Skip to content
Go to file

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time

ARIADNE (spectrAl eneRgy dIstribution bAyesian moDel averagiNg fittEr)

Characterize stellar atmospheres easily!

ARIADNE Is a code written in python 3.7 (python 3.8 changed the way multiprocessing works and that breaks ARIADNE) designed to fit broadband photometry to different stellar atmosphere models automatically using Nested Sampling algorithms.


To install ARIADNE you can clone this repository with

git clone

And then run

python install

Or try pip install astroARIADNE (Soon to be available!)

But for the code to work, first you must install the necessary dependencies.


Most can be easily installed with pip or conda but some might have special instructions (like PyMultinest!!)

ARIADNE has been tested on OS X up to Catalina and Linux. It does NOT run on Windows because healpy, a dependency of dustmaps isn't available for Windows (see

In order to plot the models, you have to download them first:

But note that plotting the SED model is optional. You can run the code withouth them!

Model Link
Phoenix v2
Phoenix v2 wavelength file
Castelli & Kurucz
Kurucz 1993

The wavelength file for the Phoenix model has to be placed in the root folder of the PHOENIXv2 models.

For the code to find these models, you have to place them somewhere in your computer as follows:

│   │
│   └───CIFIST2011
│	 │
│	 └───AGSS2009
│	 │
│	 └───AGSS2009
│	 │
│	 └───ckm05
│	 │
│	 └───ckm10
│	 │
│	 └───ckm15
│	 │
│	 └───ckm20
│	 │
│	 └───ckm25
│	 │
│	 └───ckp00
│	 │
│	 └───ckp02
│	 │
│	 └───ckp05
│	 │
│	 └───km01
│	 │
│	 └───km02
│	 │
│	 └───km03
│	 │
│	 └───km05
│	 │
│	 └───km10
│	 │
│	 └───km15
│	 │
│	 └───km20
│	 │
│	 └───km25
│	 │
│	 └───kp00
│	 │
│	 └───kp01
│	 │
│	 └───kp02
│	 │
│	 └───kp03
│	 │
│	 └───kp05
│	 │
│	 └───kp10
     └─── WAVE_PHOENIX-ACES-AGSS-COND-2011.fits


  • The Phoenix v2 models with alpha enhancements are unused
  • BT-models are BT-Settl, BT-Cond, and BT-NextGen

How to use?

Stellar information setup

To use ARIADNE start by setting up the stellar information, this is done by importing the Star module.

from import Star

After importing, a star has to be defined.

Stars are defined in ARIADNE by their RA and DEC in degrees, a name, and optionally the Gaia DR2 source id, for example:

ra = 75.795
dec = -30.399
starname = 'NGTS-6'
gaia_id = 4875693023844840448

s = Star(starname, ra, dec, g_id=gaia_id)

The starname is used purely for user identification later on, and the gaia_id is provided to make sure the automatic photometry retrieval collects the correct magnitudes, otherwise ARIADNE will try and get the gaia_id by itself using a cone search centered around the RA and DEC.

Executing the previous block will start the photometry and stellar parameter retrieval routine. ARIADNE will query Gaia DR2 for an estimate on the temperature, radius and the parallax, which can be used as priors for the fitting routine, and luminosity for completeness, as it's not used during the fit, and prints them along with its TIC, KIC IDs if any of those exist, its Gaia DR2 ID, and maximum line-of-sight extinction Av:

			Gaia DR2 ID : 4875693023844840448
			TIC : 1528696
			Effective temperature : 4975.000 +/- 104.390
			Stellar radius : 0.656 +/- 0.141
			Stellar Luminosity : 0.238 +/- 0.003
			Parallax : 3.297 +/- 0.036
			Maximum Av : 0.030

If you already know any of those values, you can override the search for them by providing them in the Star constructor with their respective uncerainties. Likewise if you already have the magnitudes and wish to override the on-line search, you can provide a dictionary where the keys are the filters and values are the mag, mag_err tuples.

If you want to check the retrieved magnitudes you can call the print_mags method from Star:


This will print the filters used, magnitudes and uncertainties. For NGTS-6 this would look like this:

		     Filter     	Magnitude	Uncertainty
		----------------	---------	-----------
		    2MASS_H     	 11.7670 	  0.0380
		    2MASS_J     	 12.2220 	  0.0330
		    2MASS_Ks    	 11.6500 	  0.0320
		GROUND_JOHNSON_V	 14.0870 	  0.0210
		GROUND_JOHNSON_B	 15.1710 	  0.0140
		  GaiaDR2v2_G   	 13.8175 	  0.0006
		  GaiaDR2v2_RP  	 13.1127 	  0.0015
		  GaiaDR2v2_BP  	 14.4012 	  0.0027
		     SDSS_g     	 14.6390 	  0.0580
		     SDSS_i     	 13.3780 	  0.0570
		     SDSS_r     	 13.7030 	  0.0320
		  WISE_RSR_W1   	 11.5550 	  0.0270
		  WISE_RSR_W2   	 11.6360 	  0.0270
		   GALEX_NUV    	 21.9520 	  0.4090
		      TESS      	 13.1686 	  0.0062

Note: ARIADNE automatically prints and saves the used magnitudes and filters to a file.

The way the photometry retrieval works is that Gaia DR2 crossmatch catalogs are queried for the Gaia ID, these crossmatch catalogs exist for ALL-WISE, APASS, Pan-STARRS1, SDSS, 2MASS and Tycho-2, so finding photometry relies on these crossmatches. In the case of NGTS-6, there are also Pan-STARRS1 photometry which ARIADNE couldn't find due to the Pan-STARRS1 source not being identified in the Gaia DR2 crossmatch, in this case if you wanted to add that photometry manually, you can do so by using the add_mag method from Star, for example, if you wanted to add the PS1_r mag to our Star object you would do:

s.add_mag(13.751, 0.032, 'PS1_r')

If for whatever reason ARIADNE found a bad photometry point and you needed to remove it, you can invoke the remove_mag method. For example you wanted to remove the TESS magnitude due to it being from a blended source, you can just run


A list of allowed filters can be found here

After the photometry + stellar parameter retrieval has finished, we can estimate the star's log g to use as prior later with the estimate_logg method:


This concludes the stellar setup and now we're ready to set up the parameters for the fitting routine.

Fitter setup

In this section we'll detail how to set up the fitter for the Bayesian Model Averaging (BMA) mode of ARIADNE. For single models the procedure is very similar.

First, import the fitter from ARIADNE

from astroARIADNE.fitter import Fitter

There are several configuration parameters we have to setup, the first one is the output folder where we want ARIADNE to output the fitting files and results, mext we have to select the fitting engine (for BMA we can only use dynesty for now), number of live points to use, evidence tolerance threshold, and the following only apply for dynesty: bounding method, sampling method, threads, dynamic nested sampler. After selecting all of those, we need to select the models we want to use and finally, we feed them all to the fitter:

out_folder = 'your folder here'

engine = 'dynesty'
nlive = 500
dlogz = 0.5
bound = 'multi'
sample = 'rwalk'
threads = 4
dynamic = True

setup = [engine, nlive, dlogz, bound, sample, threads, dynamic]

# Feel free to uncomment any unneeded/unwanted models
models = [

f = Fitter() = s
f.setup = setup
f.av_law = 'fitzpatrick
f.out_folder = out_folder
f.bma = True
f.models = models
f.n_samples = 100000

Note: While you can always select all 6 models, ARIADNE has an internal filter put in place in order to avoid having the user unintentionally bias the results. For stars with Teff > 4000 K BT-Settl, BT-NextGen and BT-Cond are identical and thus only BT-Settl is used, even if the three are selected. On the other hand, Kurucz and Castelli & Kurucz are known to work poorly on stars with Teff < 4000 K, thus they aren't used in that regime.

We allow the use of four different extinction laws:

  • fitzpatrick
  • cardelli
  • odonnell
  • calzetti

The next step is setting up the priors to use:

f.prior_setup = {
	'teff': ('rave'),
	'logg': ('default'),
	'z': ('default'),
	'dist': ('default'),
	'rad': ('default'),
	'Av': ('default')

A quick explanation on the priors:

The default prior for Teff is an empirical prior drawn from the RAVE survey temperatures distribution, the distance prior is drawn from the Bailer-Jones distance estimate from Gaia DR2, and the radius has a flat prior ranging from 0.5 to 20 R$_\odot$. The default prior for the metallicity z and log g are also their respective distributions from the RAVE survey, the default prior for Av is a flat prior that ranges from 0 to the maximum of line-of-sight as per the SFD map, finally the excess noise parameters all have gaussian priors centered around their respective uncertainties.

We offer customization on the priors as well, those are listed in the following table.

Prior Hyperparameters
Fixed value
Normal mean, std
TruncNorm mean, std, lower_lim, uppern_lim
Uniform ini, end
RAVE (Teff only) ---
Default ---

So if you knew (from a spectroscopic analysis, for example) that the effective temperature is 5600 +/- 100 and the metallicity is [Fe/H] = 0.09 +/- 0.05 and you wanted to use them as priors, and the star is nearby (< 70 pc), so you wanted to fix Av to 0, your prior dictionary should look like this:

f.prior_setup = {
	'teff': ('normal', 5600, 100),
	'logg': ('default'),
	'z': ('normal', 0.09, 0.05),
	'dist': ('default'),
	'rad': ('default'),
	'Av': ('fixed', 0)

After having set up everything we can finally initialize the fitter and start fitting


Now we wait for our results!


After the fitting has finished, we need to visualize our results. ARIADNE includes a plotter object to do just that! We first star by importing the plotter:

from astroARIADNE.plotter import SEDPlotter

The setup for the plotter is already made for you, but if you really want to change them, instructions on how to change it can be found here

Before we plot the SEDs we need to tell ARIADNE where to find our models. This step isn't necessary if you don't want or need SED plots and are happy with the HR diagram, histograms, cornerplot and RAW SED. This is done with an environmental variable called ARIADNE_MODELS, to set it up you just need to run export ARIADNE_MODELS='/path/to/Models_Dir/' in your terminal. You can also add that instruction to your .bash_profile or .bashrc and the run source ~/.bash_profile so you don't have to export everytime.

Now that ARIADNE knows where to find the models we only need to specify the results file location and the output folder for the plots!

in_file = out_folder + 'BMA_out.pkl'
plots_out_folder = 'your plots folder here'

Now we instantiate the plotter and call the desired plotting methods! We offer 5 different plots:

  • A RAW SED plot
  • A SED plot with the model and synthetic photometry
  • A corner plot
  • An HR diagram taken from MIST isochrones
  • Histograms showing the parameter distributions for each model.
artist = SEDPlotter(in_file, plots_out_folder)

The number given to plot_bma_HR is the number of extra tracks you want to plot, drawn randomly from the posterior distribution.

If you're iterating through lots of stars you can call the SEDPlotter clean method to clear opened figures with artist.clean()

If you don't have the models in your computer, then the plot_SED method will fail, as it needs the complete model grid.

An example usage file is provided in the repository called for the BMA approach and for single model fitting.


After ARIADNE has finished running, it will output a series of files and plots showing the results of the fit and other information.

The most important file is the best_fit.dat which contains the best fiting parameters with the 1 sigma error bars and the 3 sigma confidence interval. Then there are pickle files for each of the used models plus a last one for the BMA, these contain raw information about the results. There is a prior.dat file that shows the priors used and a mags.dat file with the used magnitudes and filters.

Another important output are the plots. Inside the plots folder you can find CORNER.png/pdf with the cornerplot (the plot showing the distribution of the parameters agains eachother), HR_diagram.png/pdf only for the BMA, with the HR diagram showing the position of the star, SED_no_model.png/pdf with the RAW SED showing each photometry point color coded to their respective filter, and SED.png/pdf with the SED with the catalog photometry plus synthetic photometry. If BMA was done, there's also a histograms folder inside the plot folder with various histograms of the fitted parameters and their distribution per model, highlighting the benefits of BMA.

Examples of those figures:

SED plot HR Diagram Corner plot Histogram example


Easy stellar SED fitting!





No releases published


No packages published


You can’t perform that action at this time.