World Atlas of late Quaternary Foraminiferal Oxygen and Carbon Isotope Ratios

We present a global atlas of downcore foraminiferal oxygen and carbon isotope ratios available at https://doi.org/10.1594/PANGAEA.936747 (Mulitza et al., 2021a). The database contains 2106 published and previously unpublished stable isotope downcore records with 361 949 stable isotope values of various planktic and benthic species of Foraminifera from 1265 sediment cores. Age constraints are provided by 6153 uncalibrated radiocarbon ages from 598 (47 %) of the cores. Each stable isotope and radiocarbon series is provided in a separate netCDF file containing fundamental metadata as attributes. The data set can be managed and explored with the free software tool PaleoDataView. The atlas will provide important data for paleoceanographic analyses and compilations, site surveys, or for teaching marine stratigraphy. The database can be updated with new records as they are generated, providing a live ongoing resource into the future. Published by Copernicus Publications. 2554 S. Mulitza et al.: World Atlas of late Quaternary Foraminiferal Oxygen and Carbon Isotope Ratios


Introduction
Stable oxygen and carbon isotope ratios measured on foraminiferal shells are often regarded as the foundation of Marine Geology and Paleoceanography. The importance of these proxies stems from their broad applicability in time and space, their established and efficient analytical methods and their great value for stratigraphy and paleoceanographic reconstructions (see review by Pearson, 2012). Since the pioneering work of (Urey, 1947), millions of foraminiferal isotope measurements have been performed representing time slices from the Middle Jurassic (e.g., Vetoshkina et al., 2014) into the Anthropocene (e.g., McGregor et al., 2007). Foraminiferal isotopes have substantially contributed to the reconstruction and understanding of the global climate evolution since the Early Cretaceous (Cramer et al., 2009) including the validation of the orbital theory of the ice ages (Hays et al., 1976), reconstructions of ice volume (Shackleton and Opdyke, 1973;Waelbroeck et al., 2002) and water mass structure, ocean circulation, and carbon cycling (Curry et al., 1988;Duplessy et al., 1988;Boyle and Keigwin, 1987).
Despite their importance for the understanding of the Earth system, foraminiferal isotope data have not been systematically catalogued globally or stored in a database in a consistent and standardized format. Foraminiferal isotope data are usually available in arbitrary data formats and scattered across different data repositories, which hinders an automated analysis. Harmonized data collections have the advantage that (i) information about data coverage can be immediately accessed and visualized, for example in the planning phase of research projects, (ii) data can be quickly compared for verification/quality control or to separate local signals from global signals, and (iii) that customized software can be used to visualize and analyze the data.
Here we present the first global atlas of foraminiferal stable isotope data (with uncalibrated radiocarbon ages where available). The data are stored in netCDF format (Rew and Davis, 1990) and can be directly analyzed and visualized with the free software tool PaleoDataView (PDV, Langner and Mulitza, 2019). In PDV, age information for a specific sediment core is linked to any downcore proxy series imported for that core within the same collection. This strategy ensures the long-term maintainability and consistency of the age models across different proxy records (e.g., stable isotope records of different species) in the same collection. The netCDF format also allows the data to be analyzed using programming languages such as MATLAB, R, Fortran, C++ and Python.

Data sources and harmonization
The database is provided as a collection of 2106 netCDF files with δ 18 O and δ 13 C data of species-specific foraminiferal carbonate and 598 netCDF files with raw radiocarbon ages (see references in Table A1). A detailed description of the attributes and variables in the netCDF files is provided in Sects. S1 and S2 in the Supplement. About 79 % of the files containing stable isotope records were derived from data downloaded either from PANGAEA (http://www.pangaea. de, last access: 19 May 2022) or NOAA's National Centers for Environmental Information (NCEI, http://www.ncdc. noaa.gov, last access: 19 May 2022). The remaining 21 % of the stable isotope files are based on data obtained directly from a stable isotope laboratory by one of the co-authors (8 %) or have been digitized from tables provided in papers, or paper supplements (6 %), or through personal communication (7 %). Radiocarbon data are not as frequently archived in public databases as stable isotope data. Only 62 % of the files containing radiocarbon data were obtained from NOAA or PANGAEA, whereas data in 32 % of the files were copied from tables in papers or paper supplements; 5 % of the files are based on data directly obtained from laboratories, while only 1 % were obtained through personal communication. The data set also includes 105 previously unpublished species-specific stable isotope downcore records including both δ 18 O and δ 13 C values and 45 species-specific isotope downcore records for which either δ 13 C or δ 18 O was previously unpublished (Sect. S3 in the Supplement). A table containing all data sources is available from Zenodo (https://doi.org/10.5281/zenodo.6337519, Mulitza et al., 2021b).
To generate the netCDF files, metadata, isotope data, and radiocarbon ages (if available) were first assembled in species-and site-specific Excel files in the format required by PDV. The species names were preserved as used in the original publication. If more than one stable isotope record of the same species was available for the same core, we added a suffix (e.g., size class or version) to the species name. The Excel files were then edited for units (mainly conversion from "cm" to "m" and years to kiloyears) and metadata were added. Unavailable data fields were filled with "NaN". Finally, the Excel files were converted to netCDF files using the PDV import tool. Stable isotope data and radiocarbon data were saved in separate files to allow the radiocarbon file to link to several proxy records from the same core via the core label. After import, the data were inspected and quality controlled in PDV. Every row of the downcore data fields is associated with a "use flag" indicating whether the values should be included in an analysis (use flag = 1) or not (use flag = 0). This flag can be used to exclude outliers (e.g., due to turbidites) or radiocarbon reversals in a later analysis of the data while maintaining the original data in the file. Isotope values without replicates were imported with a use flag set to "1". For replicate stable isotope measurements the use flags were set to "0" and an average of the replicates (use flag = 1) was added to the series with a comment "Mean of multiple measurements" in the same row. Raw radiocarbon ages were generally imported with a use flag set to "0" since the data are uncalibrated. Most of the data are archived with original downcore depth of the samples. If a composite depth scale was used (e.g., for International Ocean Discovery Program (IODP) and its predecessors Deep Sea Drilling Program (DSDP) and Ocean Drilling Program (ODP) cores), a comment was added and care was taken that available radiocarbon dates were imported on the same depth scale. The data are stored as raw data, with all documented corrections removed from the data. This includes a previously subtracted reservoir age and corrections applied to the stable isotope values (e.g., to account for species offsets). Variables to store downcore radiocarbon reservoir and stable isotope corrections that may be applied to the data at a later stage are already included in the netCDF files. These variables have been imported with default values of 0.4 ka (±0.1) for all radiocarbon reservoir ages and a stable isotope correction of "0" for all oxygen and carbon isotope ratios. Both reservoir ages and stable isotope corrections can be edited within PDV.

Spatial and vertical coverage
Stable isotope records are available from all major ocean basins ( Fig. 1) but tend to cluster along continental margins, where higher sedimentation rates, and thus higher temporal resolutions, can be found compared to mid-ocean ridges or deep abyssal basins. About 65 % of the downcore records are from coring locations within 400 km of the coastline (Fig. 2). The deepest record in the atlas is from 5105 m water depth (EN066-29PG, eastern tropical Atlantic (Curry and Lohmann, 1983), the shallowest record from 50 m water depth (GeoB9503-5, Senegal mud belt, Mulitza, unpublished). However, the availability of records decreases in waters shallower than about 400 m (Fig. 3), where more dynamic sedimentation regimes exist, and below 3800 m due to carbonate dissolution which often prevents the production of reliable, continuous foraminiferal stable isotope records.
Records are available in all oceanic 5 • latitude bands with the highest number in tropical latitudes and decreasing numbers towards high latitudes (Fig. 3). This pattern is likely the result of the year-round accessibility of low latitudes compared to high latitudes where, due to sea ice cover or harsh weather conditions in the cold season, expeditions are often constrained to the warm season. The largest fraction (∼ 47 %) of the stable isotope records was measured on material from the Atlantic, whereas about 9 % are from the Southern Ocean (Fig. 4). However, with 21 cores per million square kilometers, the Mediterranean has the highest density of cores followed by the Arctic Ocean (7.8 cores (million km −2 )) and the Atlantic (7.5 cores (million km −2 )). The Pacific and Indian oceans are currently only covered by 2 and 2.1 cores (million km −2 ), re-spectively, which is likely a result of relatively low accumulation rates and poor carbonate preservation over large areas. In addition, the retrieval of sediment cores in the remote and deep central areas requires more ship time compared to the Atlantic and Mediterranean Sea.

Species distribution
The majority (61 %) of all stable isotope values available in this compilation were measured on planktic Foraminifera (see individual percentages for carbon and oxygen isotopes in Fig. 5). Among the planktic species, Globigerinoides ruber (37 %) and Neogloboquadrina pachyderma (28 %) are the most commonly used species, followed by Globigerina bulloides (17 %) and Trilobatus sacculifer (6 %). These species have a relatively broad geographical coverage and are considered as mixed-layer species in their respective environment (Schiebel and Hemleben, 2017). Isotope measurements on other planktic species (summarized under "other planktics") constitute about 12 % of all values in the atlas (Fig. 5); 75 % of the included planktic oxygen isotope values and 88 % of the included benthic oxygen isotope values are reported together with the corresponding carbon isotope value. Most of the benthic isotope values (70 %) were obtained from species of the genus Cibicides/Cibicidoides. Isotope values from the in-faunal genus Uvigerina constitute about 18 % of all benthic isotope values in the atlas. The grouping of the original species names into species/genus names used in Figs. 5 and 6 is provided in Sect. S4 in the Supplement.

Species-specific and latitudinal distribution of oxygen and carbon isotope values
In the current version, the atlas contains a total of 201 593 δ 18 O values. The lowest δ 18 O value (−7.51 ‰) is observed in the species G. ruber white (Fig. 6) from the Gulf of Mexico core LOUIS1924 under the influence of Mississippi freshwater discharge (Aharon, 2003). The highest planktic δ 18 O value (6.31 ‰) can be observed in the tropical species G. ruber from core M31_2-78_PC6 (Red Sea, Geiselhart and Hemleben, 1998a). With latitude, planktic δ 18 O values follow the typical bell-shaped pattern as expected from a dominant influence of sea surface temperature (Fig. 7). Benthic δ 18 O values range from −2.85 ‰ from Cibicides corpulentus in core OC205-2-108GGC from the western tropical North Atlantic (Slowey and Curry, 1995) to 5.9 ‰ from Uvigerina bifurcata from South Atlantic site JR244-GC528 . Vertically, the δ 18 O of benthic foraminiferal species increases with water depth over the upper 800 m as expected from decreasing temperatures within the main thermocline (Fig. 8 from the decreasing temperature stratification with increasing latitude (Fig. 7). The data set contains 160 356 δ 13 C values. Planktic Foraminifera from tropical latitudes show the highest δ 13 C values (Fig. 7) of up to 3.53 ‰ in shells of the species G. ruber from the Red Sea core M31_2-78_PC6 (Geiselhart and Hemleben, 1998a). Planktic δ 13 C values get as low as −17.7 ‰ on N. pachyderma sinistral (Fig. 6) in core LV28-4-4 from the Sea of Okhotsk (Kaiser, 2001), which might be related to a potential contribution from authigenic carbonate minerals that form with the anaerobic oxidation of methane (Cook et al., 2011). Benthic foraminiferal δ 13 C gets as low as −7.99 ‰ in Elphidium batialis from western North Pacific core KT90-9_21 (Oba and Murayama, 2004) and as high as 3.36 ‰ in the aragonitic shells of Hoeglundina elegans from western North Atlantic core OC205-2-149JPC (Slowey and Curry, 1995). Benthic species of the genus Cibicides/Cibicidoides show a clear trend toward decreasing δ 13 C values in the deep ocean (Fig. 8), as expected from the global distribution of δ 13 C in dissolved CO 2 (Kroopnick, 1985).  4. Number of cores (a) and records (b) for major ocean basins. A record is a downcore series of paired oxygen and carbon isotope measurements on a foraminiferal species or species group stored in a single netCDF file. Several records can exist for a single core. The counts include records/cores for which either δ 18 O or δ 13 C is missing. Numbers in small font below ocean basin name indicate density of cores in cores/10 6 km 2 (a) and percentage from the total number of records in the atlas (b) in each basin. Ocean basins follow the definitions in the World Ocean Atlas 2001(Stephens et al., 2002. Pacific includes the Sea of Japan and the Indian Ocean includes the Bay of Bengal and the Red Sea.

Distribution of radiocarbon ages
The data set contains 6153 individual radiocarbon ages with a maximum age of about 56 ka. About 47 % of the cores are associated with at least one radiocarbon date. Most of the radiocarbon-dated cores are from the Atlantic (44 %) fol-lowed by the Pacific (28 %) and the Arctic Ocean (12 %) (Fig. 9). The temporal distribution of the radiocarbon ages ( Fig. 10) shows that the last deglaciation has been preferentially dated, which is likely a consequence of the scientific attention focused on this time period and the limited stratigraphic extent of many coring techniques. The fraction of re-    versals is higher for the deglacial and glacial periods, where the higher sampling density increases the likelihood of reversals.

Marine geology and paleoceanography
Foraminiferal oxygen isotope ratios provide one of the most reliable tools for stratigraphy in marine sediments, particularly for time periods older than the range of the radiocarbon method, or if radiocarbon is not available or associated with large uncertainties due to unknown reservoir ages. Usually, oxygen-isotope stratigraphy is applied by using global (Imbrie et al., 1984;Prell et al., 1986;Lisiecki and Raymo, 2005) or basin-wide (Lisiecki and Stern, 2016) isotope reference curves. The collection presented here may provide the opportunity to find and align new records with the closest published isotope record measured on the same species, taking events into account that may only occur locally. Through its value for stratigraphy, our collection may also provide a foundation for the global mapping of seafloor sedimentation rates. The spatial quantitative mapping of sedimentation rates will allow the development of sediment budgets for the seafloor, including carbon burial.
Oxygen and carbon isotope ratios of Foraminifera are of great value for paleoclimatology by providing information on the history of seawater temperature and isotopic composition as well as circulation, productivity, and carbon sequestration. This isotope atlas will allow for new global compilations to be undertaken to understand these processes at a global scale. Although distorted by habitat effects and vital effects, there is hope that some of these effects can be represented and quantified in foraminiferal ecosystem/calcification models (e.g., Wolf-Gladrow et al., 1999;Schmidt and Mulitza, 2002;Fraile et al., 2008). Since the number of climate models containing the cycling of oxygen and carbon isotopes is constantly growing (Marchal and Curry, 2008;Kurahashi-Nakamura et al., 2017;Tierney et al., 2020;Muglia et al., 2018;Völpel et al., 2017), foraminiferal isotopes may provide the opportunity to validate climate model experiments directly. Given this prospect and the spatial coverage, foraminiferal isotope data should be rescued, assembled, and organized to secure the information for future applications as we continue to improve our understanding of the ecological and geochemical processes that determine isotope ratios in foraminiferal shells. Depending on the scientific problem, paleoceanographic compilations usually have specific criteria (e.g., temporal resolution or the availability of radiocarbon ages) for the selection of the records to be included (e.g., Jonkers et al., 2020). An atlas product that includes the majority of the available records enables quick selection of suitable data without an extensive literature review.

Expedition planning
The planning of marine coring campaigns requires prior knowledge of existing cores. Existing core locations are often resampled to get new sediment material or to extend the stratigraphic coverage with alternative coring gear that can penetrate deeper into the sediment. For example, many IODP and ODP cores are drilled at sites where short cores were previously retrieved. The knowledge of existing core locations and their stratigraphy allows identification of sampling gaps. Many aspects of marine expeditions are unpredictable, and schedules and coring plans regularly have to be adapted, often on a daily basis. The atlas we are presenting here provides fast access to stratigraphic data and may aid the identification of suitable alternative coring locations on ocean expeditions. Both the freely available PDV software and the atlas do not require web access and are therefore suitable to be used with a standard laptop computer.

Education
Foraminiferal oxygen isotope ratios are still the most valuable stratigraphic tool in marine sediments. The atlas covers various sedimentation regimes and therefore provides numerous examples of how factors like local hydrography, species, or sedimentation rates influence the patterns of downcore isotope ratios. It therefore may be used as a resource to train students in regional isotope stratigraphy for studies in paleoceanography, paleoclimate, and marine geology. Lecturers may employ the atlas together with PaleoDataView or with custom software to show examples on how isotope stages may be identified in different geological settings and on how isotope differences between species may be explained by hydrography and foraminiferal ecology. Students may also actively explore the patterns of isotope stratigraphies from different parts of the global seafloor to actively learn how global factors such as ice volume and local factors such as sea surface temperature (SST) and freshwater input influence stable isotope records.

Data availability
All data included in the World Atlas of late Quaternary Foraminiferal Oxygen and Carbon Isotope Ratios can be downloaded at https://doi.org/10.1594/PANGAEA.936747 (Mulitza et al., 2021a). For use with the soft-ware PaleoDataView, the unzipped root directory ("WA_Foraminiferal_Isotopes_2022") of the collection with all its content can be copied into the "Documents/Pa-leoDataView/" folder (Windows) or the /PaleoDataView/ folder under "Applications" (macOS). Select the root directory "WA_Foraminiferal_Isotopes_2022" under "Data → Change Collection → Change Working Directory" to explore the data. For use with custom software, netCDF files containing stable isotopes data are stored under "WA_Foraminiferal_Isotopes_2022\Foraminiferal Isotopes\Data\" and the radiocarbon data under "WA_Foraminiferal_Isotopes_2022\Age\".
Information on the installation of PaleoDataView is available from Langner and Mulitza (2019).

Future: building a dynamic world atlas of marine sediments
The amount of proxy data from marine sediments is growing fast, and the demand of data sets that can constrain past states of the Earth system is increasing. The complexity of the data makes it challenging to maintain and reduce the data sets into spatially and chronologically coherent and meaningful data sets. We propose to initiate an atlas series that provides raw data in a consistent data format as a first step from data archived in public databases (as published) towards more sophisticated data products describing past states of the ocean and the seafloor (Fig. 11). Eventually, these harmonized data sets can form a continuously growing and sustainable public "database layer" where proxy-specific raw data can be queried and directly loaded into software that provides the tools to generate homogenized data products that can reach out into other disciplines, i.e., climate modeling. We present a simple file-based data collection where each file contains only one proxy record rather than all available data of the core. Paleoclimatic data are often analyzed and assembled in proxy-specific collections, because proxy-specific transfer functions have to be applied in order to quantify environmental variables. Furthermore, comparisons of records from different sites are preferably done on the same proxy type to ensure comparability. A single file per proxy facilitates the composition of proxy-specific collections, avoiding the additional costs (i.e., in terms of data management and disk space) of other downcore parameters in the same file. This modularity also allows individual scientists to separate their unpublished/unvalidated data from published/validated data that are ready to be included into a proxy collection. On the other hand, it is desirable to consistently apply the same stratigraphy to all proxies from a single core. PDV will automatically apply a single age model to all proxy records with the same core label. This requires that the depth scales and the core label of the different proxies are identical, when the data are imported. Figure 11. Potential workflow to form sustainable data products from raw databases.
Stable isotopes and radiocarbon ages usually provide the stratigraphic basis for further investigations. When collections of other proxies are added to PDV, these collections can rely on the stratigraphic data provided here and any changes in the stratigraphy will be applied to all proxy data in the collection. The efficient visualization of the data in PDV allows the identification of erroneous data and helps to improve the atlas product over time. The Excel export and import functions of PDV also ensure access to the data for individuals without strong programming skills.
Author contributions. SM conceptualized the atlas and wrote the paper with input from all co-authors. All authors assisted in the compilation of the data, and/or provided unpublished data and contributed to the paper.
Competing interests. The contact author has declared that neither they nor their co-authors have any competing interests.  Duplessy, J. C., Labeyrie, L. D., and Shackleton, N. J.: Changes in the distribution of δ 13 C of deep water CO 2 between the Last Glaciation and the Holocene, Paleoceanography