descriptive diagrams for a comparison of the patterns of calcification (extension rate, bulk density, calcification rate) in the modern and fossil z corals on the basis of linear regression. For a deeper understanding of the processes, we further apply quadratic polynomial regression models of experimental data calibrated with SST to account for the established non-linearity of life processes.

Stable isotope data described here are the same as reported in companion publications by Brachert et al. (2014, 2016) where all details of the methodology of sampling and analytical procedures have been reported in detail. All carbonate values are given in per mil (‰) relative to PDB according to the delta notation.

The scleractinian genus name Orbicella is used for corals previously assigned to Montastraea according to the revised taxonomic classification of the reef coral family Mussidae by Budd et al. (2012). According to the same work (op. cit.), the genus Diploria has been split into the genera Diploria and Pseudodiploria. We use the two genus names in combination as Diploria/Pseudodiploria, because our database likely incorporates material from both genera sensu (Budd et al. 2012).

Statistical analyses were performed using the PAST palaeontological statistics software package (version 3.01) for education and data analysis (freeware folk.uio.no/ohammer/past/). Variability of stable isotope data (δ18O, δ13C) was evaluated using the t test. A linear bivariate model was tested as to whether there were no statistical differences in the stable isotope values in a data set (p > 0.05) against the alternate hypothesis that there were significant differences (p < 0.05). Equality of regression slopes was tested using the f test as assumed by analyses of covariance (ANCOVA). One-way analysis of variance (ANOVA) tested if there were no statistical differences in the mean growth parameters (extension, density, calcification) between two given coral sites (p > 0.05) against the alternate hypothesis that there were significant differences (p < 0.05).

Visual inspection of the skeletons using a binocular microscope (× 15 enlargement) and SEM revealed clean skeletal surfaces not covered systematically by secondary cements (Fig. 2a, e), except for localised, micron-scaled patches of spherulitic aragonite or patches of isopachous aragonite (Böcker, 2014). SEM observation has not revealed any clear evidence for aragonite–aragonite recrystallisations (Fig. 2) but some porosity within the centres of calcification (COCs, Fig. 2b). The latter does indeed imply some dissolution has occurred, and therefore, subtle reductions of skeletal density, however, since dissolution at the COCs has also been reported from recent specimens (Perrin, 2004), this effect may also be present in the data from recent corals.

Secondary calcite is not documented by XRD analysis (detection limit of the method ∼ 1 %) and has very rarely been observed to occur within skeletal growth porosity but never within voids formed by preferential dissolution of the COCs (Fig. 2b) or microborings (Fig. 2e, f). Published geochemical screenings using LA-ICP-MS for specimen 452 K1 (Böcker, 2014) documented variable ratios of Sr / Ca and U / Ca which are in phase with serial δ18O data. These element ratios reflect SST variations consistent with reconstructions on the basis of serial δ18O values and recent instrumental seasonality along the FRT (Böcker, 2014). The positive correlation of the Sr / Ca with U / Ca and the B / Ca ratios fluctuating between 0.3 and 0.6 mmol mol-1 is fully consistent with modern z corals and implies little alteration has taken place, especially because boron is known to be a diagenetically highly volatile element (Allison et al., 2010; Böcker, 2014). According to our conviction, all these data provide no critical evidence for the alteration of the original skeleton. Because of this line of reasoning and low overall calcite content evident from XRD analysis (calcite below detection limits), we refrained from measuring element ratios sensitive to the redox conditions of calcite precipitating freshwaters or burial fluids (Fe / Ca, Mn / Ca) and other more sophisticated geochemical methods as potential measures of alteration (Anagnostou et al., 2011; Gothmann et al., 2015).

X-radiographs display very regular expressions of density bands, concordant with the growth structures of the skeleton and stable isotope records, but no cloudy density variations or patches of high (low) density as documented from diagenetically altered specimens (Böcker, 2014; Brachert et al., 2006a; Mertz-Kraus, 2009). The presence of concordant density bands implies the preservation of original density variations of the skeleton and, therefore, supports the conclusion of a pristine state of preservation for the specimens under consideration (Fig. 3). It should be noted that density was measured using X-ray densitometry along transects defined from visual inspection of radiographs, and measurements were taken only in segments of the skeleton not affected by borings (bivalves, sponges, sipunculids) or embedded encrusting biota (serpulids, bivalves). Bulk density data presented by this study and in a companion publication (Brachert et al., 2016) are, therefore, not influenced by the volume of macroscopic biogenic borings or incrustations, although these effects may also be inherent to published density data of recent corals. This is an important issue, because other approaches have used “net density” (i.e. the integrative weight of carbonate laid down by the coral and encrusting biota minus losses by bioerosion within a volume) for comparative calcification studies (Kuffner et al., 2013). In contrast to density, extension rate is not sensitive to diagenetic alterations and many data have been retrieved earlier from highly altered fossil coral specimens of Neogene age (Brachert et al., 2006b; Gischler et al., 2009; Johnson and Pérez, 2006; Reuter et al., 2005). All of these observations and reasoning suggest the z corals selected for this calcification study to be essentially unaltered by diagenesis.

The Pliocene and Pleistocene z corals from the Florida Platform display extension rates that range from 0.16 to 0.86 cm yr-1 with a mean value of 0.44 ± 0.19 cm yr-1 (n= 15, ± 1σ), bulk skeletal densities between 0.55 and 1.52 g cm-3 with a mean of 0.86 ± 0.22 g cm-3 (n= 14), and skeletal calcification rates from 0.18 to 0.54 g cm-2 yr-1 with a mean of 0.34 ± 0.11 g cm-2 yr-1 (n= 14; Fig. 4, Table 3). Annual extension rates and bulk skeletal density show a significant negative correlation (R2= 0.329; p= 0.026), i.e. density decreases with increasing extension rates. In contrast, extension rates and calcification rates display a positive relationship (R2= 0.484; p= 0.004), which implies that calcification rates also decline with increasing extension. Lastly, bulk density and calcification rate display no relationship (R2 = 0.025; p= 0.797; Fig. 4). Although no statistics were applied to the data of Orbicella (n= 2) and Porites (n= 1) their calcification systematics seem to be indistinguishable from those of Solenastrea according to visual assessment (Fig. 4). With regard to variability over geological time, extension rate, bulk density and calcification rate of the three genera Solenastrea, Orbicella and Porites from the Florida platform were plotted according to four time slices 3.2, 2.9, 1.8, and 1.2 Ma (Fig. 5, Tables 1, 2), and all calcification data were found to be indistinguishable among time slices according to ANOVA (p > 0.05). Published extension rates of z corals reported from various other fossil low-latitude sites of the WA region are ∼ 0.3 cm yr-1 in late Miocene reefs (Denniston et al., 2008b) and range from 0.3 to 0.8 cm yr-1 in Pliocene units (Johnson and Pérez, 2006), whereas they were 0.2 and 1.0 cm yr-1 in the Florida Reef Tract (FRT) during the late Pleistocene (0.13 Ma; Gischler et al., 2009). As such, they are all consistent with the low extension rates reported by our study (Fig. 4). Importantly, skeletal density data are not available from these sites due to pervasive diagenetic alterations, and therefore, skeletal density and calcification rates are unknown.

For the recent time slice (0 Ma) we use analogue data from southern Florida published in the literature and complemented in part by one new set of average values (FB-6) published here for the first time (Table 4).

The extension rates of recent Solenastrea from FB range from 0.51 to 0.89 cm yr-1 and are fully within the range found in the Pliocene and Pleistocene corals (Fig. 5). Density values have not been published from FB z corals so far; we measured a density of 1.07 g cm-3 (Table 2) which is compatible with fossil Solenastrea. The same is true for the Orbicella from FRT as compared to the two fossil Orbicella, whereas the density records available from the FRT-Porites are substantially above that from the fossil Porites which is near the lower end of the spectrum (Fig. 5, Table 4). Finally, calcification rates of all three taxa of the recent z corals in FB and FRT tend to be above the Plio-Pleistocene reconstructions (Fig. 5), and the average of all recent corals is significantly higher than the fossil average value (p < 0.05). From these observations the following three generalisations can be made: (1) the extension rates of the fossil z corals are indistinguishable from those of the recent corals, and no distinction exists between FB and FRT, nearshore and offshore. (2) Bulk density is essentially the same in recent and fossil Florida z corals, although some tendency towards higher bulk density as compared to the fossils may exist. (3) The calcification rates of the recent z corals are all higher than those of the fossils (Fig. 5).

Stable isotope proxy data of the growth environments from the corals used here for calcification records were described and interpreted in a companion paper (Brachert et al., 2016) and will not be repeated in detail. For estimates of SSTs, an equation using skeletal δ18O calibrated for Orbicella from FRT was applied (Leder et al., 1996) and making the assumption of a constant value of δ18Owater= 1.1 ‰ (recent FRT water) for all relevant interglacials (Brachert et al., 2016). On this basis, we found average annual SSTs between 19 and 26 ∘C which were likely moderated by intermittent upwelling. Reconstructed temperatures display a negative correlation with annual extension rates (p < 0.05) and a positive relationship with bulk density (p < 0.05). In contrast, no clear relation has been found between SST and calcification rate (p > 0.05), although visual inspection suggests an inverse correlation (Fig. 6). Making other assumptions for δ18Owater (but keeping the value constant for all specimens) will yield other temperature values, but the range of values between minima and maxima of average annual temperatures will remain unaffected.

Calcification of z corals responds to a complex array of environmental factors acting in concert as to control net calcification (Lough and Cooper, 2011). Next to water temperature, these factors include water depth, wave exposure, admixtures of “inimical waters” from carbonate bank interiors, high and low salinity or freshwater discharge, nutrient concentration, pH and aragonite saturation and ocean region; Cohen and Holcomb, 2009; D'Olivio et al., 2015; Ferrier-Pagès et al., 2000; Ginsburg and Shinn, 1964; Gladfelter et al., 1978; Hofmann et al., 2011; Johnson and Pérez, 2006; Klein et al., 1993; Lough and Cooper, 2011; Shinn, 1966). Thus, low calcification rates of the fossil corals can have multiple causes which are eventually hard to reconstruct. In attempting to sort out small-scale effects along environmental gradients, patterns related to taxonomy and non-linear calcification responses, we use a big picture approach beyond environmental gradients and regional acclimatisation effects and compare the reconstructed growth parameters within the frame of measured systems in southern Florida, the WA and IP (see methods section for data sources).

We use modern analogue data from southern Florida for an evaluation of the calcification rates documented here for z corals from Pliocene and Pleistocene units of the Florida Platform. In southern Florida, environments of z coral growth range from the salinity stressed environment of the FB where z corals only thrive within the most marine parts, to the open settings of the FRT variably affected by the outflow of “inimical” waters from the interior bank. Within this region, the highest rates of outflow of bankwater occur in the Middle Florida Keys where also the lowest calcification rates have been observed (Manzello et al., 2015a). Negative interference by inimical bank waters with z coral growth has been hypothesised, therefore, to be smaller in offshore reefs (> 4.5 km from coast) compared to inshore reefs (< 4.5 km from coast). Nonetheless, long-term data averaged from several Porites colonies (Manzello et al., 2015a) do not indicate to a measurable negative spatial onshore–offshore effect on z coral calcification. A proximality effect is also not inherent to the averaged analogue data shown in Fig. 5: although low calcification of Solenastrea in FB may be considered compatible with the inimical bank water hypothesis, even lower calcification rates of Porites from an offshore reef is clearly not. Apparently, small-scale spatial stress effects reported in the literature seem to be averaged out from the big picture. Because also no difference in calcification responses to environmental effects was found between Orbicella cavernosa and Porites astreoides (Manzello et al., 2015a), we consider the fossil data and recent analog data homogeneous entities not biased by systematic-taxonomical effects. From this line of reasoning we conclude the low calcification rates of the long-term fossil record from southern Florida not to reflect a restricted growth environment.

The calcification records presented by this study have been classified according to three descriptive patterns: (1) a negative relationship of extension rate with density being fully compatible with patterns of recent Orbicella. In recent Porites, the situation is more complex, because the pattern is documented only in the IP (Lough, 2008), but not in the WA (Elizalde-Rendon et al., 2010). (2) Extension rate and calcification rate showing a positive relation has been described also in recent Porites from the WA and IP (Elizalde-Rendon et al., 2010; Lough, 2008), but not in Orbicella from the WA which differ by a negative slope (Carricart-Ganivet, 2004). This is a surprising result, because the skeletal organization of Solenastrea closely resembles that of Orbicella and differs significantly from Porites, a pattern which was expected to be reflected in the systematics of calcification. (3) The fossil Solenastrea and recent Orbicella and Porites display deviating relationships with regard to bulk density and calcification rates: while the fossil Solenastrea shows no relationship, it is positive in Orbicella and WA-Porites but negative in IP-Porites (Carricart-Ganivet, 2004; Elizalde-Rendon et al., 2010; Lough, 2008). When plotted against water temperatures, the three calcification parameters and qualitative trends of the fossils are rather consistent with those of recent Orbicella from the WA (Carricart-Ganivet, 2004), both, in terms of the overall effects of temperature on extension rate and on bulk density. They differ, however, by the absence of a temperature control on calcification rates (or the presence of a likely negative slope according to visual inspection) in the fossils.

Calcification rates recorded by the fossil z corals are conspicuously low as compared to recent z corals from Florida (Fig. 5) which may represent, therefore, possibly no suitable analogue system. First of all, it should be noted, however, that the calcification data from the fossil Solenastrea (plus Orbicella and Porites) appear to be from a larger window of average annual temperatures (∼ 7 ∘C) than covered by field studies on recent z coral growth. Temperature differences behind growth data from southern Florida are rather small, and even growth data collected in the Gulf of Mexico and the Caribbean Sea both cover small gradients of average annual SSTs (∼ 1 ∘C) where Orbicella (Orbicella annularis) display positive calcification responses with increasing SST (Carricart-Ganivet, 2004). Although calcification rates are the same in both regions, average annual SSTs differ by ∼ 2 ∘C and likely reflect the acclimatisation of the same morphological taxon to regionally different SST regimes. Thus, acclimatisation effects on calcification seem to play a role within rather small observational scales. Within the same region, another species of the same genus (Orbicella falveolata), however, responds with declining calcification to this subtle gradient of ∼ 1 ∘C of average annual SST change (Carricart-Ganivet et al., 2012), either because acclimatisation is not yet fully accomplished, or because the SST regime is near the upper threshold of ecological tolerance of O. falveolata allowing no further positive acclimatisation. We assume, the latter is more likely and, therefore, calcification responses to SST seem to be non-linear over the full range of ecological tolerance of this and other taxa. This sort of non-linear response of calcification has been predicted by a modelling study on the ecological tolerance of Orbicella over a temperature window of 3–4 ∘C (Worum et al., 2007) and is also well documented by comprehensive field studies on Porites from the Great Barrier Reef system (IP; Cooper et al., 2008; De'ath et al., 2009, 2013). The tipping point between increases and decreases of calcification rates was found to be between 26 and 27 ∘C for Porites and Orbicella (Carricart-Ganivet et al., 2012; Cooper et al., 2008), or 28–29 ∘C according to modelling (Worum et al., 2007). This kind of large-scale observational data seems essential for interpreting fossil calcification data and, therefore, we discuss the calcification data in the context of the entire WA and IP.

Z coral skeletal calcification is closely linked with the saturation state of seawater with respect to aragonite, and low degrees of supersaturation have been demonstrated to cause low calcification rates (Cohen and Holcomb, 2009; Gattuso et al., 1998; Langdon et al., 2000). In the low latitudes, surface waters of upwelling regions have been shown to be low in Ωaragonite (Furnas, 2011). Upwelling has been documented also for the Plio-Pleistocene interglacials of Florida (Brachert et al., 2016), and z coral skeletons recording maximum upwelling according to their stable isotope composition, have the smallest density values but largest values of extension rate (Brachert et al., 2016). This conforms with findings from the Galapagos upwelling system, were z coral skeletal density is reduced under maximum upwelling stresses, but extension rate is higher than predicted from the ambient SST (Manzello et al., 2014). The low volumes of cements in intra-skeletal porosity of the corals and the low degree of cementation of the Plio-Pleistocene shallow-marine carbonates may reflect the effects of phosphate poisoning to carbonate precipitation in an upwelling regime (Hallock and Schlager, 1986; Manzello et al., 2014), but the benthic assemblages and low amounts of bioerosion do not provide compelling evidence for high eutrophy. If anything, these findings support intermittent upwelling inferred from stable isotope data which has positively interfered with z coral calcification on the Florida platform during the Plio-Pleistocene, but clearly documents minimal calcification rates to have coincided with episodes with minimum upwelling (Brachert et al., 2016). Thus, upwelling cannot be the prime reason for the observed low calcification rates.

Furthermore, the low extension rates of the Plio-Pleistocene z corals from Florida are fully compatible with those published from fossil z corals at various locations in the tropical WA (various taxa) which also range between 0.3 and 0.8 cm yr-1 during the Pliocene (Johnson and Pérez, 2006), ∼ 0.3 cm yr-1 in the late Miocene (Denniston et al., 2008b) and 0.2 and 1.0 cm yr-1 in the FRT during the late Pleistocene (0.13 Ma; Gischler et al., 2009; Fig. 7g). For this reason, low extension rates recorded by the Florida fossils are representative of the tropical WA at that time and were as such a large-scale regional or perhaps even global phenomenon. Reasons for globally low pH/low Ωaragonite in ambient water may be sought in high atmospheric pCO2 levels. However, for the last 3 Ma after the mid-Pliocene climatic optimum (∼ 3 Ma), reconstructed pCO2 was near pre-industrial levels and only during and before the climatic optimum it was at the levels predicted to exist by the end of this century (IPCC, 2013; Seki et al., 2010). For the long-term buffering effect of the ocean, Ωaragonite has been suggested to have not been significantly different from the present day during the Plio-Pleistocene interglacials, however (Hönisch et al., 2012). On the other hand, substantial pCO2 changes have been documented over the glacial–interglacial cycles of the Quaternary (Petit et al., 1999), concomitant with changes in calcification of calcareous plankton (Barker and Elderfield, 2002; Beaufort et al., 2011). Thus, low Ωaragonite may represent a potential driver of the observed low calcification rates.

Next to Ωaragonite, temperature is an important control of z coral calcification in the world oceans. Given the simplification in our reconstruction of SSTs discussed above, the extension rates still display a negative correlation with the average annual SST (p < 0.05) and bulk density a positive relationship with SST (p < 0.05). In contrast, no clear relation has been found between SST and calcification rate (p > 0.05), although visual inspection suggests an inverse correlation (Fig. 8). This pattern is qualitatively rather consistent with recent Orbicella (Carricart-Ganivet, 2004), however, at a substantially larger temperature window in the fossil material and an absent relationship or likely negative correlation of calcification rate with temperature (Fig. 6).

Over the large temperature window of 6.9 ∘C covered by the modern IP data, a pattern of changes driven by temperature has been documented using linear regression (Fig. 8d–f). In contrast, the temperature range documented by z corals from the WA database covers only 2.2 ∘C (Fig. 8d–f) and calcification data do not display any linear relationship. Instead of a linear fit, they can be approximated using a quadratic polynomial which should suggest the present temperature window realised by recent z corals of the WA to cover more or less the ecological spectrum of this coral province. Low extension rates documented by fossil z corals from Florida and many other locations of the Caribbean, therefore, potentially document temperatures either near their lower or upper levels of ecological tolerance. In our temperature reconstruction using skeletal δ18O values, we apply a value of δ18Owater which likely underestimates the actual SST because other methods consistently found SSTs of the WA warm pool ∼ 2 ∘C above present values during the last 5 Ma (Fedorov et al., 2013; O'Brien et al., 2014). Low calcification rates in z corals may, therefore, reflect warmer-than-present SSTs during the Plio-Pleistocene interglacials. Such an interpretation is consistent with concepts of nonlinear calcification responses to temperature in z corals (Brachert et al., 2013; Gischler et al., 2009; Worum et al., 2007). Correspondingly, approaches describing coral calcification within temperature windows of < 1 ∘C of annual temperature would not describe z coral calcification over the full spectrum of ecological tolerance of a given species and may describe calcification near the optimum or lower and/or upper threshold of calcification only. In application of this concept, z coral growth in the WA was likely under significant heat stress, and annual water temperatures 2 ∘C higher than at present were causing calcification rates 50 % lower than present day. It should be noted also that upwelling has been ascribed a mitigating effect on SST stresses depending on the depth of upwelling or the timing during the year (Chollett et al., 2010; Riegl and Piller, 2003) and maximum extension rates and/or minimum density of the Florida z corals coincided with a maximum of upwelling. Intermittent upwellings during the Plio-Pleistocene, therefore, seem to have created temporary refuges for z corals by episodically mitigating heat stresses (Brachert et al., 2016). This finding supports notions of hot SSTs during the Eemian interglacial to have resulted in reef kills at equatorial latitudes and poleward migrations of many z coral taxa (Kiessling et al., 2012). Our data also suggest recent coral reefs at equatorial latitudes to be potentially endangered from rising SSTs with ongoing climate change and ocean acidification (IPCC, 2013).