Google Flood Forecasting¶

Overview¶

google_flood_forecasting is a compact sequence-to-node forecasting baseline for flood streamflow prediction.

At a Glance¶

Hazard Family

Flood

Public catalog grouping used for this model.

Maturity

Implemented

Catalog maturity label used on the index page.

Tasks

Streamflow

Benchmark Family

Flood Benchmark

Primary benchmark-family link used for compatible evaluation coverage.

Description¶

google_flood_forecasting is a compact sequence-to-node forecasting baseline for flood streamflow prediction.

The PyHazards implementation uses a transformer encoder over per-node history windows and returns one forecast value per node.

Benchmark Compatibility¶

Primary benchmark family: Flood Benchmark

Mapped benchmark ecosystems: Caravan

External References¶

Paper: Global Flood Forecasting at a Fine Catchment Resolution using Machine Learning | Repo: Repository

Registry Name¶

Primary entrypoint: google_flood_forecasting

Supported Tasks¶

Streamflow

Programmatic Use¶

import torch
from pyhazards.models import build_model

model = build_model(
    name="google_flood_forecasting",
    task="regression",
    input_dim=2,
    out_dim=1,
    history=4,
)
preds = model({"x": torch.randn(2, 4, 6, 2)})
print(preds.shape)

Notes¶

The smoke path uses the same streamflow-style graph fixture as the other flood baselines.