Spectral¶

class pyrt.observation.Spectral(short_wavelength, long_wavelength)[source]¶

A data structure that contains spectral information required by DISORT.

It accepts the short and long wavelengths from an observation and computes their corresponding wavenumbers. It can accommodate arrays of any shape as long as they have the same shape.

Parameters

short_wavelength (ArrayLike) – The short wavelength [microns] of each spectral bin.
long_wavelength (ArrayLike) – The long wavelength [microns] of each spectral bin.

Raises

TypeError – Raised if any values in the input arrays are nonnumerical.
ValueError – Raised if either of the input arrays are not the same shape, if any values in short_wavelength are not larger than the corresponding values in long_wavelength, or if either of the input arrays contain values outside of 0.1 to 50 microns (I assume this is the valid range to do retrievals).

Notes

If you do not plan to use thermal emission, there is probably little benefit to making an instance of this class. See ThermalEmission for a discussion on thermal radiation in DISORT.

See also

constant_width: Create instances of this class if the wavelengths are equally spaced.

Examples

Import the relevant modules

>>> import numpy as np
>>> from pyrt.observation import Spectral

Instantiate this class for a simple set of wavelengths.

>>> short = np.array([9, 10])
>>> long = short + 1
>>> Spectral(short, long)
Spectral:
   low_wavenumber = [1000.          909.09090909]
   high_wavenumber = [1111.11111111 1000.        ]

Instantiate this class for measurements taken at 1 to 30 microns in 1 micron increments, with each channel having a 50 nm spectral width.

>>> center = np.linspace(1, 30, num=30)
>>> half_width = 0.05 / 2
>>> wavelengths = Spectral(center - half_width, center + half_width)
>>> wavelengths.low_wavenumber.shape
(30,)

Instantiate this class for an image of shape (50, 60), where each pixel contains the same 20 wavelengths—wavelengths that span 1 to 20 microns in 1 micron increments with each channel having a 50 nm spectral width. Note that this is not memory efficient, as each pixel along the “measurement shape” contains the same information, but it may be useful for keeping as many arrays as possible equally shaped.

>>> center = np.linspace(1, 20, num=20)
>>> half_width = 0.05 / 2
>>> wav_grid = np.broadcast_to(center, (50, 60, 20))
>>> wavelengths = Spectral(wav_grid - half_width, wav_grid + half_width)
>>> wavelengths.low_wavenumber.shape
(50, 60, 20)
>>> np.array_equal(wavelengths.low_wavenumber[0, 0, :],
...                wavelengths.low_wavenumber[9, 17, :])
True

property high_wavenumber: numpy.ndarray¶

Get the high wavenumber [cm ^-1]—the wavenumber corresponding to short_wavelength.

Notes

Each element along the observation dimension(s) is named WVNMHI in DISORT. It is only needed by DISORT if thermal_emission is set to True.

property low_wavenumber: numpy.ndarray¶

Get the low wavenumber [cm ^-1]—the wavenumber corresponding to long_wavelength.

Notes

Each element along the observation dimension(s) is named WVNMLO in DISORT. It is only needed by DISORT if thermal_emission is set to True.

sky_image constant_width