The stable isotope composition of water (δ18O and δ2H) provides a powerful tracer of the global water cycle, tracking water as it continuously cycles between the ocean, atmosphere, and land. The isotopic composition of seawater is nearly conservative when no phase change occurs. As water molecules undergo phase changes however, the lighter, more abundant isotope (¹⁶O and ¹H) is preferentially vaporized during evaporation with respect to the heavier, less abundant isotope (¹⁸O and ²H), while the heavy isotope is preferentially condensed during precipitation (Dansgaard, 1964). This partitioning of isotopes based on mass allows the isotope values (where δ18O=18O16Osample18O16Ostandard-1⋅1000) of water to be used as a tracer of the hydrologic cycle (Dansgaard, 1954; Galewsky et al., 2016; Gat, 1996). In the ocean, the isotope values of seawater (δ18O_sw and δ2H_sw) can provide valuable information on an array of processes, including heat and mass exchange with the atmosphere (via precipitation and evaporation), large-scale ocean circulation, and freshwater input from rivers and ice sheets (Akhoudas et al., 2021; Benetti et al., 2016; Benway and Mix, 2004; Biddle et al., 2019; Craig and Gordon, 1965; Dee et al., 2023; Frew et al., 2000; Imbrie et al., 1984; Jacobs et al., 1985; Lisiecki and Raymo, 2005; Meredith et al., 1999; Strain and Tan, 1993). δ18O_sw and δ2H_sw values can also provide insight into other ocean tracers such as salinity, since they covary strongly due to the influence of evaporation and precipitation on each of these variables (Craig and Gordon, 1965; LeGrande and Schmidt, 2011). However, because key processes act differentially on salinity as compared to the stable isotope values, δ18O_sw and δ2H_sw provide additional constraints on ocean mixing and the local moisture budget.

Seawater isotope values also create a common unit that uniquely links paleoclimate reconstructions to modern climate observations and isotope-enabled model simulations. Modern δ18O_sw data are essential for the calibration of paleoclimate proxies of past ocean variability based on the δ18O of marine carbonates such as corals, foraminifera, molluscs, ostracods, and coralline algae. Recent paleoclimate data assimilation efforts such as the Last Millennium Reanalysis project (e.g., Tardif et al., 2019) would greatly benefit from a spatial network of δ18O_sw data to improve quantification of proxy uncertainty and for training the proxy system models that underlie those efforts. Modern δ2H_sw data are used in the calibration of paleoceanographic proxies based on the δ2H of alkenones and other lipid biomarkers in marine sediments (e.g., Eglinton and Eglinton, 2008). When used in tandem with δ18O data (i.e., to calculate d-excess in surface ocean and overlying water vapor), δ2H data can be used to constrain evaporation parameters (e.g., Benetti et al., 2014). As such, observational and reconstruction efforts based on seawater isotope values enable scientists to better understand the underlying physics that govern the water cycle, and to extend hydroclimate records back to the preindustrial era, thus contextualizing anthropogenic climate change and improving the skill of future climate projections. In modern environments, δ18O in marine biominerals and δ2H in lipids can be used to trace plankton and animal movement and provide provenance for ecology, conservation, archaeology, and food forensics studies (Doubleday et al., 2022). Given these wide-ranging applications, seawater isotope data are used in a wide range of fields, including paleo and modern oceanography, atmospheric science, geology, marine biology, archaeology, and geography.

Observational δ18O_sw data can also be used to provide boundary conditions in climate models and to assess model performance and skill. The increasing integration of oxygen isotopes of water in climate models – from models of intermediate complexity to fully coupled Earth System Models (e.g., Blossey et al., 2010; Bong et al., 2024; Bony et al., 2008; Brady et al., 2019; Cauquoin et al., 2019; Dee et al., 2015; Field et al., 2014; Fiorella et al., 2021; Kurita et al., 2011; Lee and Fung, 2008; Noone and Simmonds, 2002; Nusbaumer et al., 2017; Risi et al., 2010, 2020, 2021; Schmidt et al., 2007; Tada et al., 2021; Wei et al., 2018; Werner et al., 2011; Yoshimura et al., 2008) – bolsters the interpretation of modern and paleoclimate observations, while also providing opportunities to test model performance in resolving key features of the hydrologic cycle, e.g., the representation of moisture transport, circulation, and surface water fluxes.

Paralleling recent advances in the numerical simulation of water isotopes, new analytical capabilities have also developed in recent years, including new in situ atmospheric measurement techniques and strategies (Finkenbiner et al., 2022; Gupta et al., 2009; Henze et al., 2022), and the development of global atmospheric data products from a variety of remote sensors (e.g., Diekmann et al., 2021; Schneider et al., 2022; Worden et al., 2019). As a result, measurements of water isotopes have become increasingly incorporated in coordinated observing networks and monitoring studies of precipitation and atmospheric water vapor, including the Global Network of Isotopes in Precipitation (http://www.iaea.org/services/networks/gnip. last access: 20 February 2026) and the National Ecological Observatory Network (http://www.neonscience.org/, last access: 20 February 2026).

However, no such coordinated observing network for seawater δ18O currently exists. Unlike meteorological observations on land, observations of ocean hydrological properties (e.g., precipitation, evaporation, and salinity) are largely either limited to the past few decades (via satellite remote sensing, the ARGO (Wong et al., 2020) and GOOS (Dexter and Summerhayes, 2010) programs, the TAO/TRITON array, and other mooring, drifter, and ship-of-opportunity measurements), or are confined to selected coastal and island locations that have the necessary infrastructure to support sustained in situ measurements of ocean surface properties. Furthermore, these ocean observations rarely include δ18O_sw because there is currently no cost-effective, easily deployable instrumentation to measure seawater isotopes in situ. Thus, seawater samples must be taken back to a laboratory for isotopic analysis. Despite these structural challenges, a growing number of δ18O_sw datasets have been generated in recent decades due to the accelerated collection of δ18O_sw samples, new instrumentation based on cavity-ring down spectroscopy (CRDS) with reduced analytical costs, the capability to measure both δ18O and δ2H in parallel, and new sampling devices that enable long-term seawater sample collections (e.g., Jannasch et al., 2004; Khare et al., 2021).

In recognition of the broad value of δ18O_sw data to the Earth Sciences, a major effort to gather δ18O_sw data occurred in the 1990s (Bigg and Rohling, 2000; Schmidt, 1999; Schmidt et al., 1999) and resulted in the development of the NASA's Goddard Institute for Space Studies (GISS) Global Seawater Oxygen-18 database (https://data.giss.nasa.gov/o18data/, last access: 20 February 2026), which contains over 26 000 global measurements of δ18O_sw (and some δ2H data) from the 1950s to 2000s. In 2006, that database was used to construct a global gridded dataset of δ18O_sw and to characterize regional relationships between δ18O_sw and salinity (LeGrande and Schmidt, 2006) and it has subsequently been used in a broad range of studies involving δ18O_sw. However, the NASA GISS database is no longer actively updated, with the last δ18O_sw measurement added in 2011. As a result, a growing number of new δ18O_sw datasets published since 2011 remain without an active δ18O_sw-specific data repository in which to archive the data. Researchers have instead provided the δ18O_sw data in the supplemental tables of journal articles, or have archived the δ18O_sw data with other geochemical data (e.g., coral δ18O), in data repositories such as the National Centers for Environmental Information (NCEI) for Paleoclimatology (https://www.ncei.noaa.gov/products/paleoclimatology, last access: 20 February 2026) and PANGAEA (https://www.pangaea.de/, last access: 20 February 2026). Because these datasets can be difficult to find, non-machine-readable, and/or decentralized, they are not easily accessible to the wide range of research communities that would benefit from this data (see a related review by Chamberlain et al., 2021). Furthermore, many publishers and several funding agencies now require researchers to archive their data in FAIR and public repositories. For these reasons, a comprehensive database of δ18O_sw data that is publicly available and actively maintained is critically needed.

CoralHydro2k and its Seawater δ18O Database started in 2017 as a project in Phase 3 of the Past Global Changes (PAGES) 2k network, a long-running initiative to study past global changes over the last 2000 years and to compile paleoclimate data in publicly available, machine-readable databases (Konecky et al., 2023; PAGES 2k Network Coordinators, 2017; Tierney et al., 2015) The CoralHydro2k project was formed to investigate the variability of hydrology and temperature in the tropical surface ocean during the past 2000 years based on the combination of coral δ18O, which varies with temperature and δ18O_sw, and the strontium-to-calcium ratio (Sr/Ca) in corals, which is a temperature proxy. To aid in the calibration and interpretation of the paired coral δ18O and Sr/Ca records in the database, and derive coral-based reconstructions of seawater δ18O, the CoralHydro2k project started to compile δ18O_sw data.

In recognition of the growing number of δ18O_sw datasets that have been generated during the last two decades, the CoralHydro2k Seawater δ18O Database project was launched in 2020 to recover “hidden” δ18O_sw data that were not easily findable. During the past five years, we have integrated these records, along with any associated δ2H, salinity, and temperature data, with data from public databases and repositories to create a new, centralized, machine-readable, and metadata-rich database that aligns with findability, accessibility, interoperability, and reusability (FAIR) standards (Wilkinson et al., 2016). Here we provide a detailed description of the PAGES CoralHydro2k Seawater δ18O Database, highlighting the opportunities and limitations of the current database. We also provide a data submission template with comprehensive metadata for guiding the reporting of seawater isotope data in the future.

CoralHydro2k included team members from the Phase 1 PAGES Ocean2k working group (Tierney et al., 2015) and Phase 2 Iso2k working group (Konecky et al., 2020) and many new members, particularly from the coral paleoclimate community. CoralHydro2k continued into Phases 3 and 4 of PAGES 2k, focusing on reconstructing past changes in tropical ocean temperature and hydroclimate using paired Sr/Ca and δ18O from coral archives over the last 2000 years (Hargreaves et al., 2020; Walter et al., 2023). Recurring calls went out within the international paleoclimate community for working group members, coral experts, and paleo data assimilation experts to join the effort with monthly teleconference meetings and one in-person meeting in 2019 (Hargreaves et al., 2020). As a result, the CoralHydro2k database was produced, a global, actively curated compilation of coral δ18O and Sr/Ca proxy records of tropical ocean hydrology and temperature for the Common Era (Walter et al., 2022, 2023). A number of sub-projects were developed in conjunction with CoralHydro2k, including a project to develop new proxy system models (PSM) for coral δ18O. The group working on this sub-project realized that the spatial and temporal coverage of observational δ18O_sw data were too sparse to integrate into the PSM framework and that many new δ18O_sw datasets produced during the last few decades are not easily findable or accessible.

CoralHydro2k thus formed a new sub-project in 2020 to compile existing seawater δ18O data with rich metadata following FAIR standards (Atwood et al., 2024; DeLong et al., 2022). Researchers were invited to submit their data to the CoralHydro2k Seawater δ18O Database via a Qualtrics survey and accompanying YouTube video that provided instructions on how to submit data.

The workload for assembling the seawater data and metadata was performed by CoralHydro2k members and new members of the Seawater δ18O Database sub-project. The team was made of volunteer scientists from all academic levels, including undergraduate and graduate students, postdoctoral researchers, and early- to senior-level scientists from several international academic and research institutions. The work was completed remotely in synchronous working sessions and asynchronously across several virtual platforms (Google Suite, Slack, and Zoom). Data discovery, metadata protocols, and compilation were done collaboratively as the project progressed.

To facilitate future curation of seawater isotope data by the research community, we established the Seawater Oxygen Isotopes Community (https://www.earthchem.org/communities/seawater-oxygen-isotopes/, last access: 20 February 2026) in the EarthChem Library (ECL), a data repository that archives, publishes, and provides access to data in the geosciences. The ECL offers a suite of services for data preservation and access, including long-term archiving and data registration with a Digital Object Identifier (DOI). Through the new Seawater Oxygen Isotopes Community, new seawater δ18O (and δ2H) datasets can be submitted and assigned a DOI, which allows the datasets to be cited and tracked when used by other researchers. The CoralHydro2k members promoted this new database at international conferences in the United States and Europe, in the PAGES newsletter (Atwood et al., 2024), and in Eos, the monthly magazine of the American Geophysical Union (AGU) (DeLong et al., 2022).

The CoralHydro2k Seawater δ18O Database was designed to be as inclusive and comprehensive as possible in its record-selection criteria to support the project goal of developing a FAIR database of global seawater δ18O measurements, paired with δ2H and salinity measurements, and to include as much “hidden” data as possible. Thus, the Seawater δ18O Database selection criteria were less restrictive than other PAGES 2k efforts, and the database includes data from peer-reviewed scientific literature, student theses and dissertations, public data repositories, and direct author submission.

For hidden data, we searched for and included datasets spanning all depths and all latitudes across the global ocean. For publicly available data, given the substantial time commitment involved in finding and adding the extensive metadata, we typically only included data from the upper 50 m between 35° N to 35° S (to aid in CoralHydro2k's seawater δ18O reconstruction studies using δ18O and Sr/Ca in tropical-subtropical shallow-water corals). In subsequent versions of the database, we will target the inclusion of all publicly available datasets.

In alignment with FAIR data principles, the Seawater δ18O Database contains extensive metadata. 12 metadata fields are required, with an additional 38 optional metadata fields that provide important supporting information on the sampling site, sample collection and storage, isotope analysis method, evaporation and correction flags, and error information. Where available, paired seawater δ2H, salinity, and temperature data are also reported. The full set of required and optional metadata fields in the database are intended to establish a set of best practices for future reporting of seawater isotope data. The Scientific Committee on Oceanic Research (SCOR) working group 171 “Towards best practices for Measuring and Archiving Stable Isotopes in Seawater (MASIS)” further intends to build upon the set of best practices established for the CoralHydro2k Seawater δ18O Database. We note that many of the optional metadata fields in the database are essential for proper quality control, inter-comparison, and interpretability across datasets. However, this information was often not reported in the original datasets and publications, and thus this metadata could not be made mandatory without greatly restricting the number of datasets in the database. However, we strongly encourage the inclusion of all metadata fields in this database (required and optional) for future reporting of seawater isotope data. To assist researchers with this process, a blank data submission template (with examples) is provided with the CoralHydro2k Seawater δ18O Database.

The metadata fields are described in Tables 1–2. The CoralHydro2k Seawater δ18O Database team implemented several rounds of quality control measures for the data and metadata. Following the Iso2k database procedure (Konecky et al., 2020), each metadata field has an associated quality control certification “Level” from 1 to 5, described below and in Table 1. Level 1 and Level 2 metadata fields constitute “essential” metadata, and if a dataset lacked one of these fields, it was excluded from the database.

Level 1 fields are required for inclusion in the database and they contain standardized vocabularies, according to Table 2. They are recommended as primary fields for filtering and querying records in the database. They were subject to the highest Quality Control (QC) standard. Examples of Level 1 metadata are: “Collection year”, “Collection month”, “Latitude”, “Longitude”, and “Depth”.

The CoralHydro2k Seawater δ18O Database contains 18 598 data points from 106 datasets (Fig. 1A, B). 53 % of the data (9862 data points) is categorized as “hidden” data (i.e., data not currently available in public databases or public repositories; Fig. 1C, D), and the remaining 47 % of the data is from public databases or public repositories (Table 3). 10,407 measurements (56 %) are from the sea surface (depth ≤ 5 m; Fig. 1A), 3693 (20 %) are from the mixed layer (between 5–50 m), and 4498 (24 %) are below 50 m. The time span of the database covers 1972 to 2021 (Fig. 2) and the depth range covers the surface to 5797 m below sea level. The earliest data point in the database was collected in September 1972 and the most recent data point was collected on 8 October 2021. 3480 data points (19 %) were collected before the year 2000, and 15 118 data points (81 %) were collected on or after the year 2000 (Fig. 2). Because the addition of public datasets focused on the region between 35° N and 35° S, 75 % of the measurements in the database are located within the tropical-subtropical region.

In addition to δ18O measurements, the database also includes paired δ2H, salinity, and temperature measurements when available. 16 098 data points (87 %) have paired salinity values (Fig. 3A), 13 871 data points (75 %) have paired temperature values (Fig. 3B), and 9769 data points (53 %) have paired δ2H measurements (Fig. 3C). 185 measurements have an evaporation flag (Fig. 3D), which allows the user to filter out samples that may be influenced by post-collection evaporation from the database.

Compared to the seawater δ18O database presented in LeGrande and Schmidt (2006), data coverage in the surface ocean (upper 5 m) is substantially improved in the tropics and subtropics, particularly in the northern Indian Ocean, the eastern Atlantic Ocean, the northeast coast of South America, the Mediterranean Sea, and the equatorial Pacific Ocean. However, poor data coverage still exists in the western Pacific Ocean and Maritime Continent region, the southeastern Indian Ocean, and the subtropical Pacific Ocean regions in both hemispheres. Below 5 m depth, the data coverage is even more limited (Fig. 1B). At all depths, regions with reasonable spatial coverage of δ18O_sw data contain limited temporal coverage (Fig. 4A). Typically, only a few years of regular measurements are available from the most highly sampled regions. For example, only 21 % of locations contain at least 12 measurements spanning two years within a 2° latitude × 2° longitude grid box and only 8 % of the grid boxes contain data that cover at least 50 % of the annual cycle (i.e., 6/12 months of the calendar year; Fig. 4B). While the coverage of seawater isotope data has been growing over the last decade, these measurements are still sparse in space and time, thus highlighting the need for globally coordinated sampling campaigns and archiving efforts.

Outside of the tropics and subtropics, the coverage of δ18O data in the CoralHydro2k Seawater δ18O Database is sparser, since only hidden datasets were collected from all latitudes and all depths across the global ocean, while datasets from public repositories were only incorporated into the database if the measurements were made in the upper 50 m between 35° N and 35° S. Future database development efforts will include incorporating additional hidden and public datasets.

To assess how the δ18O_sw data in the database compares with isotope-enabled climate model simulations and other products, we compare the climatological annual cycle in δ18O_sw at different island sites using four data products: two simulations of isotope-enabled General Circulation Models (the National Center for Atmospheric Research Community Earth System Model Last Millennium Ensemble (1000 years; Brady et al., 2019) and the NASA Goddard Institute for Space Studies E2-R last millennium simulation (ensemble member E4rhLMgTck; 255 years; Colose et al., 2016)), a regional ocean model of the Pacific called isoROMS (44 years; Stevenson et al., 2018), and a gridded dataset of global monthly mean δ18O_sw based on data assimilation with the MITgcm (Breitkreuz et al., 2018). The Breitkreuz dataset is based on a 400-year quasi-equilibrated simulation of a water isotope-enabled global ocean general circulation model constrained by global monthly δ18O_sw data collected from 1950 to 2011 and climatological salinity and temperature data collected from 1951 to 1980.

The characteristics of δ18O_sw variability at the four selected sites in the tropical Pacific and Atlantic Oceans vary widely across the different data products and the δ18O_sw observations from the CoralHydro2k Seawater δ18O Database, with large differences in both the amplitude and phase of the annual cycle of δ18O_sw (Fig. 5). These differences could be due to deficiencies in the models (associated with subgrid-scale parameterizations, treatment of atmospheric exchange and ocean mixing processes, and/or the limited spatial resolution of the models not capturing the local influence of evaporation, precipitation, runoff, and upwelling at the sampling sites), and/or uncertainties in the observational data given the low temporal resolution of the δ18O_sw measurements. Clearly, more observational data is needed to determine the source of the discrepancies, pointing to the need for more coordinated and sustained seawater isotope sampling programs. Seawater isotope sampling at multinational observing systems that are already in place, such as the Tropical Pacific Observing System (TPOS), Bermuda Atlantic Time-series Study (BATS), GO-SHIP, and GEOTRACES, could expand and complement existing observational programs. For example, incorporating new sampling devices such as long-term osmotically pumped fluid samplers (Jannasch et al., 2004; Khare et al., 2021) could provide a relatively straightforward way to add δ18O_sw measurements to existing programs. The development of sustained seawater isotope measurements at a network of observational hotspots around the global ocean would provide powerful new constraints on hydrologic changes in the modern ocean, generating data that could be used to test theoretical predictions, assess climate model performance and skill, and calibrate paleoclimate proxies for improved paleoclimate reconstruction.

The CoralHydro2k Seawater δ18O Database follows the FAIR data principles (Wilkinson et al., 2016) that strive to make scholarly data findable, accessible, interoperable, and reusable. The CoralHydro2k Seawater δ18O Database uses the Comma Separated Values (*.CSV) file format, a machine-readable format for archiving and describing seawater isotope data. Access to the database has been granted for reviewers and editors during the review phase. The data are also available upon request for members of the public that wish to participate in the review process by emailing the corresponding author. The database is available on the NOAA NCEI World Data Service for Paleoclimatology landing page: https://www.ncei.noaa.gov/access/paleo-search/study/34575 (last access: 11 February 2026) (10.25921/ap7d-2k16, Atwood et al., 2026). We recommend accessing the database through the NOAA NCEI landing page to find the latest instructions on using the database.

Example Jupyter notebooks and MATLAB scripts are available on the CoralHydro2k Seawater Database GitHub page: https://github.com/CoralHydro2k/ch2kSeawater_Database (last access: 11 February 2026, and Zenodo at 10.5281/zenodo.18828491, Hargreaves, 2026) to help users search, filter, and visualize the database. We encourage users of the database to share their scripts on GitHub as well for improved access.

The CoralHydro2k Seawater δ18O Database includes records (0–50 mbsl, 35° N to 35° S) from ten international databases, including GEOTRACES, NASA GISS, PANGAEA, CISE LOCEAN, GLODAPv2, WIDB, NOAA NCEI, and BODC (Table 3). Literature searches were also conducted to find hidden seawater δ18O data (from all depths and latitudes) published only in tables and supplemental data files of published papers, theses, and dissertations. Data was also sourced from author contributions sent directly to this project or the EarthChem community (2026): https://earthchem.org/communities/seawater-oxygen-isotopes (last access: 11 February 2026). Users of this database should adhere to the data use policies for the underlying data sources (see Table 3 and Supplement Sects. S1–S4).