A simplified overview of the WALIS U-series fossil coral dataset and workflow is provided in Fig. 2, and the database field descriptors can be found here: 10.5281/zenodo.3961543 (Rovere et al., 2020). Published U-series analyses, elevation measurements and relevant metadata for each dated fossil coral are uploaded into WALIS either (1) manually via an online user interface or (2) with multiple entries at once using a spreadsheet template. Once entered into WALIS, all of the uploaded information is added to the WALIS fossil coral U-series database, and each analysis is assigned a unique identifier (WALIS U-series ID). Finally, all analyses from corals that can be used as RSL indicators (i.e., that are both in primary growth position and have an associated elevation measurement) are further subset into a fossil coral RSL database. Both databases can then be downloaded by any WALIS user.

This dataset contains several new features that have been added since the Dutton and Lambeck (2012) and Hibbert et al. (2016) compilations. Several key updates are as follows:

New sample identifiers are included, which make it easier to identify which U-series analyses are associated with the same coral colony. Sample IDs are reported in the format XX00-000-000. The first four digits denote the study that the coral age was published in, whereas the following sets of three numbers represent the coral sample and U-series analysis, respectively. For example, CH91-001-002 is the second U-series age reported for the first coral (here denoted with the number “001”), published in Chen et al. (1991). In situations where a coral specimen has been reanalyzed in multiple studies (e.g., for many samples from Barbados), the first four digits will refer to the oldest paper in which U-series ages were reported, while the “references” field will indicate the study the analysis came from. This ensures that the user can easily distinguish which samples came from the same coral colony, which was not always clear in earlier iterations of this database. This sample naming system has also been utilized for samples dated using other techniques that are reported in WALIS by other workers.

Corals precipitate their skeletons directly from dissolved ions in seawater, forming a calcium carbonate mineral called aragonite. As part of this process, uranium (U) is incorporated at parts-per-million concentrations as impurities within the aragonite crystal lattice, and in ideal, closed-system conditions thorium (Th) concentrations are negligible. This is because of the high particle reactivity of Th, which causes the element to have a relatively short residence time in the water column. Once the coral skeleton has formed, the U-series radiometric clock is effectively started, and the elapsed time is measured by the ingrowth of 230Th from the radioactive decay of 234U and 238U as the system returns to secular equilibrium (Edwards et al., 1987a, 2003). It is the disequilibrium that arises from the combination of high U concentrations and negligible detrital Th content that enables high-precision U-series dating of coral skeletal material, thus making fossil corals both valuable RSL indicators and an important source of absolute age control for other marine-derived sediments (e.g., marine terrace deposits).

Unfortunately, coral skeletal material is also highly susceptible to post-depositional alteration (i.e., diagenesis), particularly after exposure to meteoric waters, as is often the case with emergent LIG reef units (Thompson et al., 2003). As a result, a U-series date (i.e., calculated from U-series measurements without interpretation) must be carefully evaluated for signs of geochemical open-system behavior before it can be used to constrain a fossil coral age, which is an interpretation of the U-series date. Prior to U-series dating, coral samples are frequently prescreened using X-ray diffraction (XRD) and thin-section microscopy to identify evidence of recrystallization and/or alteration of coralline aragonite to secondary calcite minerals. Even coral samples that do not have detectable calcite and are not recrystallized can still yield anomalously young/old ages for an LIG deposit, indicating that mineralogically pristine samples can still display open-system behavior with respect to U-series isotopes (e.g., Fig. 3). Therefore, additional geochemical variables are often used to evaluate the quality of U-series ages (e.g., see Sect. 2.3).

Several models have been proposed to correct U-series ages that display open-system behavior, but it is well understood that patterns of diagenesis in altered corals at a study site follow multiple diagenetic pathways that cannot be explained by a single model (Henderson and Slowey, 2000; Scholz et al., 2007; Thompson et al., 2003; Villemant and Feuillet, 2003). While there are circumstances in which altered coral samples may be good candidates for open-system correction, this would require further analysis of diagenetic trends at each site, which is beyond the scope of this study, as no single open-system model can explain all of diagenetic variability in the dataset. For example, the Thompson et al. (2003) open-system model is well suited to correct diagenetic arrays common to Barbados and some localities in Western Australia. It does not, however, explain all modes of diagenesis present in the fossil coral record (e.g., Figs. 3a, 4). Hence, this analysis focuses on assessing closed-system ages.

Previous studies typically adopted a set of geochemical screening criteria to remove U-series data that have been altered through open-system behavior (e.g., Scholz and Mangini, 2007). Three of the most common geochemical variables used are

calcite content;

The strict screening protocol follows the general approach of Dutton and Lambeck (2012) and Hibbert et al. (2016), with some modification in the case where multiple subsamples from a single coral specimen were dated. To be accepted, a sample must have

calcite content < 2 %,

Many screening decisions are context based and were addressed separately for each site, but some general modifications to the strict screening protocol are addressed here. First, we expanded the calcite screening threshold to include all corals that are below the limit of quantification for the XRD method employed, which can be as high as 4 or 5 % for some studies.

Second, we allowed for a higher 232Th threshold of 12 ppb (i.e., a 230Th / 232Th activity ratio of ∼ 500) when age reproducibility can be verified by multiple subsamples from the same coral. This roughly corresponds to a 1 ‰ (or ∼ 0.13 kyr) effect on the measured U-series age, assuming a bulk upper-continental-crust contaminant (Dutton et al., 2015; Taylor and McLennan, 1995; Wedepohl, 1995). Although it has been demonstrated that the composition of detrital thorium contamination can depart from bulk crustal values at different study sites (Cobb et al., 2003; Shen et al., 2008), our approach nonetheless offers a first-order estimate that should approximate the degree of contamination. Additionally, we accepted samples that do not have detrital Th information reported in cases where rejecting these samples would have effectively removed the study site from the dataset. Cases where this has been done are noted explicitly in the site summaries.

Finally, we expanded the upper limit of the δ234Ui threshold by 2 ‰, so that the new range of acceptable δ234Ui values is 140 ‰–152 ‰, provided that the newly accepted ages are stratigraphically consistent with the other ages from the site. This was done, in part, because the average δ234Usw value for the LIG is not constrained and there is evidence that the uranium isotopic composition of seawater has varied by several per mil on glacial–interglacial timescales (Chen et al., 2016; Chutcharavan et al., 2018). More importantly, it is also clear that there are likely subtle, unresolved biases in interlaboratory calibration protocols that could result in systematic offsets of a few per mil, depending on the lab where a sample was dated (Chutcharavan et al., 2018, and references therein).

The purpose of these screening protocols is, specifically, to identify the highest-quality closed-system fossil coral U-series ages that can be used to provide constraints on sea-level change within the LIG (i.e., on suborbital/millennial timescales). We acknowledged that some users may only be interested in knowing whether a geologic feature is broadly constrained to the LIG by the fossil coral U-series data, and we have endeavored to make such distinctions where applicable in the site descriptions (see Sect. 3). Users are also cautioned that the screening protocols provided in this paper are only intended as guidelines to assist users with identifying coral U-series ages that display closed-system behavior. A U-series measurement fitting a set of predetermined geochemical parameters does not automatically imply that an age is robust or that it can provide meaningful radiometric age constraints on LIG sea-level change. Therefore, it is important for the user to carefully evaluate whether a screened age is consistent with the available geologic context. Additionally, the two example screening protocols provided here are by no means the only way to screen fossil coral U-series data, and we have included a functionality within the WALIS U-series database to upload alternative screening interpretations.

In general, site-specific fossil reef RSL histories for the LIG diverge from GMSL because of processes such as regional tectonics, glacial isostatic adjustment (GIA) and dynamic topography (Broecker et al., 1968; Farrell and Clark, 1976; Mitrovica and Milne, 2003; McMurtry et al., 2010; Austermann et al., 2017). Although correcting fossil coral RSL records for these processes was not the main focus of this study, it is nonetheless important for a user to be cognizant of this complication when comparing sea-level records from different sites. It is also worth keeping in mind that, although all three factors affect the uncertainty in the absolute elevation for coral-derived RSL reconstructions, the relative contribution of each varies from site to site.

Corals were U-series-dated from emergent LIG reef deposits on Great Inagua, San Salvador and the Abaco Islands in The Bahamas (Chen et al., 1991; Hearty et al., 2007; Thompson et al., 2011). A total of 200 U-series ages from 142 unique coral specimens were reported, with 29 of these corals dated at least in duplicate. In total, the original study authors accepted 49 U-series ages from 37 coral samples as closed-system ages. Thompson et al. (2011) did not use closed-system ages and instead applied an open-system correction to each sample. Under the strict screening protocol, 35 U-series ages from 26 coral samples were accepted. This number increased to 43 U-series ages from 29 corals when the flexible screening protocol was applied. Sample ages that passed flexible screening ranged from 131.3 ± 1.4 kyr for CH91-002 to a weighted mean age of 119.8 ± 0.3 kyr for TH11-023 (weighted mean ages are reported where multiple subsamples of the same coral passed the screening criteria).

Several site-specific adjustments were made under the flexible screening protocol. First, the 232Th concentrations for Chen et al. (1991) were recalculated using the published 230Th / 232Th activity ratios from their supplement, as in certain cases only one 232Th concentration was reported for multiple subsamples of the same coral. Second, we only considered samples that were dated at least in duplicate from Thompson et al. (2011), as calcite content was not reported in that study and there are no elevation data from which stratigraphic relationships can be derived. Finally, we accepted ages from sample CH91-023 as closed-system since the ages were reproducible and calcite content was on the cutoff threshold at 2 %.

Assessing whether corals from The Bahamas dataset were in primary growth position is challenging. Chen et al. (1991) applied the term “in situ” to describe both growth position corals that are part of the reef framework and detrital coral rubble that had been cemented in place. For the present compilation, we categorized all corals derived from rubble layers as “not in primary growth position”. Using this approach, a total of 14 ages from Chen et al. (1991), derived from 11 coral specimens, can be treated as RSL indicators under the flexible screening protocol. Previous workers assigned a paleowater depth range of 3 to 4 m for the Cockburn Town and Devil's Point sites (Chen et al., 1991). A more recent study, however, reevaluated the vertical position and facies characteristics of the two fossil coral reefs using high-precision surveying techniques and published new paleowater depth interpretations (Skrivanek et al., 2018). It is difficult to compare the present dataset to the reef zones described in Skrivanek et al. (2018), as Chen et al. (1991) did not distinguish between reef units in their study. All of the corals in primary growth position, however, were colonies of Pseudodiploria clivosa or Orbicella annularis, which were found in units with interpreted paleowater depths of 0.2–5 m at Devil's Point reef and 0.2–3 m at Cockburn Town (Skrivanek et al., 2018).

Thompson et al. (2011) distinguished corals that were derived from a rubble layer from those collected from the in situ reef framework but gave no elevation information associated with each sample, so none of the samples are used as RSL indicators here. Though elevation estimates were provided in Thompson et al. (2011) for each reef unit, these elevations do not always match those subsequently surveyed at the same sites, calling into question the use of those approximate elevations (Skrivanek et al., 2018). Primary-growth-position corals can, however, still be used to constrain the maximum age of each fossil reef, even without published elevation data. A total of five corals (11 analyses in total) from Thompson et al. (2011) are in primary growth position and passed the flexible screening criteria. These ages range from 127.3 ± 0.6 to 119.8 ± 0.3 kyr for the Devil's Point reef and 125.2 ± 1.5 to 122.2 ± 1.7 kyr for the Cockburn Town reef.

U-series coral ages were reported for three locations along the Pacific coast of Baja California, Mexico (Muhs et al., 2002a). Corals collected for that study came from detrital sedimentary deposits on marine terraces and, therefore, were not in primary growth position and cannot be used as strictly reliable RSL indicators. Instead, the study authors used coral U-series ages as a constraint on the maximum age of the terraces. In total, 26 corals were dated, and the study authors accepted 16 of the U-series ages. None of these ages passed the strict or flexible closed-system criteria.

Barbados is one of the most intensely studied LIG fossil reef localities in the world, with 141 U-series analyses reported for 107 corals from 11 separate studies (Bard et al., 1990; Blanchon and Eisenhauer, 2001; Cutler et al., 2003; Edwards et al., 1997, 1987b; Gallup et al., 1994, 2002; Hamelin et al., 1991; Muhs and Simmons, 2017; Speed and Cheng, 2004; Thompson et al., 2003). The island is located on an accretionary prism and has experienced differential uplift since the LIG. Local uplift rates vary from ∼ 0.2 m/kyr in the north and south of Barbados to as high as ∼ 0.5 m/kyr near the Clermont Nose/University of the West Indies transect near the middle of the island (e.g., Muhs and Simmons, 2017; Taylor and Mann, 1991), so care must be taken when applying subsidence corrections to the Barbados dataset. Additionally, the dataset can be challenging to interpret, as there are multiple names for some localities, and some coral samples have been dated in two or more studies. To facilitate data accessibility, we standardized site location names (e.g., all LIG samples from the Clermont Nose area were given the site name “Univ. West Indies (UWI) Transect”), and we endeavored to link U-series measurements across multiple studies that were derived from the same coral colony.

Of the U-series ages reported, the original study authors accepted 40 ages from 33 unique coral specimens. It should be noted that Thompson et al. (2003) did not apply closed-system screening criteria; rather, an open-system model was used. Under the strict screening protocol, a total of 24 U-series ages were accepted from 17 corals, whereas the flexible protocol accepted 41 ages from 28 corals. Ages from the flexible screening protocol range from 103.8 ± 1.0 kyr (BL01-001-001) to 172.5 ± 1.4 kyr (GA02-006-001). The oldest age was sampled from “Lazaretto unit”, which is part of the LIG Rendezvous Hill terrace, but this unit is actually MIS 6 in age (Speed and Cheng, 2004). It should be noted that one sample which passed both screening protocols, GA02-032-001 (136.4 ± 0.9 kyr), was rejected, as this age was eventually retracted by the Gallup et al. (2002) study authors after multiple dated subsamples from the same colony were unable to reproduce the reported age. Several of the corals were also dated using Pa–Th methods (Cutler et al., 2003; Edwards et al., 1997; Gallup et al., 2002).

Of the samples that passed flexible screening, a total of eight U-series ages (six corals in total) came from corals that were stated as being in primary growth position, with ages ranging from 103.8 ± 1.0 kyr (BL01-001) to 172.5 ± 1.4 kyr (GA02-006). The number of RSL data points is increased to 29 ages from 21 corals by including all samples that were not explicitly identified as transported clasts or cobbles. One primary-growth-position coral with an age of 129.6 ± 0.8 kyr (GA02-014) was treated as marine limiting as no coral taxonomic information was provided. In cases where primary-growth-position corals are derived from a reef crest facies, we assigned a paleowater depth of < 5 m, which is the typical depth range for modern Caribbean reef crest environments (Lighty et al., 1982). In all other cases, we applied the taxon-derived modern depth distributions (OBIS, 2014).

LIG-aged corals are present at Grape Bay on the southern side of Bermuda (Ludwig et al., 1996; Muhs et al., 2002b). These deposits are inferred to be originally derived from patch reefs but are not in primary growth position and may have been storm derived (Muhs et al., 2002b). Therefore, the ages presented in these studies represent a constraint on the maximum age of the rubble deposit but cannot be used as RSL indicators. In total, nine corals were dated from Grape Bay, and the authors originally accepted seven of the ages. Only two corals pass the strict closed-system criteria: MU02-019, with an age of 116.9 ± 0.9 kyr, and MU02-020, with an age of 113.7 ± 0.9 kyr. A third coral, MU02-026 (118.7 ± 0.9 kyr), is also accepted once the flexible protocol has been applied.

Several studies reported U-series coral ages from marine deposits along the southwest coast of California and several of the Channel Islands (Muhs et al., 2012a, 2006, 2002b; see global review of MIS 5a/c sea-level records by Thompson and Creveling (2021) for an overview of MIS 5 marine terraces along the California coast). These samples are solitary Balanophyllia elegans corals from detrital sedimentary deposits on marine terraces and are therefore not in primary growth position and cannot be used as RSL indicators. Instead, the study authors used the coral U-series ages as a constraint on the maximum age of terrace formation. A total of 153 U-series ages were reported for 148 unique coral specimens, with four of the corals dated in duplicate. The study authors accepted 34 of the ages (32 corals in total). None of the ages passed the strict closed-system criteria, but four ages from three corals were accepted under the flexible protocol: MH02-075 (118.3 ± 0.6 kyr), MH02-077 (119.9 ± 0.8 kyr) and MH06-013 (weighted mean age of 118.8 ± 0.7 kyr).

Two LIG ages are reported from Gran Canaria and Lanzarote in the Canary Islands (Muhs et al., 2014). Both corals were Siderastrea radians fragments collected from marine deposits and were used to determine the maximum age of the deposits and constrain local uplift rates. The authors accepted both ages, assigned a 0.017–0.050 m/kyr uplift rate for the Gran Canaria site and determined that the Lanzarote site had not been subjected to significant uplift since the LIG. Neither age passed the strict or the flexible screening protocols, but they do broadly constrain the age of their respective deposits to the LIG.

Zazo et al. (2007) reported U-series coral ages that were used to constrain the age of marine terrace conglomerates on Sal Island, Cabo Verde. In total, 10 U-series ages were reported for five corals (and one hydrozoan), with one coral sample (ZA07-004) analyzed five times. All coral ages were accepted by the study authors, but only one age from coral ZA07-004 (129.5 ± 4.0 kyr) passed the flexible protocol. This coral is not in primary growth position and cannot be used to constrain RSL.

U-series ages have been reported for multiple outcrops of the LIG Hato unit on the island of Curaçao, for a total of 25 ages from 15 unique coral colonies (Hamelin et al., 1991; Muhs et al., 2012b; for a regional overview, see Rubio-Sandoval et al., 2021). Curaçao is slowly uplifting, with an estimated uplift rate of 0.026 to 0.054 m/kyr, based on the “highest inner edge” elevation of the Hato unit at 12.4 m (Muhs et al., 2012b). In total, the study authors accepted five U-series ages from four unique coral specimens. Under the strict screening protocol, this is reduced to a single age of 118.8 ± 0.8 kyr from sample SC78-005-002. The flexible protocol adds three additional ages: two from MU12-001, with a weighted mean age of 126.6 ± 0.7 kyr, and an age of 118.7 ± 1.2 kyr from coral SC78-002-002.

All samples which passed the flexible screening criteria were in primary growth position. Based on the paleoenvironmental interpretations of Muhs et al. (2012b), samples SC78-005 and MU12-001 were part of an Acropora palmata-dominated reef crest facies growing in 0–5 m water depth, which we adopted for this study. Paleoenvironmental interpretations and stratigraphic context were not provided for sample SC78-002. Therefore, the modern depth distribution for Diploria spp. was applied in this case (OBIS, 2014).

Fossil corals of LIG age were reported for the Abdur Reef Limestone on the Red Sea coast of Eritrea (Walter et al., 2000). The Eritrean coast is tectonically active and is estimated to be uplifting by 0.06 m/kyr based on the elevation of the LIG reef deposits (Hibbert et al., 2016). In total, seven U-series ages were reported for six corals, with one coral dated in duplicate. In the original study, the authors accepted all ages, except for one from coral WA00-006, which had anomalously low U content and an age that was older than expected. None of the ages passed the strict or the flexible closed-system screening criteria.

Fossil corals have been dated from multiple sites across the Florida Keys (Fruijtier et al., 2000; Muhs et al., 2011; Multer et al., 2002). In total, 55 U-series ages were reported for 51 unique coral samples, with four corals dated in duplicate. In total, the study authors accepted 15 of the ages from 13 coral samples. Under the strict screening criteria, four ages were accepted from three unique coral specimens from Windley Key: MU11-026, with a weighted mean age of 115.1 ± 0.6 kyr; MU11-034, with an age of 114.1 ± 0.6 kyr; and MU11-037, with an age of 120.4 ± 0.8 kyr. Using the flexible screening protocol, the total number of analyses accepted increases to 13, from a total of 10 unique coral specimens ranging from a weighted-mean age of 123.0 ± 0.6 kyr (MU11-012) to a weighted-mean age of 113.7 ± 0.6 kyr (MU11-034). One of these samples, MU11-005, was accepted despite failing the strict 232Th criterion, as it was only marginally higher (2.4 ppb) and passed both the calcite and δ234Ui thresholds. It should be noted that samples from Muhs et al. (2011) and Multer et al. (2002) appeared to have a 5 % limit of quantification for their XRD techniques, so all samples from these studies were interpreted as having acceptable calcite content.

All 10 corals that passed the flexible screening protocol were in primary growth position and, therefore, can be used as RSL indicators. These samples were collected from outcrops of the Key Largo Limestone at Windley Key and Key Largo, with sample elevations ranging from 2–5 m a.m.s.l. (Muhs et al., 2011). The dominant coral taxa in the outcrops studied at both localities were massive Orbicella annularis and Pseudodiploria strigosa, with the Windley Key site also containing Colpophyllia natans. Several estimates of paleowater depth for the Key Largo Limestone have been published and range from < 3 m to as much as 12 m water depth (Fruijtier et al., 2000; Muhs et al., 2011; Perkins, 1977; Stanley, 1966). Most recently, Muhs et al. (2011) interpreted this facies as having a minimum water depth of 3 m based on the optimal depth range for these three coral species from a modern ecological survey of reefs in the Florida Keys and Dry Tortugas (Shinn et al., 1989).

We adopted the 3 m estimate of Muhs et al. (2011) as the most probable paleowater depth for the LIG deposits at Key Largo and Windley Key and further parameterized the possible range of paleowater depths. As stated by Muhs et al. (2011), the optimum water depths for Pseudodiploria strigosa and Copophyllia natans are 3–10 and 2–10 m, respectively, whereas the optimal depth range for Orbicella annularis is substantially wider, at 3–45 m (Shinn et al., 1989). However, the modern Montastrea annularis distribution has a median depth closer to 10 m (upper 95 % confidence interval =17 m water depth; OBIS, 2014). Given this additional information, we parameterized the paleowater depth uncertainty for these samples as 3 +7/-0 m.

A regional overview of LIG sea-level records for French Polynesia is provided by Hallmann et al. (2020). Many of the islands and archipelagos in French Polynesia are former hot-spot volcanoes that are subsiding because of volcanic loading. In these cases, LIG deposits are often located below sea level and can only be accessed via scientific drilling. U-series ages from corals have been published from two locations in French Polynesia: Mururoa atoll in the Tuamotu Archipelago and offshore drilling at Tahiti during IODP Expedition 310 (Camoin et al., 2001; Thomas et al., 2009). Both Mururoa Atoll and Tahiti have been subject to subsidence since at least the Late Pleistocene. The subsidence rate at Mururoa was estimated to be ∼ 0.07–0.08 m/kyr using K–Ar dating of the volcanic basement and the location of the LIG unit, which was 3 m below the modern reef (Trichet et al., 1984; Camoin et al., 2001). At Tahiti, the subsidence rate is an order of magnitude greater and is commonly estimated to be 0.25 m/kyr with a total possible range of 0.2–0.4 m/kyr (Le Roy, 1994; Bard et al., 1996; Thomas et al., 2012). To date, no corals of LIG age have been discovered at Tahiti, but several corals from the IODP record have been dated to late MIS 6 (Thomas et al., 2009).

The existing dataset for French Polynesia contains 19 U-series dates from 12 corals, with five corals from Tahiti dated in duplicate and one in triplicate. Of the six corals analyzed from Mururoa, only one age (CA01-007) was accepted by the study authors. This sample also passed the strict screening protocol and has a recalculated age of 126.0 ± 2.2 kyr. For the samples from Tahiti, two corals (TH09-001 and TH09-003) were accepted by the study authors. Based on the strict screening protocol, dates from two corals and four unique U-series measurements passed screening: TH09-003, with a weighted-mean age of 133.9 ± 0.4 kyr, and TH09-008, with a single age of 134.0 ± 0.4 kyr. By employing the flexible screening protocol, four additional U-series ages can be included from two corals: TH09-001 (weighted mean age of 138.0 ± 0.5 kyr) and TH09-005 (weighted mean age of 137.0 ± 0.5 kyr).

Sample CA01-007 from Mururoa was reported as being “reworked” (Camoin et al., 2001). Samples TH09-001, TH09-003, TH09-005 and TH09-008 are both interpreted as being in growth position and thus can be used as RSL indicators. TH09-003 and TH09-008 are both massive Porites spp., which is commonly associated with depth ranges of 0–25 m, while TH09-001 and TH09-005 were associated with a shallower facies interpreted as growing in 0–6 m water depth (Thomas et al., 2009).

Fossil corals have been dated from multiple localities across Grand Cayman (Coyne et al., 2007; Vezina et al., 1999). In total, 15 corals from the LIG and late MIS 6 were dated, with the authors accepting all but three ages. All of the ages were rejected by the strict and flexible closed-system screening protocols.

Two studies have reported LIG U-series ages from the Great Barrier Reef, which were collected via scientific drilling on modern reef flats (Braithwaite et al., 2004; Dechnik et al., 2017). A total of 40 ages from 14 unique coral specimens were reported, and the authors originally accepted 11 of the ages from five corals. All but one of these corals (BR04-001) were dated in triplicate. Under the strict screening protocol, three ages were accepted from two coral samples (DE17-001, DE17-003). When the flexible protocol was applied, the total number of accepted ages expanded to include seven ages from three corals: DE17-001, with a weighted mean age of 128.7 ± 0.7 kyr; a single age from DE17-003, which was dated to 126.1 ± 0.5 kyr; and DE17-004, with a weighted mean age of 127.7 ± 0.5 kyr. Coral DE17-004 had higher calcite content (6.5 %) but was nevertheless accepted because the ages were stratigraphically consistent with DE17-001/003 and the sample passed the 232Th and δ234Ui thresholds.

All three corals that passed the flexible screening protocol were in primary growth position and can be used as RSL indicators. Using the coralgal assemblage interpretations, Dechnik et al. (2017) assigned these samples a 0–6 m paleowater depth range, which we also adopted.

One study reported corals with LIG ages from uplifted terraces on the Perachora Peninsula, Greece, which were originally used to constrain local uplift rates (Dia et al., 1997). The authors reported that the two corals with LIG ages (DI97-002, DI97-003) showed signs of open-system behavior, based on uranium isotopes (δ234Ui > 200 ‰), high detrital 232Th concentrations (∼ 7 ppb) and anomalously low 87Sr / 86Sr ratios. DI97-002 and DI97-003 passed neither the strict nor the flexible closed-system screening criteria.

Three studies reported U-series ages on corals from uplifted terraces along the Gulf of Aqaba (Bar et al., 2018; Manaa et al., 2016; Yehudai et al., 2017). Bar et al. (2018) inferred an uplift rate of 0.13 m/kyr for the northeastern Gulf of Aqaba based on the present-day elevation of the coral terraces and the timing of diagenesis for altered fossil corals dated by Yehudai et al. (2017). In total, 67 U-series ages were reported for 18 unique coral specimens, with the majority of samples dated in triplicate or greater. In total, the study authors accepted six of these ages from four coral samples. Under both the strict and flexible closed-system screening criteria, only two samples were accepted from the upper Haql terrace: MA16-003, with an age of 119.7 ± 0.5 kyr, and MA16-004, with an age of 120.2 ± 0.6 kyr.

It is unclear which coral samples were collected in primary growth position, so MA16-003 and MA16-004 were not treated as RSL indicators within the database.

Bard et al. (1990) reported U-series ages from the northwest coast of Haiti. In total, three analyses were conducted on two unique coral specimens, with one of the corals dated in duplicate. All three ages were accepted by the study authors, but detrital Th content was not reported, which meant we were unable to accept the ages based on the strict screening protocol. Under the flexible protocol, all three ages were accepted from two corals: BA90-021 (122.8 ± 1.1 kyr) and BA90-022 (weighted mean age of 125.3 ± 1.4 kyr). Both samples were identified as Acropora palmata corals that were part of a reef crest facies, which typically grows in < 5 m water depth (Lighty et al., 1982). The northeastern coastline of Haiti is tectonically active, and it is estimated that the uplift rate is approximately 0.36 m/kyr based on the elevation of the local LIG terrace reported by Dodge et al. (1983).

Several studies have published coral U-series ages from the Waimanalo Formation on the Hawaiian island of Oahu, which dates to the LIG (Hearty et al., 2007; McMurtry et al., 2010; Muhs et al., 2002b; Szabo et al., 1994; for a regional overview, see Hallmann et al., 2020). Oahu is slowly uplifting at a rate of ∼ 0.06 m/kyr because it is located on the peripheral bulge of the island of Hawaii, which is subsiding as a consequence volcanic loading from Hawaiian hot-spot volcanism (McMurtry et al., 2010, and references therein; Szabo et al., 1994). A total of 82 U-series analyses were published for Oahu from 72 unique coral specimens. Eight of the samples have been dated in duplicate and one in triplicate.

In the original studies, the authors accepted 59 ages from 52 coral samples, whereas under the strict screening protocol we accepted 25 U-series ages from 23 coral samples. Using the flexible screening protocol, this number increased to 34 ages from 29 coral samples, with ages ranging from 110.84 ± 3.9 kyr (SZ94-007) to 133.0 ± 3.3 kyr (SZ94-021). The 232Th concentrations for Szabo et al. (1994) were recalculated using the published 230Th / 232Th activity ratios. Additionally, we interpreted the limit of quantification for the XRD measurements in Szabo et al. (1994) to be 5 % for calcite, which led to two additional ages (SZ94-002-001, SZ94-016-001) being accepted. Many of the dated samples from Oahu were either clasts or collected from marine conglomerates, so the number of samples that can be used as RSL indicators is substantially smaller than the total number of corals that passed screening. Under the flexible protocol, a total of nine corals can be treated as RSL indicators, with ages ranging from 110.8 ± 3.9 kyr (SZ94-007) to 126.5 ± 0.7 kyr (MU02-055). The only constraint on paleowater depth was given in Szabo et al. (1994), in which the authors noted that typical water depths for Pacific Pocillopora and Porites are between 1 and 27 m based on a previous synthesis paper (Wells, 1954). Both Szabo et al. (1994) and Muhs et al. (2002b), however, ultimately treated primary-growth-position corals as marine limiting, with a minimum water depth of 1 m. Without any additional constraints on water depth, we applied the taxon-specific modern coral depth distributions (OBIS, 2014).

Coral U-series ages were reported for two locations in Indonesia: Sumba Island and Southeast Sulawesi (Bard et al., 1996; Pedoja et al., 2018; see Maxwell et al., 2021, for a regional overview). Both locations are tectonically active, with uplift rates for Sumba Island and Southeast Sulawesi estimated by the study authors to be 0.2–0.5 and 0.12–0.29 m/kyr, respectively. Between these two study sites, a total of 21 corals were dated, and 14 of these ages were accepted by the study authors. Based on the strict closed-system criteria, four of the ages from Pedoja et al. (2018) were accepted: PE18-001, with an age of 133.7 ± 3.0 kyr; PE18-002, with an age of 131.2 ± 3.0 kyr; PE18-005, with an age of 112.8 ± 3.0 kyr; and PE18-008, with an age of 127.8 ± 2.0 kyr. Under the flexible criteria, six additional ages were accepted from Bard et al. (1996), ranging from 86.9 ± 0.6 kyr (BR96-008) to 133.1 ± 1.0 kyr (BR96-012).

Although sample elevations for Pedoja et al. (2018) were reported, the elevation uncertainty is large (± 10 m), and the authors did not provide facies information or state whether the corals were in primary growth position. All six Bard et al. (1996) ages were identified as primary-growth-position corals. Sample BR96-016 was originally interpreted as growing in 5–15 m water depth, so we have used the midpoint of this range as the assigned paleowater depth uncertainty (i.e., 10 +5/-5 m). The other five ages were not associated with facies/paleowater depth interpretations, so the modern taxa depth distributions were assigned (OBIS, 2014). However, it should be noted that sample BR96-017 was identified as a Porites microatoll, which implies that the colony was likely growing within the subtidal/intertidal zone.

One study reported U-series ages from corals cored on Amédée Islet, New Caledonia (Frank et al., 2006; for regional review, see Hallmann et al., 2020). In total, 19 analyses were reported for 15 corals, with one coral dated five times. The study authors used open-system ages (Thompson et al., 2003; Villemant and Feuillet, 2003) to confirm the existence of an LIG reef deposit within the core records and estimated a subsidence rate of 0.16 m/kyr for the study site. The strict and flexible closed-system criteria rejected all of the ages from New Caledonia.

A single coral with a late MIS 6 age (KE12-001; 133.5 ± 1.0 kyr) was reported for the South Pacific island of Niue and has been interpreted as a 2 m high Porites colony that infilled a karstic channel (Kennedy et al., 2012; for a regional overview, see Hallmann et al., 2020). The authors interpreted this deposit as being LIG in age. This U-series age, however, has an anomalously low initial uranium isotopic value (δ234Ui=121.7 ± 3.3 ‰) and fails both the strict and flexible closed-system screening criteria.

LIG fossil coral U-series ages are available from uplifted coral reef terraces on the Huon Peninsula, Papua New Guinea (Cutler et al., 2003; Esat et al., 1999; Stein et al., 1993; for a regional overview, see Hallmann et al., 2020). The region has experienced substantial uplift since the LIG, with local uplift rates estimated to be ∼ 2 m/kyr. As a result, the LIG fossil reef deposits are presently located ∼ 140–230 m a.m.s.l. A total of 47 analyses were reported from Huon Peninsula fossil reefs from 34 unique coral specimens. One coral was dated in duplicate, and five corals have U-series ages from multiple subsamples. Of the 47 U-series analyses performed, the study authors accepted 11 ages from seven coral samples (although the actual number of ages accepted is likely higher, as Esat et al. (1999) did not specify the acceptable δ234Ui thresholds used in their study). Under the strict screening protocol, 11 ages from five coral samples (CU03-011, ST93-005, ST93-006, ST93-007, ST93-009) were accepted, whereas 13 ages from seven samples (CU03-011, CU03-023, ES99-020, ST93-005, ST93-006, ST93-007, ST93-009) were accepted under the flexible protocol. The ages that pass the flexible screening protocol range from a weighted mean age of 115.2 ± 0.7 kyr for CU03-011 to 136.8 ± 1.8 kyr for a single analysis from ST93-006.

Corals CU03-011, CU03-023 and ES99-020 do not have any contextual information that can be used to determine if they are in primary growth position, so we did not treat these samples as RSL indicators. ST93-005, ST93-006, ST93-007 and ST93-009 are in primary growth position, but there are no published paleowater depth interpretations provided. Additionally, there are insufficient modern observations for Gardineroseris planulata to produce a robust modern depth distribution for these samples, so samples ST93-005 through ST93-007 can only be treated as marine limiting (Hibbert et al., 2016; OBIS, 2014). The final sample, ST93-009, is a colony of Porites lutea, which has a modern depth range of 0 +0/-45 m (OBIS, 2014).

One study published coral U-series ages for emergent coral reef terraces along the Red Sea coast of Saudi Arabia (Manaa et al., 2016). In total, study authors reported U-series ages for 25 coral samples and accepted 17 of the ages. Using the strict closed-system screening criteria, that number is reduced to three samples collected from reef terraces near the port city of Yanbu. Two of the samples (MA16-009 and MA16-010) were collected from the lower Yanbu terrace and yielded ages of 42.2 ± 0.1 and 51.4 ± 0.1 kyr, respectively, but were rejected because the ages were not stratigraphically consistent with the rest of the unit (Manaa et al., 2016). The remaining age, from the upper Yanbu terrace (MA16-013), yielded an LIG age of 127.9 ± 0.4 kyr. A second sample from the upper Yanbu terrace (MA16-012) was also accepted under the flexible screening protocol, yielding an age of 112.6 ± 0.3 kyr. The authors did not state whether these samples were in primary growth position, so these samples were not treated as RSL indicators.

Two studies from the Seychelles published U-series coral ages (Dutton et al., 2015b; Israelson and Wohlfarth, 1999; for a regional overview, see Boyden et al., 2021), containing a total of 67 U-series measurements for 31 individual coral specimens. Approximately half (15) of the corals were dated in triplicate, with three corals measured in duplicate. In the original studies, 24 of the corals yielded acceptable ages (38 unique U-series ages). Under the strict screening protocol, only five unique U-series ages from three corals are accepted. This is increased to 25 U-series ages from 14 corals once the flexible screening criteria are applied, with ages ranging from a weighted mean age of 122.2 ± 0.5 kyr from sample DU15-017 to 129.1 ± 1.6 kyr from sample IS99-007.

Multiple sources have interpreted the Seychelles deposits as having formed in an intertidal to upper subtidal zone, which, based on modern analogues from these same islands, results in a maximum water depth of 2 m (Dutton et al., 2015b; Israelson and Wohlfarth, 1999; Montaggioni and Hoang, 1988). Here, the maximum paleowater depth of 2 m was adopted, with one exception. Sample IS99-010, a Porites sp., was not explicitly tied to the subtidal facies. Therefore, we assigned IS99-010 a water depth uncertainty based on the modern depth distribution for Porites spp. (4 +62/-4 m; OBIS, 2014). Of the 14 samples that met the flexible closed-system screening criteria, all except DU15-017 and DU15-019 were identified as being in primary growth position.

LIG coral ages were reported for subtidal deposits on the Yorke Peninsula near Adelaide, South Australia (Pan et al., 2018). Four U-series ages were reported by the authors from four unique specimens of the solitary coral Plesiastrea versipora. One of the ages (PA18-002) was originally accepted by the authors, but none of the four samples met the strict or the flexible closed-system screening criteria. The Plesiastrea versipora were not in primary growth position, and these deposits were interpreted by Pan et al. (2018) as being wave and/or storm derived.

In the US Virgin Islands, sediment cores from Holocene reefs off the island of St. Croix possess LIG reef deposits in the underlying substrate (Toscano et al., 2012). St. Croix is unique among many other Caribbean LIG sites in that all but one of the LIG fossil reef localities are presently submerged below modern sea level, and there is an apparent 0.62 m/km gradient between LIG deposits from the northwestern and northeastern ends of the island. The authors interpreted this gradient as having resulted from differential subsidence or tilting caused by regional tectonism (Toscano et al., 2012, and references therein). Six corals from this study yielded LIG ages, five from drill cores on Tague Reef on the northeastern end of St. Croix and one from a drill core farther west, on Long Reef. Toscano et al. (2012) accepted all ages except the one from Long Reef (TO12-010). The strict closed-system criteria yielded similar results, but we rejected coral TO12-008 from Tague Reef because of an elevated δ234Ui value. The four corals that passed the strict screening criteria are TO12-005, with an age of 115.1 ± 0.3 kyr; TO12-006, with an age of 124.6 ± 0.3 kyr; TO12-007, with an age of 123.44 ± 0.4 kyr; and TO12-009, with an age of 129.4 ± 0.4 kyr. An additional sample, TO12-008, can also be included under the flexible protocol, yielding an age of 125.7 ± 0.3 kyr.

In Toscano et al. (2012), corals that were not in primary growth position were listed as “fragments”. Since none of the corals that passed geochemical screening were stated as being fragments in the Toscano et al. (2012) supplement, we have treated them as being in primary growth position. The authors interpreted the Tague Reef LIG deposit as being part of a reef flat/back-reef setting in < 5 m water depth, so an interpreted paleowater depth range of 5 +0/-5 m was adopted. Based on the U-series dating, Toscano et al. (2012) estimated subsidence rates of 0.08 m/kyr for Tague Reef and 0.07 m/kyr for the Long Reef site.

U-series ages have been reported for 19 corals collected from the LIG Boat Cove and South Reef units on West Caicos (Kindler and Meyer, 2012; Kerans et al., 2019). Additional mass spectrometric U-series ages have been reported by Simo et al. (2008), but the type of material dated ranged from “well preserved coral, to skeletal grains and ooids”. As Simo et al. (2008) did not specify which carbonate material was dated for each of their ages, their dataset was not included in the present compilation.

Of all the ages reported, a total of 13 were accepted by the study authors. Under both the strict and flexible screening criteria, the two ages from Kindler and Meyer (2012) were rejected due to high calcite content. For the Kerans et al. (2019) study, calcite content was not reported for each sample, but the authors stated that XRD measurements indicated “a range of calcitization 100 % calcite to 3 % calcite”, so the authors instead used Sr element mapping to identify the best-preserved sections of coral to date. Based on the range of calcite concentrations given, none of the samples would pass the strict/flexible protocols, so these samples were also rejected.

Edwards et al. (1987b) reported LIG U-series ages for uplifted coral terraces on Efate Island, Vanuatu. In total, there are three U-series ages for two corals, with one coral (ED87-010) measured in duplicate. All three ages were accepted by the study authors and also passed the strict and flexible closed-system criteria. ED87-010 has a weighted mean age of 130.6 ± 1.1 kyr, which constrains the age of the lower Efate terrace, and ED87-011 has an age of 126.5 ± 1.4 kyr, which constrains the age of the upper terrace. No elevation information was reported for these samples, so they cannot be used as RSL indicators. For a regional overview, the reader is directed to Hallman et al. (2020).

Fossil corals from the coastline of Western Australia represent perhaps the broadest geographic region reported here, spanning more than 1400 km from Cape Range in the north to Foul Bay near the southwestern tip of the Australian continent. It has one of the largest number of U-series ages of any study area covered by WALIS, with 176 U-series ages reported for 156 unique coral specimens (Collins et al., 2003; Eisenhauer et al., 1996; Hearty et al., 2007; O'Leary et al., 2008a, b, 2013; McCulloch and Mortimer, 2008; Stirling et al., 1995, 1998). These sites are considered to be tectonically stable, with one notable exception being sites near the Cape Cuvier anticline, for which there is strong evidence of neotectonism since the LIG (Whitney and Hengesh, 2015).

In the original studies, at least 59 of the U-series ages from 55 coral were accepted. Under the strict screening protocol, the total number of accepted ages dropped substantially to just nine, from five corals, largely because calcite content was not reported in many of the studies. This resulted in 75 % of the dataset being summarily rejected without any assessment of age quality. To remedy this, for the flexible screening protocol, we allowed samples without reported calcite content to be screened using 232Th and δ234Ui. This is a reasonable judgment call to make, as in previous studies by the same authors only corals with calcite contents below detection limit were dated (e.g., Stirling et al., 1995). Under the flexible screening protocol, we also accepted ZH93-001-001, which had slightly elevated 232Th concentration (∼ 3 ppb), and both samples from SI96-002, as they fell along the 129 kyr, closed-system isochron. Samples that passed flexible screening were comparable to those that were accepted in the original publications, with 61 ages accepted from 56 unique coral samples. Ages ranged from 116.3 ± 0.3 kyr (ST98-012) to 134.3 ± 1.9 kyr for sample EI96-006.

Of the samples that passed the flexible screening protocol, 42 were in primary growth position (45 analyses total) and can be used as RSL indicators. Explicit paleowater depth interpretations were not provided in most cases, so we assigned the modern coral depth distributions for the relevant taxa (OBIS, 2014). It should be noted, however, that many of the samples in the dataset were collected from the very top of the LIG reef outcrops and were likely growing within a few meters of sea level. For samples from Shark Bay (OL08-002, OL08-003, OL08-009, OL08-010), no coral taxonomic information was provided, so these samples should be treated as marine limiting. Additionally, two samples from the Houtman Abrolhos islands (ZH93-001, ZH93-005) were interpreted as being intertidal or subtidal deposits and can be constrained to < 2 m paleowater depth (Zhu et al., 1993).

Fossil coral U-series ages were reported for emergent reef terraces in Yemen, along the Al-Hajaja coast and on Perim Island (Al-Mikhlafi et al., 2018). In total, 35 U-series ages were reported for 33 coral specimens, with two corals dated in duplicate. Al-Mikhlafi et al. (2018) concluded that terrace Tr3 from the study area was LIG in age but decided against using any of the samples collected as RSL indicators, as most of the corals were diagenetically altered. None of the U-series ages from Yemen met the strict or flexible closed-system criteria.

Two reef tracts were uncovered during the excavation and construction of the Xcaret theme park near Playa del Carmen, Mexico (see Simms, 2021, for a regional overview). U-series ages are available from both reef tracts, which included Acropora palmata and Siderastrea siderea corals (Blanchon et al., 2009). In total, 33 U-series ages were reported from 10 unique coral specimens, with each coral dated at least in triplicate. In total, the study authors accepted seven analyses from five corals. Both the strict and flexible screening criteria rejected all but three analyses, primarily because multiple subsamples that were dated for the rejected coral specimens did not yield reproducible U-series ages. The three ages that passed screening were both from upper reef tract sample BL09-006, giving a weighted mean age of 123.9 ± 0.7 kyr. This sample, however, was identified as a clast by Blanchon et al. (2009) and therefore cannot be used as an RSL indicator.

Multiple studies published coral U-series ages > 150 kyr, suggesting the corals grew during previous glacial–interglacial cycles (e.g., Andersen et al., 2008, 2010b; Bard et al., 1991; Camoin et al., 2001; Gallup et al., 1994; Hearty et al., 1999; Kennedy et al., 2012; McMurtry et al., 2010; Muhs et al., 2011; Stirling et al., 2001; Thomas et al., 2012; Vezina et al., 1999; Zazo et al., 2007). Assessing the quality of pre-LIG fossil coral U-series ages would require open-system modeling, which is beyond the scope of this study.

Holocene coral U-series ages are not included in this study. These data are, however, being compiled by the HOLSEA working group (Khan et al., 2019; https://www.holsea.org, last access: 28 June 2021).

One of the outstanding controversies for fossil coral RSL reconstructions is whether fossil reef sites record evidence of millennial- or centennial-scale sea-level change within the LIG. Constraining the anatomy (pattern) of GMSL change within the LIG is crucial for our understanding of ice sheet dynamics in warm interglacial periods, such as today, and has direct bearing on future projections of sea-level response to anthropogenic forcing (Church et al., 2013; DeConto and Pollard, 2016; Sweet et al., 2017). The analytical precision of U-series dating and field surveying techniques has advanced dramatically over the past 30 years, but this key question remains unresolved (Kopp et al., 2017).