SmokeForecast

In Beta through H2 2026.

A mobile application giving prescribed burn managers detailed weather forecasting alongside smoke plume dynamics, so burns can be planned and scheduled with confidence.

SmokeForecast automates and tracks the hardest part of prescribed burn planning — smoke plume estimation — using live weather data, satellite imagery, and contemporary plume dynamics modeling.

Easy-Distributed-CUF

An extension to the NVIDIA Collective Communications Library (NCCL) that reduces the effort required to scale CUDA Fortran applications across NVIDIA HPC clusters.

Scaling a Fortran model past a single node usually means hand-writing the distribution logic: enumerating the hardware, checking that every device is healthy before a long run starts, and hard-coding halo widths that have to be revisited every time the code moves to a different machine. EDC handles that layer for you, providing:

• Automatic hardware discovery — the library enumerates available devices and interconnects rather than requiring them to be declared up front
• Hardware health checks — devices are validated before a run begins, so a failing GPU surfaces immediately instead of twelve hours into a job
• Automatic halo sizing — exchange regions are sized against the compute architecture actually present, so the same source runs on a workstation and a cluster without modification

The result is CUDA Fortran that scales without the distribution code becoming a second project.

QF-Bin-Compare

Analysis tooling for sets of QUIC-Fire binary output files, built to answer a question that is otherwise tedious: what actually changed?

QUIC-Fire is the coupled fire-atmosphere model developed by Los Alamos National Laboratory with the USDA Forest Service and Tall Timbers Research Station, and it has become a standard tool for prescribed fire planning. Its outputs are raw binaries, which makes comparison across runs awkward at exactly the moments it matters most.

qf-bin-compare reads those outputs directly and surfaces the differences, so users can:

• Compare ensemble runs — quickly identify where members of an ensemble diverge, and by how much
• Compare across versions — understand what changed in model output between QUIC-Fire releases, and confirm that a version bump did what was expected

Useful for anyone validating a model configuration, tracking down an unexpected result, or documenting reproducibility for a publication or burn plan.