GraphCast TC Adapter¶

Overview¶

graphcast_tc is an experimental foundation-weather adapter that keeps the shared storm trajectory interface while remaining lightweight enough for CI.

At a Glance¶

Hazard Family

Tropical Cyclone

Public catalog grouping used for this model.

Maturity

Experimental Adapter

Catalog maturity label used on the index page.

Tasks

Track + Intensity

Benchmark Family

Tropical Cyclone Benchmark

Primary benchmark-family link used for compatible evaluation coverage.

Description¶

graphcast_tc is an experimental foundation-weather adapter that keeps the shared storm trajectory interface while remaining lightweight enough for CI.

The PyHazards version is intentionally wrapper-style and should be treated as an adapter contract rather than a full reproduction of the original weather model.

Benchmark Compatibility¶

Primary benchmark family: Tropical Cyclone Benchmark

Mapped benchmark ecosystems: IBTrACS

External References¶

Paper: GraphCast: Learning skillful medium-range global weather forecasting | Repo: Repository

Registry Name¶

Primary entrypoint: graphcast_tc

Supported Tasks¶

Track + Intensity

Programmatic Use¶

import torch
from pyhazards.models import build_model

model = build_model(name="graphcast_tc", task="regression", input_dim=8, horizon=5)
preds = model(torch.randn(2, 6, 8))
print(preds.shape)

Notes¶

Experimental adapter: intended for shared-evaluator prototyping rather than exact weather-model parity.