perturbation.BredVector

Import path: earth2studio.perturbation.BredVector

View source on GitHub

Documentation

Bred Vector perturbation method, a classical technique for pertubations in ensemble forecasting.

Parameters:

• model (Callable[[Tensor], Tensor]) –

  Dynamical model, typically this is the prognostic AI model. TODO: Update to prognostic looper

• noise_amplitude (float | Tensor, default: 0.05 ) –

  Noise amplitude, by default 0.05. If a tensor, this must be broadcastable with the input data.

• integration_steps (int, default: 20 ) –

  Number of integration steps to use in forward call, by default 20

• ensemble_perturb (bool, default: False ) –

  Perturb the ensemble in an interacting fashion, by default False

• seeding_perturbation_method (Perturbation, default: Brown() ) –

  Method to seed the Bred Vector perturbation, by default Brown Noise

Note

For additional information:

• journals.ametsoc.org/view/journals/bams/74/12/1520-0477_1993_074_2317_efantg_2_0_co_2.xml
• en.wikipedia.org/wiki/Bred_vector

__call__

__call__(
    x: Tensor, coords: CoordSystem
) -> tuple[Tensor, CoordSystem]

Apply perturbation method

Parameters:

• x (Tensor) –

  Input tensor intended to apply perturbation on

• coords (CoordSystem) –

  Ordered dict representing coordinate system that describes the tensor

Returns:

• tuple[torch.Tensor, CoordSystem]: –

  Output tensor and respective coordinate system dictionary