Models¶
Browse PyHazards model implementations across hazard families, compare scope and maturity, and navigate to model-specific detail pages.
At a Glance¶
4
Catalog tabs grouped by the normalized public hazard taxonomy.
24
Public core baselines plus additional implemented variants.
3
Prototype weather-model integrations kept separate from the stable catalog.
27
Models with explicit benchmark-family or ecosystem links on this page.
Catalog by Hazard¶
Use the hazard tabs below to browse the public catalog. Each card keeps the index-page summary short, then links into model-specific detail pages and compatible benchmark coverage.
Wildfire models cover danger forecasting, weekly activity forecasting, and spread prediction under the shared wildfire benchmark family.
Implemented Models
This table includes both core baselines and public variants or additional implementations for the hazard family.
A temporal convolution baseline for weekly wildfire activity forecasting.
Wildfire Forecasting Implemented
A two-stage wildfire framework with a DNN stage for incident-level cause and size prediction plus an LSTM + autoencoder stage for weekly forecasting.
Wildfire Classification Forecasting Implemented
A compact encoder-decoder baseline for wildfire spread mask prediction.
Wildfire Spread Implemented
A lightweight simulator-style wildfire spread adapter inspired by front-propagation systems.
Wildfire Spread Implemented
A sequence forecasting baseline for next-window wildfire activity across weekly count features.
Wildfire Forecasting Implemented
A temporal convolution wildfire spread baseline over short raster history windows.
Wildfire Spread Implemented
A lightweight raster wildfire spread adapter inspired by WRF-SFIRE style transport.
Wildfire Spread Implemented
An explainable CNN segmentation model with an ASPP mechanism for next-day wildfire spread prediction.
Wildfire Spread Implemented
Earthquake models span phase picking and dense-grid forecasting, with detail pages linked to the shared earthquake benchmark coverage.
Implemented Models
This table includes both core baselines and public variants or additional implementations for the hazard family.
A transformer-style earthquake phase-picking baseline for modern sequence modeling comparisons.
Earthquake Phase Picking Implemented
A bidirectional sequence encoder for joint earthquake phase picking with attention pooling over waveform windows.
Earthquake Phase Picking Implemented
A compact CNN baseline for generalized phase detection and historical earthquake picking comparisons.
Earthquake Phase Picking Implemented
Paper: Generalized Seismic Phase Detection with Deep Learning | Repo: Repository
A lightweight phase-picking baseline that predicts P- and S-arrival indices from multichannel waveform windows.
Earthquake Phase Picking Implemented
A ConvLEM-based sequence-to-sequence model for dense-grid earthquake wavefield forecasting and early-warning style rollout experiments.
Earthquake Wavefield Forecasting Implemented
Flood models cover streamflow and inundation forecasting, ranging from sequence baselines to dense-grid flood-mapping architectures.
Implemented Models
This table includes both core baselines and public variants or additional implementations for the hazard family.
An entity-aware hydrology baseline with static-feature gating over streamflow histories.
Flood Streamflow Implemented
A compact spatiotemporal flood-inundation baseline for raster forecast experiments.
Flood Inundation Implemented
A transformer-style sequence baseline for nodewise streamflow forecasting.
Flood Streamflow Implemented
An adapter-style LSTM baseline for nodewise streamflow forecasting on graph-temporal inputs.
Flood Streamflow Implemented
A U-Net style urban inundation baseline for dense-grid flood prediction.
Flood Inundation Implemented
A physics-informed graph neural network for flood forecasting with interpretable KAN-style components, residual message passing, and delta-state decoding.
Flood Streamflow Implemented
Storm models are organized under one tropical-cyclone family, including basin-specific hurricane baselines and shared all-basin forecasting models.
Implemented Models
This table includes both core baselines and public variants or additional implementations for the hazard family.
A compact multimodal storm baseline for hurricane track and intensity forecasting.
Tropical Cyclone Track + Intensity Implemented
A spatiotemporal tropical-cyclone baseline with an intensity-focused head and shared trajectory output.
Tropical Cyclone Track + Intensity Implemented
A knowledge-guided fusion baseline for tropical cyclone track and intensity forecasting.
Tropical Cyclone Track + Intensity Implemented
A compact MLP baseline for hurricane track and intensity forecasting.
Tropical Cyclone Track + Intensity Implemented
A GRU plus attention baseline for all-basin tropical cyclone forecasting.
Tropical Cyclone Track + Intensity Implemented
Experimental Adapters
These entries remain public as lightweight wrapper or prototype integrations and should not be counted as stable implemented methods.
An experimental wrapper-style storm adapter inspired by FourCastNet forecast fields.
Tropical Cyclone Track + Intensity Experimental Adapter
An experimental wrapper-style storm adapter inspired by GraphCast/GenCast forecast fields.
Tropical Cyclone Track + Intensity Experimental Adapter
An experimental wrapper-style storm adapter inspired by Pangu-Weather forecast fields.
Tropical Cyclone Track + Intensity Experimental Adapter
Recommended Entry Points¶
If you are new to PyHazards, these four models provide the clearest starting point for each hazard family.
Start with: FireCastNet
A compact encoder-decoder baseline for wildfire spread mask prediction.
Benchmark: Wildfire Benchmark
Start with: PhaseNet
A lightweight phase-picking baseline that predicts P- and S-arrival indices from multichannel waveform windows.
Benchmark: Earthquake Benchmark
Start with: FloodCast
A compact spatiotemporal flood-inundation baseline for raster forecast experiments.
Benchmark: Flood Benchmark
Start with: Hurricast
A compact multimodal storm baseline for hurricane track and intensity forecasting.
Benchmark: Tropical Cyclone Benchmark
Programmatic Use¶
Use pyhazards.models package for the developer registry workflow, builder examples, and package-level API lookup. Use Benchmarks to compare compatible benchmark families before selecting a model for evaluation.