PyNRLMSISE-00

Python interface for the NRLMSISE-00 empirical neutral atmosphere model

This python version of the NRLMSISE00 upper atmosphere model is based on the C-version of the code, available at https://www.brodo.de/space/nrlmsise. The C code is imported as a git submodule from git://git.linta.de/~brodo/nrlmsise-00.git (browsable version at: https://git.linta.de/?p=~brodo/nrlmsise-00.git).

:warning: This python interface is in the beta stage, that is, it should work but may still have some bugs. The interface is supposed to be stable but may still change slightly in future versions.

Documentation can be found at https://pynrlmsise00.readthedocs.io, and development happens on github https://github.com/st-bender/pynrlmsise00.

Quote from https://ccmc.gsfc.nasa.gov/models/modelinfo.php?model=MSISE:

“The MSISE model describes the neutral temperature and densities in Earth's atmosphere from ground to thermospheric heights. The NRLMSIS-00 empirical atmosphere model was developed by Mike Picone, Alan Hedin, and Doug Drob.”

Install

Requirements

• numpy - required
• spaceweather and xarray - optional, for the datatset sub-package, see below
• pytest - optional, for testing
• sphinx - optional, to build the documentation

To compile the C source code, additional system header files may be required. For example on Debian/Ubuntu Linux, the package libc6-dev is needed.

pynrlmsise00

A pip package called nrlmsise00 is available from the main package repository, and can be installed with:

$ pip install nrlmsise00

In some cases this will install from the source package and the note above about the additional requirements applies.

As binary package support is limited, pynrlmsise00 can be installed with pip directly from github (see https://pip.pypa.io/en/stable/reference/pip_install/#vcs-support and https://pip.pypa.io/en/stable/reference/pip_install/#git):

$ pip install [-e] git+https://github.com/st-bender/pynrlmsise00.git

The other option is to use a local clone:

$ git clone https://github.com/st-bender/pynrlmsise00.git
$ cd pynrlmsise00
$ git submodule init
$ git submodule update

and then using pip (optionally using -e, see https://pip.pypa.io/en/stable/reference/pip_install/#install-editable):

$ pip install [-e] .

or using setup.py:

$ python setup.py install

Optionally, test the correct function of the module with

$ py.test [-v]

or even including the doctests in this document:

$ py.test [-v] --doctest-glob='*.md'

Usage

The python module itself is named nrlmsise00 and is imported as usual:

>>> import nrlmsise00

Basic class and method documentation is accessible via pydoc:

$ pydoc nrlmsise00

Python interface

The Python interface functions take datetime.datetime objects for convenience. The local solar time is calculated from that time and the given location, but it can be set explicitly via the lst keyword. The returned value has the same format as the original C version (see below). Because of their similarity, gtd7() and gtd7d() are selected via the method keyword, gtd7 is the default.

The return values are tuples of two lists containing the densities (d[0]--d[8]) and temperatures (t[0], t[1]).

The output has the same order as the C reference code, in particular:

• d[0] - He number density [cm⁻³]
• d[1] - O number density [cm⁻³]
• d[2] - N2 number density [cm⁻³]
• d[3] - O2 number density [cm⁻³]
• d[4] - Ar number density [cm⁻³]
• d[5] - total mass density [g cm⁻³]) (includes d[8] in gtd7d())
• d[6] - H number density [cm⁻³]
• d[7] - N number density [cm⁻³]
• d[8] - Anomalous oxygen number density [cm⁻³]
• t[0] - exospheric temperature [K]
• t[1] - temperature at alt [K]

The flags and ap_a value array are set via keywords, but both default to the standard setting, such that changing them should not be necessary for most use cases. For example setting flag[0] to 1 changes the output to metres and kilograms instead of centimetres and grams (0 is the default).

>>> from datetime import datetime
>>> from nrlmsise00 import msise_model
>>> msise_model(datetime(2009, 6, 21, 8, 3, 20), 400, 60, -70, 150, 150, 4, lst=16)
([666517.690495152, 113880555.97522168, 19982109.255734544, 402276.3585712511, 3557.464994515886, 4.074713532757222e-15, 34753.12399717142, 4095913.2682930017, 26672.73209335869], [1250.5399435607994, 1241.4161300191206])

NumPy interface

A numpy compatible flat version is available as msise_flat(), it returns a 11-element numpy.ndarray with the densities in the first 9 entries and the temperatures in the last two entries. That is ret = numpy.ndarray([d[0], ..., d[8], t[0], t[1]]).

>>> from datetime import datetime
>>> from nrlmsise00 import msise_flat
>>> msise_flat(datetime(2009, 6, 21, 8, 3, 20), 400, 60, -70, 150, 150, 4)
array([5.65085279e+05, 6.79850175e+07, 1.18819263e+07, 2.37030166e+05,
       1.32459684e+03, 2.39947892e-15, 5.32498381e+04, 1.07596246e+06,
       2.66727321e+04, 1.10058413e+03, 1.09824872e+03])

All arguments can be numpy.ndarrays, but must be broadcastable to a common shape. For example to calculate the values for three altitudes (200, 300, and 400 km) and two latitude locations (60 and 70 °N) simultaneously, one can use numpy.newaxis (which is equal to None) like this:

>>> from datetime import datetime
>>> import numpy as np
>>> from nrlmsise00 import msise_flat
>>> alts = np.arange(200, 401, 100.)  # = [200, 300, 400] [km]
>>> lats = np.arange(60, 71, 10.)  # = [60, 70] [°N]
>>> # Using broadcasting, the output will be a 2 x 3 x 11 element array:
>>> msise_flat(datetime(2009, 6, 21, 8, 3, 20), alts[None, :], lats[:, None], -70, 150, 150, 4)
array([[[1.36949418e+06, 1.95229496e+09, 3.83824808e+09, 1.79130515e+08,
         4.92145034e+06, 2.40511268e-13, 8.34108685e+04, 1.74317585e+07,
         3.45500931e-08, 1.10058413e+03, 9.68827485e+02],
        [8.40190601e+05, 3.25739060e+08, 1.82477392e+08, 5.37973134e+06,
         6.53609278e+04, 1.75304136e-14, 5.92944463e+04, 4.36516218e+06,
         1.03939126e+02, 1.10058413e+03, 1.08356514e+03],
        [5.65085279e+05, 6.79850175e+07, 1.18819263e+07, 2.37030166e+05,
         1.32459684e+03, 2.39947892e-15, 5.32498381e+04, 1.07596246e+06,
         2.66727321e+04, 1.10058413e+03, 1.09824872e+03]],
<BLANKLINE>
       [[1.10012225e+06, 1.94725472e+09, 4.08547233e+09, 1.92320077e+08,
         6.65460281e+06, 2.52846563e-13, 6.16745965e+04, 2.45012145e+07,
         5.21846603e-08, 1.13812434e+03, 1.00132640e+03],
        [6.83809952e+05, 3.42643970e+08, 2.13434661e+08, 6.43426889e+06,
         1.01162173e+05, 1.95300073e-14, 4.36031132e+04, 6.70490625e+06,
         1.59911615e+02, 1.13812434e+03, 1.12084651e+03],
        [4.65787225e+05, 7.52160226e+07, 1.51795904e+07, 3.13560147e+05,
         2.32541183e+03, 2.76353370e-15, 3.92811827e+04, 1.73321928e+06,
         4.12296154e+04, 1.13812434e+03, 1.13580463e+03]]])

Xarray Dataset interface

Output to a 4-D xarray.Dataset is supported via the dataset submodule which can be installed with:

pip install [-U] 'nrlmsise00[dataset]'

This module provides a 4-D version msise_4d() to broadcast the 1-D inputs for time, altitude, latitude, and longitude. It also uses the spaceweather package by default to automatically obtain the geomagnetic and Solar flux indices. The variable names are set according to the MSIS output.

>>> from datetime import datetime
>>> from nrlmsise00.dataset import msise_4d
>>> alts = np.arange(200, 401, 100.)  # = [200, 300, 400] [km]
>>> lats = np.arange(60, 71, 10.)  # = [60, 70] [°N]
>>> lons = np.arange(-70., 71., 35.)  # = [-70, -35,  0, 35, 70] [°E]
>>> # broadcasting is done internally
>>> ds = msise_4d(datetime(2009, 6, 21, 8, 3, 20), alts, lats, lons)
>>> ds
<xarray.Dataset>
Dimensions:  (alt: 3, lat: 2, lon: 5, time: 1)
Coordinates:
  * time     (time) datetime64[ns] 2009-06-21T08:03:20
  * alt      (alt) float64 200.0 300.0 400.0
  * lat      (lat) float64 60.0 70.0
  * lon      (lon) float64 -70.0 -35.0 0.0 35.0 70.0
Data variables:
    He       (time, alt, lat, lon) float64 8.597e+05 1.063e+06 ... 4.936e+05
    O        (time, alt, lat, lon) float64 1.248e+09 1.46e+09 ... 2.635e+07
    N2       (time, alt, lat, lon) float64 2.555e+09 2.654e+09 ... 1.667e+06
    O2       (time, alt, lat, lon) float64 2.1e+08 2.062e+08 ... 3.471e+04
    Ar       (time, alt, lat, lon) float64 3.16e+06 3.287e+06 ... 76.55 67.16
    rho      (time, alt, lat, lon) float64 1.635e-13 1.736e-13 ... 7.984e-16
    H        (time, alt, lat, lon) float64 3.144e+05 3.02e+05 ... 1.237e+05
    N        (time, alt, lat, lon) float64 9.095e+06 1.069e+07 ... 6.765e+05
    AnomO    (time, alt, lat, lon) float64 1.173e-08 1.173e-08 ... 1.101e+04
    Texo     (time, alt, lat, lon) float64 805.2 823.7 807.1 ... 818.7 821.2
    Talt     (time, alt, lat, lon) float64 757.9 758.7 766.4 ... 818.7 821.1
    lst      (time, lon) float64 3.389 5.722 8.056 10.39 12.72
    Ap       (time) int32 6
    f107     (time) float64 66.7
    f107a    (time) float64 69.0

C model interface

The C submodule directly interfaces the model functions gtd7() and gtd7d() by importing nrlmsise00._nrlmsise00.

>>> from nrlmsise00._nrlmsise00 import gtd7, gtd7d
>>> # using the standard flags
>>> gtd7(2009, 172, 29000, 400, 60, -70, 16, 150, 150, 4)
([666517.690495152, 113880555.97522168, 19982109.255734544, 402276.3585712511, 3557.464994515886, 4.074713532757222e-15, 34753.12399717142, 4095913.2682930017, 26672.73209335869], [1250.5399435607994, 1241.4161300191206])

This module also provides "flat" variants of the C functions as gtd7_flat() and gtd7d_flat(). For example using gtd7() the same way as above:

>>> import numpy as np
>>> from nrlmsise00 import gtd7_flat
>>> alts = np.arange(200, 401, 100.)  # = [200, 300, 400] [km]
>>> lats = np.arange(60, 71, 10.)  # = [60, 70] [°N]
>>> # Using broadcasting, the output will be a 2 x 3 x 11 element array:
>>> gtd7_flat(2009, 172, 29000, alts[None, :], lats[:, None], -70, 16, 150, 150, 4)
array([[[1.55567936e+06, 2.55949597e+09, 4.00342724e+09, 1.74513806e+08,
         6.56916263e+06, 2.64872982e-13, 5.63405578e+04, 4.71893934e+07,
         3.45500931e-08, 1.25053994e+03, 1.02704994e+03],
        [9.58507714e+05, 4.66979460e+08, 2.31041924e+08, 6.58659651e+06,
         1.16566762e+05, 2.38399390e-14, 3.86535595e+04, 1.43755262e+07,
         1.03939126e+02, 1.25053994e+03, 1.20645403e+03],
        [6.66517690e+05, 1.13880556e+08, 1.99821093e+07, 4.02276359e+05,
         3.55746499e+03, 4.07471353e-15, 3.47531240e+04, 4.09591327e+06,
         2.66727321e+04, 1.25053994e+03, 1.24141613e+03]],
<BLANKLINE>
       [[1.31669842e+06, 2.40644124e+09, 4.21778196e+09, 1.89878716e+08,
         8.17662024e+06, 2.71788520e-13, 4.64192484e+04, 5.13265845e+07,
         5.21846603e-08, 1.24246351e+03, 1.04698385e+03],
        [8.22632403e+05, 4.52803942e+08, 2.53857090e+08, 7.50201654e+06,
         1.53431033e+05, 2.46179628e-14, 3.20594861e+04, 1.62651506e+07,
         1.59911615e+02, 1.24246351e+03, 1.20963726e+03],
        [5.73944168e+05, 1.10836468e+08, 2.19925518e+07, 4.58648922e+05,
         4.68600377e+03, 4.10277781e-15, 2.89330169e+04, 4.65636025e+06,
         4.12296154e+04, 1.24246351e+03, 1.23665288e+03]]])

Note

All functions require the solar 10.7 cm radio flux and and the geomagnetic Ap index values to produce correct results. In particular, according to the C source code:

• f107A: 81 day average of F10.7 flux (centered on the given day of year)
• f107: daily F10.7 flux for previous day
• ap: magnetic index (daily)

The f107 and f107A values used to generate the model correspond to the 10.7 cm radio flux at the actual distance of the Earth from the Sun rather than the radio flux at 1 AU. The following site provides both classes of values (outdated): ftp://ftp.ngdc.noaa.gov/STP/SOLAR_DATA/SOLAR_RADIO/FLUX/

More up-to-date index files can be found at https://celestrak.com/SpaceData/, which are also provided by the spaceweather package.

f107, f107A, and ap effects are neither large nor well established below 80 km and these parameters should be set to 150., 150., and 4. respectively.

License

This python interface is free software: you can redistribute it or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, version 2 (GPLv2), see local copy or online version.

The C source code of NRLMSISE-00 is in the public domain, see COPYING.NRLMSISE-00.