A browser tool called GPlates Web Service lets researchers place a fossil in deep time without a supercomputer

On 20 July 2026, researchers behind the GPlates Web Service published a browser-based tool that lets a user type in a modern latitude and longitude for a fossil site and watch the position of that point migrate across a half-billion-year animated map of the Earth's surface. The portal is built on GPlates, the open desktop software long used by geologists to reconstruct ancient plate configurations, and turns what was once a desktop workflow requiring installed software and curated rotation files into something a fieldworker can finish before lunch.

The release matters because paleogeographic reconstruction has been a quiet bottleneck in the discipline. A trilobite collected last week in Morocco may have lived on the southern margin of a vanished ocean. A fossil pollen grain lifted from a Patagonian core may have travelled thousands of kilometres before it settled. Pinning down where the organism actually lived when it died has, for two decades, meant either installing GPlates, downloading the right set of plate-rotation models, and running the analysis locally, or paying a specialist with access to curated infrastructure. The new service pushes that capability into a browser tab, with the rotation models hosted server-side and the heavy lifting done remotely.

What the tool actually does

Users enter a sample's present-day coordinates and a target geological age. The service returns the paleocoordinates and the uncertainty envelope around them. Output can be downloaded as a CSV or as a file compatible with GPlates Desktop for users who want to keep working in the more capable offline environment. The interface exposes the underlying rotation models so researchers can cite which version they used, a small but important point: paleogeographic reconstructions are sensitive to the choice of model, and citing the model is what lets a result be checked.

The team built the service on GPlates, the open-source desktop platform developed over more than a decade at the University of Sydney and the California Institute of Technology, and on the pyGPlates Python library that wraps the same reconstruction engine for scripted use. Both have been community-maintained since the original EarthByte group at Sydney released them. By exposing that engine through a web API, the new portal extends the reach of a tool that has, until now, asked a lot of its users.

The standing problem of paleo-reconstruction

Reconstructing past positions is hard for structural reasons. Continents drift, ocean basins open and close, and the rotation files that describe the motion of each plate at each geological interval carry their own uncertainties. The dominant models today are built from a combination of marine magnetic anomalies, hotspot tracks, and paleomagnetic measurements, and they disagree with each other in ways that matter at the resolution of a single fossil site. Two researchers running the same present-day coordinates through two different rotation models can end up with paleocoordinates several hundred kilometres apart, particularly for intervals older than about 200 million years.

The new service does not paper over that problem. Its documentation flags model uncertainty explicitly and exposes the choice of rotation model as a parameter the user must select. That matters because the older pattern in some published work has been to run a single reconstruction, report the result, and leave the model choice implicit. A tool that forces the user to name the model in the output is, in a small way, a corrective.

Why the timing is right

Three converging pressures make a hosted service useful now. First, the data layers that reconstruction depends on have multiplied: paleoclimate simulations, species-occurrence databases, and basin-scale stratigraphic compilations all want to be cross-referenced against plate position, and pulling them into a desktop workflow is friction. Second, the community has converged on a small number of well-documented rotation models, so a hosted service can ship a curated set rather than ask each user to assemble their own. Third, the open-source footprint around GPlates has matured to the point where a public-facing service can lean on community maintenance rather than a single grant cycle.

There is a quiet geopolitics of this kind of infrastructure. Open plate-rotation models, free paleogeographic reconstructions, and community-run software have lowered the cost of doing palaeontology in countries that do not host the major western research laboratories. A graduate student in Quito, Nairobi, or Bandung can now run a reconstruction that, ten years ago, would have required either a sabbatical visit or a paid licence to commercial software. That is not a small thing for a discipline whose fossil record is global and whose collections sit disproportionately in the Global South.

What it does not solve

The tool does not resolve the underlying geological disagreements. It does not, on its own, improve the resolution of the rotation models, and it does not replace the local field knowledge that tells you whether a coordinate is sensible. A fossil's present-day location can be mapped back through time, but the reconstruction is only as good as the model chosen and the stratigraphic age assigned. None of that changes.

The published documentation is also candid about limits: the service depends on the hosted rotation files remaining available, the paleomagnetic and magnetic-anomaly inputs are unevenly distributed across the globe, and the older the target interval, the wider the uncertainty envelope becomes. For deep-time work, the output is a starting position, not an answer.

What to watch

The next test will be whether the community treats the service the way the documentation hopes: as an entry point that pushes users toward the desktop tool and the underlying rotation files, rather than as a black box that returns a coordinate without context. Adoption will be visible in the citations. If the bulk of new paleobiogeographic papers begin citing GPlates Web Service alongside the rotation-model version, the tool will have done what its authors intended. If it ends up cited in place of the underlying model, that will be a sign the documentation has not done enough work.

Desk note: This piece runs on a single primary source; the science desk preferred an early technical read on a release that the popular press has not yet picked up. Monexus will revisit with comparative coverage if the service is adopted at scale across published literature.