sciapy 0.0.7

Python tools for (some) SCIAMACHY data

SCIAMACHY data tools

Overview

These SCIAMACHY tools are provided as convenience tools for handling SCIAMACHY level 1c limb spectra and retrieved level 2 trace-gas densities.

More extensive documentation is provided on sciapy.rtfd.io.

Level 1c tools

The sciapy.level1c submodule provides a few conversion tools for SCIAMACHY level 1c calibrated spectra, to be used as input for trace gas retrieval with scia_retrieval_2d.

Note that this is not a level 1b to level 1c calibration tool.

For calibrating level 1b spectra (for example SCI_NL__1P version 8.02 provided by ESA via the ESA data browser) to level 1c spectra, use the SciaL1C command line tool or the free software nadc_tools. The first produces .child files, the second can output to HDF5 (.h5).

Further note: .child files are currently not supported.

Level 2 tools

The sciapy.level2 submodule provides post-processing tools for trace-gas densities retrieved from SCIAMACHY limb scans. Support simple operations as combining files into netcdf, calculating and noting local solar time at the retrieval grid points, geomagnetic latitudes, etc.

The level 2 tools also include a simple binning algorithm.

Regression

The sciapy.regress submodule can be used for regression analysis of SCIAMACHY level 2 trace gas density time series, either directly or as daily zonal means. The regression tools support various parameter fitting methods using scipy.optimize and uncertainty evaluation using Markov-Chain Monte-Carlo sampling with emcee. Further supports covariance modelling via celerite and george.

Install

Prerequisites

Sciapy uses features from a lot of different packages. All dependencies will be automatically installed when using pip install or python setup.py, see below. However, to speed up the install or for use within a conda environment, it may be advantageous to install some of the important packages beforehand:

• numpy at least version 1.13.0 for general numerics,
• scipy at least version 0.17.0 for scientific numerics,
• matplotlib at least version 2.2 for plotting,
• netCDF4 for the low level netcdf4 interfaces,
• h5py for the low level hdf5 interfaces,
• dask,
• toolz,
• pandas and
• xarray for the higher level data interfaces,
• astropy for (astronomical) time conversions,
• parse for ASCII text parsing in level1c,
• pybind11 C++ interface needed by celerite
• celerite at least version 0.3.0 and
• george for Gaussian process modelling,
• emcee for MCMC sampling and
• corner for the sample histogram plots.

Out of these packages, numpy is probably the most important one to be installed first because at least celerite needs it for setup. It may also be a good idea to install pybind11 because both celerite and george use its interface, and both may fail to install without pybind11.

Depending on the setup, numpy and pybind11 can be installed via pip:

pip install numpy pybind11

or conda:

conda install numpy pybind11

sciapy

There is no pip package yet, but sciapy 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/sciapy.git

The other option is to use a local clone:

$ git clone https://github.com/st-bender/sciapy.git
$ cd sciapy

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

Usage

The whole module as well as the individual submodules can be loaded as usual:

>>> import sciapy
>>> import sciapy.level1c
>>> import sciapy.level2
>>> import sciapy.regress

Basic class and method documentation is accessible via pydoc:

$ pydoc sciapy

The submodules' documentation can be accessed with pydoc as well:

$ pydoc sciapy.level1c
$ pydoc sciapy.level2
$ pydoc sciapy.regress