MaNGA Caveats

The Secondary sample was designed to have a higher density of targets than is required to give the desired 2:1 ratio of Primary+ to Secondary galaxies. Therefore, before allocating IFUs we randomly sample the Secondary sample to the desired target density. Due to a very recently discovered bug in the target selection code this random sampling is not truly random and in fact samples in such a way as to make the number density distribution flat as a function of stellar mass. This is a small change, since the density distribution was already quite close to being flat with stellar mass. However, it does mean the observed Secondary sample is no longer selected purely by i-band absolute magnitude and redshift, but also has a weak dependence on stellar mass. It also has consequences for calculating the appropriate weights to apply to any sample containing the Secondary sample (see the weights FAQ).

For more information, see Wake et al. 2017 (AJ, 154, 86).

While MaNGA does not usually reobserve the same targets, the DR16 sample (identical to the DR15 sample) contains a number of galaxies that have been observed more than once, either deliberately for data quality assessment purposes or due to early issues with targeting and/or plate drilling. In almost all cases the duplicate observations have the same mangaid but different plate-ifu, and thus are easy to differentiate. Five galaxies, however, have been observed more than once but with different mangaids. The following table lists the five galaxies with their two assigned mangaids:

1-322159	12-84620
1-199287	12-49536
1-322172	12-193534
1-458396	12-129618
1-458515	12-129646

So, what happened here? Galaxies with mangaid 12-XXX correspond to an early sample selection done using the NSA catalogue v1b_0_0_v2 and were originally intended only for a commissioning run in March 2014. Given that the data obtained for those 12- galaxies is of comparable quality to the main sample MaNGA galaxies, DR15 and DR16 include those galaxies but not as part of the main samples. Some of those galaxies were reobserved later as part of the final sample selection and were assigned new 1- mangaids (1- indicates the v1_0_1 version of the NSA catalogue).

Our advice is that you should not use 12- galaxies for your statistical analysis, since they are not part of MaNGA's final sample selection. If you are not concerned about sampling issues and want to use them, please be aware of this caveat.

Thanks to Cheng Du and collaborators for noticing and reporting this issue.

Although the vast majority of cosmic rays and other transient features are detected by the DRP and flagged (either for removal or for masking), lower-intensity glitches are sometimes missed and can make it into the final datacubes where they show up as unmasked hot pixels. Further improvements to the DRP to handle these spurious cosmic rays is still under development. Currently, some of these artifacts (e.g., satellite trails) are being addressed on a case-by-case basis; however, individual hot glitches are not addressed. It is particularly important to be wary of this when searching for isolated emission features in the data cubes.

After the DR15 reductions were complete it was realized that some fiber bundles have reference metrology (i.e., lab-measured fiber locations within the IFU bundle) that are systematically incorrect in scale factor by 2%. This means that the astrometric solution of the corresponding data cubes will similarly be off by 2% relative to the center of the IFU bundle. This effect is generally too small to detect for individual galaxies since it corresponds to just 0.5 arcsec error (i.e., 1/5 of the PSF FWHM) in relative position at the edge of the largest bundles compared to the center of the bundle. However, science cases that depend critically on astrometric accuracy at the edges of the largest bundles should be aware of this issue. This problem will be fixed in the next data release.

If you find a data cube in the DRP data products that does not have the associated DAP output files, there are two reasons why this may have happened.

• In the automated run of the DAP, any observation staged to be analyzed must satisfy the following criteria:
  • MANGAID is not 'NULL',
  • MANGA_TARGET1 > 0 or MANGA_TARGET3 > 0,
  • and cz > -500 .

  These items basically mean that the target must be part of a main-survey or ancillary target and have a valid initial estimate of its cz redshift; this is allowed to be negative for some very local targets.

• The DAP data could be missing because the code faulted on a specific data cube. For DR16 (identical to DR15), 13 observations faulted, which was not considered critical to this release. Solving these issues is part of ongoing improvement to the DAP code. The affected observations are: 7443-3703, 8140-6101, 8146-3702, 8158-3703, 8309-3703, 8312-6101, 8481-6103, 8549-12703, 8993-1901, 9025-12702, 9507-12702, 9677-12703, 9888-9102. Note: Observations 8481-6103, 8549-12703, 8993-1901, and 9507-12702 faulted but only after creating the VOR10 MAPS and DAP model LOGCUBE files. These files have been distributed with DR15 and DR16 and are available in Marvin, but should be used with caution.

Errors are "formal" errors determined either by the covariance (inverse Hessian) matrix provided by the optimization algorithm (stellar and emission-line kinematics) or a direct propagation of the error based on the inverse variances provided by the DRP (as for the emission-line moments and spectral indices). Idealized experiments and analysis of repeat observations have shown that the formal errors for the stellar kinematics are within a factor of two of the statistical error; see the DAP technical papers listed here.

There are still severe deficiencies in the flagging, in general. Measurements can exhibit pixel-to-pixel variations that are inconsistent with random error (because of the strong covariance between neighboring spaxels) or fiber-level deviations that are inconsistent with physical intuition. These issues tend to occur at low S/N, however, they might not all be caught by a simple S/N cut. Please consider the limited robustness of the flagging for your science goals. These limitations apply to essentially all derived products, and will be improved for future releases.

As mentioned above, the DRP will inevitably miss some cosmic rays. Occasionally cosmic rays will land on or near emission lines, which will play havoc with the emission-line fitter. Beyond possibly affecting the flux of the nearest line, it can also pull off all of the lines because the lines are forced to have the same velocity. An example of this is 8134-9102 where a cosmic ray lands on [SII]6718 for e.g., spaxel (x,y) = (25,35). The cosmic ray is slightly blueward of the [SII] line center and leads to all the remaining lines being too far to the blue with incorrect fluxes and dispersions. We are testing ways of removing the influence of these artifacts.

Due to a logic error in the code, the velocity dispersion measurements for the [OII] lines are almost entirely masked and the inverse variance is provided as 0. This will be corrected in future releases.

• Errors in spectral indices are over-estimated, as demonstrated in comparing DAP measurements to the Firefly VAC, Monte-Carlo-generated errors.
• Spectral indices are all measured at input (NSA) redshift, not fitted one!
• Spectral-index velocity-dispersion corrections are done adopting the NSA redshift for all spectra. This means that there will be a velocity dependence of the correction that can be more significant that the correction for the velocity dispersion itself. The approach to the velocity dispersion corrections will be improved in future releases.

• In the MAPS files, the HDUCLAS1 = 'CUBE' even for single-channel extensions.
• Data cubes with no good stellar-continuum models are flagged as CRITICAL, but not DAPCRIT.
• The BINID in model LOGCUBE files does not have the emission-line moment and spectral-index BINIDs populated. Use the BINID maps from the MAPS files.