A global spatially explicit database of changes in island paleo-area and archipelago configuration during the late Quaternary

Disclaimer/Complaints regulations If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You will be contacted as soon as possible.

see Borregaard et al., 2017 for a review) and the glacial sensitive model (GSM; incorporate the role of longterm changes in island geography in predictions of biodiversity patterns. On time-scales of < 100 kyr, geographical changes on most oceanic islands are shaped mainly by sea level fluctuations. These fluctuations involve the repetitive fusion and fission of islands, area contraction and expansion, and changes in the number of stepping stones (reducing inter-island isolation) by the emergence and drowning of seamounts .
Until now, temporal dynamics of palaeo-geography (e.g., the duration over which islands were connected and the rates at which island area changed) remained relatively underexplored. Biogeographical and macroecological studies so far have mainly explored the role of one (static) sea level stand, such as the current high sea level stand or the extreme low stand at the LGM, although both are highly exceptional when considering the last 1 Myr (Supporting Information Appendix S1; Bintanja, van de Wal, & Oerlemans, 2005;van Andel, 1989;Woodruff, 2010). In this paper, we present a global, spatially explicit database with a quantification of changes in island palaeo-area and reconstructions of archipelago configuration driven by sea level fluctuations during the late Quaternary. This Palaeo-Islands and Archipelago Configuration (PIAC) database consists of 178 islands located in 27 archipelagos spread around the globe (see Figure 1 for an overview). Here, we extend previous work by reconstructing island palaeo-geography dynamics on a multi-millennial time-scale and with global coverage. In the subsequent sections, we describe the methods by which the data were derived, we provide the technical validation and discuss the reliability of the database and possible applications.

| M ETH OD S
A method to calculate palaeo-area change driven by sea level fluctuations was developed by Rijsdijk, Hengl, Norder, Avila, & Fern andez-Palacios (2013) and Rijsdijk et al. (2014). To create the PIAC database, we have enhanced the method developed by Rijsdijk et al. (2013Rijsdijk et al. ( , 2014; we used higher-quality input data, the computation time is reduced, the contouring method has been improved, and the model is applied to a larger number of archipelagos ( Figure 1). The workflow was subdivided into the following four steps ( Figure 1): collecting input data, intermediate processing, generating output data, and data validation.
To quantify palaeo-area, three types of input data were used: The GEBCO_2014

| Description of the database
Palaeo-area and archipelago configurations were reconstructed for 178 islands within 27 archipelagos. The PIAC database is stored for a time span of 35 kyr (Lambeck et al., 2014) and 140 kyr (Cutler et al., 2003).
The workflow for producing the PIAC database (R scripts and associated files) is organized within separate folders for each processing step In Figure 2, the data on archipelago configuration and palaeo-area are visualized for the Canary Islands. As a result of bathymetric differences, each island shows a unique area change signature. Similar graphs of palaeo-area change for other archipelagos are presented in Supporting Information Appendix S2 for a time span of 35 kyr (Lambeck et al., 2014) and 140 kyr (Cutler et al., 2003).

| Validation
The comparison of the area calculated at the present-day sea level with the real current island area indicates that they are highly correlated (Pearson product-moment correlation coefficient, r 5 0.99; Supporting Information Appendix S3). For most islands, the calculated area is slightly larger than the real area, but in a few cases it is smaller.
The deviations stem from the spatial resolution (30 arc-second) of the bathymetry DEM. Consequently, the proportional deviations are larger for small islands (Supporting Information Appendix S3).

| Accuracy
Possible inaccuracies in the calculated palaeo-area and reconstructed archipelago configuration stem from two factors. The first factor is the resolution of the bathymetry DEM (30 arc-second; c. 1 km at sea level at the equator). To our knowledge, there is no publicly available global bathymetry DEM with a higher resolution than the one used to produce this database. As DEMs with higher resolution become available, the code accompanying this paper can be applied to generate outputs with increased accuracy. Second, we used a global mean sea level curve The island centre points are used to identify the islands for which palaeo-area is calculated. The calculated area for the present sea level are compared with the current area by calculating the Pearson product-moment correlation coefficient (r) for reconstructing regional geographies. Sea level is not uniform across the globe; regional deviations are caused by the complex interactions between oceans, ice-sheets and the Earth's crust, including geological processes (e.g., glacial isostatic adjustment, and vertical land movement resulting from both gradual and sudden tectonic processes), variations in ocean mass and density, and 'fingerprint' effects, such as the gravitational attraction between water and ice-sheets (Clark, Farrell, & Peltier, 1978;Farrell & Clark, 1976;Kopp, Hay, Little, & Mitrovica, 2015;Lambeck et al., 2014;Milne & Mitrovica, 2008;Raymo, Mitrovica, Leary, Deconto, & Hearty, 2011). Although these different processes might negate each other, regional deviations from the global sea level curve up to tens of metres are possible (Kopp et al., 2015;Milne & Mitrovica, 2008;Woodroffe, McGregor, Lambeck, Smithers, & Fink, 2012    | 503 over long time-scales is currently unfeasible for many archipelagos owing to lack of data, and is therefore currently impossible on a global scale.

| Applications
The global PIAC database enables movement beyond a focus on extreme archipelago configurations, such as during the present interglacial high or the LGM (see Ferentinos et al., 2012;Voris, 2000;Warren, Strasberg Fuerteventura and Lanzarote; Figure 2) on biodiversity and phylogenetic patterns? And, is present-day beta diversity related to inter-island palaeodistance? The database also allows for the reconstruction of coastal environments for archaeological studies, exploration of potential archaeological sites in currently submerged regions (Erlandson & Fitzpatrick, 2006;Lambeck & Chappell, 2001;Rick, Kirch, Erlandson, & Fitzpatrick, 2013), and the development of simulation models in human dispersal studies (Ferentinos et al., 2012;Montenegro et al., 2016).
Area or isolation metrics derived from the PIAC database can be used as explanatory variables in macroecological and evolutionary studies. Examples of metrics that could be calculated from the database are as follows: rate of area change; minimal, maximal or average area over a biogeographically relevant time period; or the mean, minimal and maximal inter-island distance. For local scale biogeographical, phylogenetic, population genetic and archaeological studies, we recommend use of a regional sea level curve where available (Simaiakis et al., 2017;van Andel, 1989;Warren et al., 2010). In addition, regional crustal tectonic or other effects leading to temporal deviations from the global mean curve are not incorporated. If precise timings of geographical change are required, reference to known local deviations from the global mean sea level curve is recommended. The database presented in this paper should be regarded as an approximation of palaeo-area and archipelago configuration shaped by sea level fluctuations to be refined as new data and methods become available. The R scripts and functions included in the database allow more detailed calculations for specific islands or archipelagos where a regional sea level curve or a high-resolution bathymetry DEM exists.

ACKNOWLEDGMENTS
We are very grateful to the three anonymous referees who provided very helpful comments on our manuscript. Their suggestions greatly