.. This file is generated by scripts/render_dataset_docs.py. Do not edit by hand.

AEFA Forecast
=============

Synthetic-backed dense-grid forecasting adapter aligned to the AEFA earthquake forecasting workflow.

Overview
--------

AEFA Forecast is the public forecasting adapter used by the earthquake benchmark when exercising dense-grid wavefield forecasting models.

The current implementation is synthetic-backed, but it preserves the task shape, tensor layout, and reporting surface used by the shared earthquake evaluator.

At a Glance
-----------

.. list-table::
   :widths: 28 72
   :stub-columns: 1

   * - Provider
     - AEFA forecasting ecosystem surfaced through a PyHazards adapter
   * - Hazard Family
     - Earthquake
   * - Source Role
     - Forecast Benchmark
   * - Coverage
     - Benchmark-aligned earthquake forecasting samples
   * - Geometry
     - Dense-grid wavefield tensors
   * - Spatial Resolution
     - Benchmark-defined dense sensor grid
   * - Temporal Resolution
     - Short history and forecast windows
   * - Update Cadence
     - Generated locally for smoke and benchmark-alignment runs
   * - Period of Record
     - Synthetic-backed benchmark adapter
   * - Formats
     - PyTorch tensors via the dataset registry
   * - Registry Entry
     - ``aefa_forecast``

Data Characteristics
--------------------

- Multichannel dense-grid history tensors paired with future dense-grid targets.
- Registry-backed benchmark adapter rather than a raw external archive loader.
- Intended for forecasting-path validation and report generation.

Typical Use Cases
~~~~~~~~~~~~~~~~~

- Smoke tests for WaveCastNet-style earthquake forecasting.
- Shared forecasting benchmark runs under the earthquake evaluator.
- Validation of report exports aligned to the forecasting path.

Access
------

Use the links below to access the upstream source or its public documentation.

- `AEFA repository <https://github.com/chenyk1990/aefa>`_

PyHazards Usage
---------------

Use this adapter when you want the public earthquake forecasting benchmark surface rather than the private synthetic dataset name.

Registry Workflow
~~~~~~~~~~~~~~~~~

Primary dataset name: ``aefa_forecast``

.. code-block:: python

   from pyhazards.datasets import load_dataset

   data = load_dataset(
       "aefa_forecast",
       micro=True,
       temporal_in=5,
       temporal_out=4,
   ).load()

   train = data.get_split("train")
   print(train.inputs.shape, train.targets.shape)

- micro=True keeps the synthetic-backed forecasting path lightweight for validation.

Related Coverage
~~~~~~~~~~~~~~~~

**Benchmarks:** :doc:`Earthquake Benchmark </benchmarks/earthquake_benchmark>`, :doc:`AEFA </benchmarks/aefa>`

**Representative Models:** :doc:`WaveCastNet </modules/models_wavecastnet>`

Inspection Workflow
-------------------

This dataset is currently surfaced as a registry-backed benchmark adapter,
so there is no standalone inspection CLI documented for it.

Notes
-----

- This is a benchmark adapter, not a full external AEFA ingestion pipeline.

Reference
---------

- `AEFA <https://github.com/chenyk1990/aefa>`_.