Astrobase is a Python package for analyzing light curves and finding variable stars. It includes implementations of several period-finding algorithms, batch work drivers for working on large collections of light curves, and a web-app useful for reviewing and classifying light curves by stellar variability type. This package was spun out of a bunch of Python modules I wrote and maintain for my work with the HAT Exoplanet Surveys. It’s applicable to many other astronomical time-series observations, and includes support for the light curves produced by Kepler and TESS in particular.

Most functions in this package that deal with light curves (e.g. in the modules astrobase.lcfit, astrobase.lcmath, astrobase.periodbase, astrobase.plotbase, astrobase.checkplot) usually require three Numpy ndarrays as input: times, mags, and errs, so they should work with any time-series data that can be represented in this form. If you have flux time series measurements, most functions also take a magsarefluxes keyword argument that makes them handle flux light curves correctly.

The astrobase.lcproc subpackage implements drivers for working on large collections of light curve files, and includes functions to register your own light curve format so that it gets recognized and can be worked on by other Astrobase functions transparently.

  • Guides for specific tasks are available as Jupyter notebooks at Github: astrobase-notebooks.
  • The full API documentation generated automatically from the docstrings by Sphinx is available.
  • The code for Astrobase is maintained at Github.

Install Astrobase from PyPI using pip:

# preferably in a virtualenv

# install Numpy to compile Fortran dependencies
$ pip install numpy

# install astrobase
$ pip install astrobase

Package contents

  • astrobase.astrokep: contains functions for dealing with Kepler and K2 Mission light curves from STScI MAST (reading the FITS files, consolidating light curves for objects over quarters), and some basic operations (converting fluxes to mags, decorrelation of light curves, filtering light curves, and fitting object centroids for eclipse analysis, etc.)
  • astrobase.astrotess: contains functions for dealing with TESS 2-minute cadence light curves from STScI MAST (reading the FITS files, consolidating light curves for objects over sectors), and some basic operations (converting fluxes to mags, filtering light curves, etc.)
  • astrobase.hatsurveys: modules to read, filter, and normalize light curves from various HAT surveys.

This package contains parallelized implementations of several period-finding algorithms.

  • astrobase.lcfit: functions for fitting various light curve models to observations, including sinusoidal, trapezoidal and full Mandel-Agol planet transits, eclipses, and splines.
  • astrobase.lcmath: functions for light curve operations such as phasing, normalization, binning (in time and phase), sigma-clipping, external parameter decorrelation (EPD), etc.
  • astrobase.lcmodels: modules that contain simple models for several variable star classes, including sinusoidal variables, eclipsing binaries, and transiting planets. Useful for fitting these with the functions in the astrobase.lcfit module.
  • astrobase.varbase: functions for dealing with periodic signals including masking and pre-whitening them, ACF calculations, light curve detrending, and specific tools for planetary transits.
  • astrobase.plotbase: functions to plot light curves, phased light curves, periodograms, and download Digitized Sky Survey cutouts from the NASA SkyView service.
  • astrobase.lcproc: driver functions for running an end-to-end pipeline including: (i) object selection from a collection of light curves by position, cross-matching to external catalogs, or light curve objectinfo keys, (ii) running variability feature calculation and detection, (iii) running period-finding, and (iv) object review using the checkplotserver webapp for variability classification. This also contains an Amazon AWS-enabled lcproc implementation.
  • astrobase.checkplot: contains functions to make checkplots: a grid of plots used to quickly decide if a period search for a possibly variable object was successful. Checkplots come in two forms:

    Python pickles: If you want to interactively browse through large numbers of checkplots (e.g., as part of a large variable star classification project), you can use the checkplotserver webapp that works on checkplot pickle files. This interface allows you to review all phased light curves from all period-finder methods applied, set and save variability tags, object type tags, best periods and epochs, and comments for each object using a browser-based UI (see below). The information entered can then be exported as CSV or JSON for the next stage of a variable star classification pipeline.

    PNG images: Alternatively, if you want to simply glance through lots of checkplots (e.g. for an initial look at a collection of light curves), there’s a checkplot-viewer webapp available that operates on checkplot PNG images.

screenshots of the checkplotserver web application
  • astrobase.cpserver: contains the implementation of the checkplotserver webapp to review, edit, and export information from checkplot pickles produced as part of a variable star classification effort run on a large light curve collection. Also contains the more light-weight checkplot-viewer webapp to glance through large numbers of checkplot PNGs.
  • astrobase.varclass: functions for calculating various variability, stellar color and motion, and neighbor proximity features, along with a Random Forest based classifier.
  • modules and functions to query various astronomical catalogs and data services, including GAIA, SIMBAD, TRILEGAL, NASA SkyView, and 2MASS DUST.

Other useful bits


  • astrobase.coordutils: functions for dealing with coordinates (conversions, distances, proper motion).
  • astrobase.timeutils: functions for converting from Julian dates to Baryocentric Julian dates, and precessing coordinates between equinoxes and due to proper motion; this will automatically download and save the JPL ephemerides de430.bsp from JPL upon first import.


  • astrobase.fakelcs: modules and functions to conduct an end-to-end variable star recovery simulation.



This package requires the following other packages:

  • numpy
  • scipy
  • astropy
  • matplotlib
  • Pillow
  • jplephem
  • requests
  • tornado
  • pyeebls
  • tqdm
  • scikit-learn

For optional functionality, some additional packages from PyPI are required:

Installing with pip

If you’re using:

  • 64-bit Linux and Python 2.7, 3.4, 3.5, 3.6, 3.7
  • 64-bit Mac OSX 10.12+ with Python 2.7 or 3.6
  • 64-bit Windows with Python 2.7 and 3.6

You can simply install astrobase with:

(venv)$ pip install astrobase

Otherwise, you’ll need to make sure that a Fortran compiler and numpy are installed beforehand to compile the pyeebls package that astrobase depends on:

## you'll need a Fortran compiler.                              ##
## on Linux: dnf/yum/apt install gcc gfortran                   ##
## on OSX (using homebrew): brew install gcc && brew link gcc   ##

## make sure numpy is installed as well!                        ##
## this is required for the pyeebls module installation         ##

(venv)$ pip install numpy # in a virtualenv
# or use dnf/yum/apt install numpy to install systemwide

Once that’s done, install astrobase:

(venv)$ pip install astrobase

Other installation methods

To Install all the optional dependencies as well:

(venv)$ pip install astrobase[all]

To install the latest version (may be unstable at times):

$ git clone
$ cd astrobase
$ python install
$ # or use pip install . to install requirements automatically
$ # or use pip install -e . to install in develop mode along with requirements
$ # or use pip install -e .[all] to install in develop mode along with all requirements

Citing Astrobase

Released versions of Astrobase are archived at the Zenodo repository. Zenodo provides a DOI that can be cited for each specific version. The following bibtex entry for Astrobase v0.3.8 may be useful as a template. You can substitute in values of month, year, version, doi, and url for the version of astrobase you used for your publication:

      author       = {Waqas Bhatti and
                      Luke G. Bouma and
                      Joshua Wallace},
      title        = {\texttt{Astrobase}},
      month        = feb,
      year         = 2018,
      version      = {0.3.8},
      publisher    = {Zenodo},
      doi          = {10.5281/zenodo.1185231},
      url          = {}

Alternatively, the following bibtex entry can be used for all versions of Astrobase (the DOI will always resolve to the latest version):

      author       = {Waqas Bhatti and
                      Luke G. Bouma and
                      Joshua Wallace},
      title        = {\texttt{Astrobase}},
      month        = oct,
      year         = 2017,
      publisher    = {Zenodo},
      doi          = {10.5281/zenodo.1185231},
      url          = {}

Also see this AAS Journals note on citing repositories.

Period-finder algorithms

If you use any of the period-finder methods implemented by astrobase.periodbase, please also make sure to cite their respective papers as well.

  • the generalized Lomb-Scargle algorithm from Zechmeister & Kurster (2008)
  • the phase dispersion minimization algorithm from Stellingwerf (1978, 2011)
  • the AoV and AoV-multiharmonic algorithms from Schwarzenberg-Czerny (1989, 1996)
  • the BLS algorithm from Kovacs et al. (2002)
  • the ACF period-finding algorithm from McQuillan et al. (2013a, 2014)


Please see for the latest changelog for tagged versions.


Astrobase is provided under the MIT License. See the LICENSE file for the full text.

Indices and tables