Open source / Coming soon
Earth’s complexity.
Less time to compute.
We’re rewriting snow and fire simulation models in Rust, accelerated by GPUs, to make detailed physical modeling more practical at planetary scale.
pi-snow
Rust + GPU / SNOWPACK rewrite
SNOWPACK models how snow cover evolves with the weather, tracking its layers, microstructure, and exchanges of mass and energy with the atmosphere and soil. It supports applications from avalanche forecasting to hydrology and climate research.
Faster snow simulations make it more practical to model more locations and seasons, expanding the snow component of large-scale Earth models.
pi-fire
Rust + GPU / FDS rewrite
NIST’s Fire Dynamics Simulator (FDS) models low-speed fluid flows, with an emphasis on the movement of smoke and heat from fires. Its companion tool, Smokeview, visualizes simulation results.
Faster fire simulations make it more practical to explore more fire, heat, and smoke scenarios as part of larger Earth-modeling workflows.
Performance figures reflect our testing so far at full precision, not a guarantee across all workloads or hardware. Lower-precision testing is currently underway for both projects; final testing is still in progress. Source code will be available after the planned open-source release.
Why speed matters
More scenarios. A bigger picture.
Building Earth models at scale means simulating many places, time periods, and possible futures. Faster physical models can put more of those runs within a given compute budget, enabling broader coverage, larger ensembles to explore uncertainty, and quicker iteration on how Earth’s systems fit together.
Looking ahead
More to come later in 2026.
Autonomous, collaborative science data aggregation, an open data project, and more. Stay tuned for further announcements.
For future open-source work, we’re also interested in exploring how tools and shared protocols could connect human decisions, ecological knowledge, and AI. Long Now Labs’ Interspecies Protocols is a source of inspiration for this direction—not an announced collaboration or a committed release.