Version 0.4 Updates¶

Version 0.4.3¶

Dependency updates¶

Ensure compatibility between earthkit components by restricting the following dependencies:

earthkit-utils<0.2

Version 0.4.1¶

Array formats¶

Made almost all of the methods array format agnostic with support for Numpy, Torch and CuPy arrays/tensors as an input. The array backend is automatically detected from the input data type.

from earthkit.meteo import thermo

# Example with Numpy array
import numpy as np

t = np.array([264.12, 261.45])
p = np.array([850, 850]) * 100.0
theta = thermo.potential_temperature(t, p)

# Example with Torch tensor
import torch

t = torch.tensor([264.12, 261.45])
p = torch.tensor([850.0, 850.0]) * 100.0
theta = thermo.potential_temperature(t, p)

New features¶

Added the earthkit.meteo.vertical submodule
Added the earthkit.meteo.stats.array.value_to_return_period() and earthkit.meteo.stats.array.return_period_to_value() methods. They are based on a Gumbel-distribution fit to the sample data (#29). See the following notebook example:
- /how-tos/return_period.ipynb.
Added earthkit.meteo.score.array.correlation.pearson() to compute the Pearson correlation over fields.
Implemented the symmetric Compute Crossing Point Forecast (CPF) method earthkit.meteo.extreme.array.cpf() (#28).
Added earthkit.meteo.thermo.array.specific_gas_constant() to compute the specific gas constant for moist air
Enabled earthkit.meteo.solar.array.julian_day() to use timezone aware datetime objects as input (#41).
Added the nan_policy option to earthkit.meteo.score.array.crps() to handle nans. The possible values are: “raise”, “propagate”, and “omit”. The default is “propagate”. (#45)

New dependencies¶

earthkit-utils >= 0.0.1