Foreground: Reddening (E(B-V)) Map

The data made available through this page have been superseded by the Green et al. 2018 reddening map, which is available here

Reddening (E(B-V)) Map (Green et al. 2015)

Green et al. (2015) derived a three-dimensional map of dust reddening to a distance of several kpc from Pan-STARRS1 and 2MASS stellar photometry. The map covers almost the entire sky north of declination -30 degrees, mostly at a resolution of 3.4 to 13.7 arcminutes. It is based on Pan-STARRS1 photometry for 800 million stars and matching 2MASS photometry for about 200 million of these stars. The three-dimensional data can be queried or downloaded from the 3D dust mapping site of Green, Schlafly, and Finkbeiner. This site provides the following notes on the three-dimensional data.

• Units

  We use the same definition of E(B-V) as the Schlegel, Finkbeiner & Davis (1998) dust map. Although this was originally supposed to be the excess B-V in the Landolt filter system, Schlafly & Finkbeiner (2011) found that it differs somewhat from the true stellar B-V excess. Therefore, in order to convert our values of E(B-V) to extinction in various filter systems, consult Table 6 of Schlafly & Finkbeiner (2011) (use the values in the RV = 3.1 column).

• Samples

  For each sightline, we provide multiple estimates of the distance vs. reddening relationship. Alongside the maximum-probability density estimate (essentially, the best-fit) distance-reddening curve, we also provide samples of the distance-reddening curve, which are representative of the formal uncertainty in the relation. Most statistics you may wish to derive, like the median reddening to a given distance, are best determined by using the representative samples, rather than the best-fit relation.

• Quality Assurance

  We include a number of pieces of information on the reliability of each pixel. A convergence flag marks whether our fit to the line-of-sight reddening curve converged. This is a formal indicator, meaning that we correctly sampled the spread of possible distance-reddening relations, given our model assumptions. It does not, however, indicate that our model assumptions were correct for that pixel. This convergence flag is based on the Gelman-Rubin diagnostic, a method of flagging Markov Chain Monte Carlo non-convergence.

  Additionally, minimum and maximum reliable distances are provided for each pixel, based on the distribution of stars along the sightline. Because we determine the presence of dust by observing differential reddening of foreground and background stars, we cannot trace dust beyond the farthest stars in a given pixel. Our estimates of dust reddening closer than the nearest observed stars in a pixel are similarly uncertain. We therefore urge caution in using our maps outside of the distance range indicated for each pixel.

LAMBDA provides a two-dimensional map of the cumulative reddening to the farthest distance bin for each pixel. This was calculated as the median over the Markov chain samples in the 3D dataset for the farthest distance bin, following the method given here. We also provide a 1-sigma uncertainty for the cumulative reddening. Following a recommendation from Gregory Green, this was calculated as half of the difference between the 84th percentile cumulative reddening sample and 16th percentile sample, with a floor of 0.03.

The two-dimensional reddening map is available as a binary FITS table, ordered according to the HEALPix scheme. The original three-dimensional dataset contains data for pixels of different sizes. The two-dimensional data have been upsampled to the minimum pixel size (HEALPix res=11, or Nside=2048). The data are set to NaN for unobserved pixels. Each record in the table has the following fields:

• EBV
  E(B-V) on the scale of Schlegel, Finkbeiner, and Davis 1998


• EBV_ERR
  E(B-V) uncertainty estimate


• DM_RELIABLE_MAX
  maximum reliable distance modulus


• CONVERGED
  flag indicating whether the fit to the line of sight reddening curve converged


• GR_DIAGNOSTIC
  the Gelman-Rubin diagnostic for the farthest distance bin, < 1.2 is good


• NSIDE
  specifies the original pixel size (HEALPix nside) in the three-dimensional dataset


• HEALPIX_INDEX
  the original nested pixel index in the three-dimensional dataset

For the inner Galactic plane, the 3D extinction map of Marshall et al. 2006 may also be of interest. This map penetrates to larger distance, but with poorer angular resolution.

More details can be found in:

A Three-dimensional Map of Milky Way Dust", Green, Schlafly, Finkbeiner et al. 2015, ApJ, 810, 25. ADS

3D Dust Mapping site

Dataverse site

Back To Foreground Derivative Diffuse Components Page