APOGEE Caveats

Note that the APOGEE data in DR15 are identical to those in DR14, but use the DR15 documentation pages for reference.

This page relays a list of known issues in the DR15 APOGEE data release. We have separated these into two categories: (i) problems and bugs and (ii) fundamental limitations from either data or methodology. The former are known issues in the catalogs themselves (many discovered after the release), whereas the latter are more fundamental to what we can provide.

For additional discussion of certain issues, see the pages on Using APOGEE spectra, Using ASPCAP stellar parameters, and Using APOGEE stellar abundances.

Problems and Bugs

VHELIO is in Barycentric Frame

In the allVisit and allStar files, the VHELIO entry is in the barycentric frame, not the heliocentric frame as the name suggests. The naming convention has been maintained from earlier releases for historical reasons. The data model has been corrected in DR15 to more clearly convey this information to the user.

Target bit not set

APOGEE2_TARGET2 bit is not set for the following bit flags:

  • 22 -- APOGEE2_1M_TARGET
  • 30 -- APOGEE2_OBJECT

This will be corrected in future releases.

logg Calibration for RC stars

Second red clump (2RC) stars were not included in the calibration of ASPCAP log g to APOKASC log g. As a result the calibrated log g values for the 2RC stars are incorrect.

Users are encouraged to recalibrate their data for all RC stars following the updated calibration relations described in the External Calibration section of the ASPCAP documentation.

Inconsistent IDs

For a few objects, alternate object/star names were used during the reduction. 2MASS identifications have been adopted in the final summary files and associated database tables, but if users want to find the individual star spectrum files (e.g., the apStar or aspcapStar files), they will need to know the star name used during the reduction, which is stored in the REDUCTION_ID tag. This is particularly true for the 1m+APOGEE observations.

Duplicate Targets

For certain fields which overlap one another, a few stars were inadvertently targeted in both fields. Since spectra are combined only within a field, these stars appear more than once in the combined spectra and summary files/tables. When multiple combined spectra of the same object exist, the lower S/N observations all have a bit set in the EXTRATARG flag that appears in the summary allStar file and in the CAS table.

The EXTRATARG bit should allow users to avoid the duplicate use of the same target in an analysis.

Multiple target classes

Some targets may have been selected independently for different programs within different visits to the same field. When we combine spectra, the target flag that we adopt for the combined spectrum is a bitwise OR of all of the target flags of the individual visit spectra. As a result, there are stellar spectra constructed from multiple visits for which a target flag bit may be set in the combined spectra, but not in all of the visit spectra that were used to construct it.

Y is Misidentified

Due to a typographical error, the measurements called Y in our data were made on a line from Yb. The error, however, is deeper than just a "typo" because the associated transition parameters were also from Y. Thus, ignore this element in scientific applications.

Invalid C and N abundances in dwarfs

For stars fit with the dwarf grids, the C and N abundances are not valid. The dwarf grids do not have separate dimensions for C and N, so the [M/H] dimension is used to determine C and N abundances. Because most of the information for C and N come from molecules, varying C and N with [M/H] is incorrect; whenever the C or N abundance is varied, the abundances of all other elements are also varied! Since many of the windows for C and N are the same, this also has the effect of making the C and N abundances almost the same. This was a problem as early as DR13 that was not recognized at the time.

Users should not use C and N abundances for stars fit with the dwarf grids (ASPCAP_CLASS = Fd, GKd, or Md)!

The Cannon

All of the caveats listed on this page apply to the Cannon results. Unfortunately, there is no way for the user to recalibrate the Cannon results.

Users should not use any results for stars with log g > 2.9 due to the problem with the ASPCAP surface gravity calibration.

Methodology and Data Limitations

Lingering effects of LSF

Although ASPCAP currently attempts to account for LSF variations by splitting the analysis into groups by mean fiber number, there are indications that the mean and scatter in abundances of stars varies at a low level with the mean fiber number (few hundredths of a dex).

Erroneous extinctions for bright stars

For very bright objects, the WISE photometry employed to generate some extinction estimates can have significant problems, which leads to estimated extinctions that may be appreciably in error. This is particularly true for the bright star sample observed with the NMSU 1-meter.

Users are cautioned against using the tabulated extinctions for very bright stars.

Persistent effects of persistence

While the improved treatment of persistence seems to result in significantly better performance in the derivation of stellar parameters and abundances for stars affected by persistence, there are still some stars for which persistence likely leads to issues.

Users of the APOGEE spectra should pay careful attention to the persistence flags and should consider the use of the inflated uncertainties for persistence-affected pixels. .

Atypical Abundance Ratios

Because ASPCAP works by varying element families together, inaccuracies can occur if aberrant/non-standard element abundance ratios are present in stars. For example, the ASPCAP parameter derivation is somewhat flawed in second generation globular cluster stars, where non-standard oxygen abundances lead to systematic offsets in effective temperature and gravity, which in turn result in offsets in other quantities.


For stars in which all of the visit spectra were recorded in regions affected by persistence, the uncertainties for the pixels/wavelengths affected by persistence are significantly inflated. This has the effect of down-weighting them relative to pixels/wavelengths that are not affected by persistence in the ASPCAP fits. Because the S/N that characterizes the combined spectra is the median S/N of all of the pixels, the standard S/N reported for these stars is reduced. To provide a better S/N estimate, we have also calculated an alternate S/N, SNREV, that is determined over a wavelength region in the middle chip that should not have many pixels that can be affected by persistence.

SNREV is the recommended quantity to use for S/N assessment.