AstroNote 2019-15

Primary tabs

2019-05-20 13:31:06
Type: Announcement-Tool/Utility
Revising the treatment of object coordinates (astrometric accuracies) on the TNS *** CALL FOR FEEDBACK ***
Authors: Ofer Yaron, Avishay Gal-Yam, Eran Ofek, Avner Sass, Nikola Knezevic (Weizmann)

We are now in a process of implementing an improved treatment of object coordinates on the TNS, based on measured astrometric accuracies of the major reporting surveys, as detailed below.

As described in the bottom section relating to the estimation of the surveys' accuracies, we plan to apply this revision on June 10 (postponed to the 23rd), 2019, and we encourage the community's feedback and inputs by that date.

The new treatment will have an effect on both of the following aspects:
1. For every object that has multiple public discovery (AT) reports from different surveys, the principal object coordinates will be set according to the report of the group for which the highest astrometric accuracy was measured (relating to the latest report of that group, if the same group submitted several reports for the same object).
So, e.g., if an additional AT report is received for a certain object after discovery, and this additional report is from a group that has a better astrometric accuracy, the object coordinates will be updated according to the new report. All reported coordinates are clearly kept and remain unchanged within the list of AT reports. The discovery credit (certificate) is in any case not influenced by this change.
2. The measured astrometric accuracies also come into effect in the automated decision whether to associate a new discovery report with an existing object or if to create a new object. Until now a constant 5 arcsec separation threshold (as well as a temporal threshold of 5 years) was in effect; so, a new report with coordinates that were within 5" (and within 5 yrs) of an existing object would be automatically associated with that object. 

In the proposed revision, a new object will be created whenever the separation of a new report from all known objects is larger than the square root of the sum of the 95%-level astrometric uncertainties squared of the new report and existing objects, each assigned based on the astrometric accuracy of the reporting survey. In effect this is roughly equivalent to the 99.7% confidence level.
So, e.g., if an existing object was discovered by a group having an accuracy of 0.4 arcsec, if that same group now submits a new report with coordinates separated by >sqrt(2)*0.4", a NEW object will be created based on the group’s defined accuracy. If the group submitting a new report has an accuracy of say 2", a new object will be created only if the separation between the new and existing object surpasses sqrt(2^2+0.4^2), and so on.

Estimation of the accuracies for the most frequently reporting surveys.
As our fiducial definition of the survey accuracies, open to comments and suggestions by the community, we used the transient reports by Gaia as "ground truth". We then made a comparison of the separations relative to the Gaia reported coordinates for all objects with multiple reports by Gaia and the other surveys. We inspected the distribution of offsets and determined the 95% percentile of these distributions for all surveys with >20 coincident reports with Gaia, and adopted these values as estimators of survey astrometric accuracy. See the attached histograms for plots of the top contributing surveys (panels on the right are shown with a fixed x-axis span, 0.0-2.0, for easier comparison of the separation distributions).

Besides the comparisons against Gaia, an additional complexity comes into play and that is the internal scatter of an object's coordinates within a given survey. Some surveys have submitted multiple AT reports for some objects (like ZTF, ATLAS, Pan-Starrs), and analysis of the distributions of the internal scatter revealed that this scatter can be slightly larger than the derived accuracy (based on the comparison with Gaia). Reaching a solid conclusion is difficult (in the process we've encountered cases where objects having multiple reports, from either several surveys or even a single one, should actually be separated to different events; this will be taken care of), therefore, in order to be on the safe side, we propose to add a ~20% safety margin to the dervied accuracies, and relate to those final values as representing the surveys' accuracies.

The table below provides the derived and proposed survey accuracy values (in arcsec). For Gaia itself we took a safety margin slightly above the expected accuracy.


Survey 0.95 quantile Applied Accuracy (20% margin rounded)
GaiaAlerts ----- 0.20
Pan-Starrs1 0.38 0.46
OGLE 0.43 0.52
ZTF 0.44 0.53
ATLAS 0.83 1.00
POSS 1.34 1.60
PTSS 2.82 3.40
ASAS-SN 3.40 4.00


All other surveys currently remain with the default value of 5 arcsec (in accordance with the existing threshold value), since we do not have in the system enough reports coincident with Gaia reports from which to draw a meaningful estimate.

This situation is expected to continue and improve as more reports from Gaia and other surveys accumulate, however, we do encourage comments from the community, especially from the existing active surveys on the TNS, to inform us if a specific value (other than the default 5") of the astrometric accuracy should be defined for a survey, supported by convincing evidence.

As mentioned, this revision is due to be deployed on June 10 (postponed to the 23rd), 2019; so please provide your comments and inputs by then.

The process will also apply backwards to all existing objects on the TNS for which multiple discovery reports exist; so if the current coordinates of an object differ from the coordinates reported for the same object by the group with the highest defined accuracy, the principal coordinates will be updated. Clearly, a log of the changes will be provided after processing.



Here are the values for both the 0.5 (median) and 0.95 quantiles, sorted by the 0.95 quantile, as were obtained for the separation distributions on 2019-05-20. N denotes the number of coincident reports (and as mentioned, we relate to only the surveys having >20 coincident reports with Gaia).

Gaia vs. N 0.5 0.95
Pan-STARRS1 668 0.12 0.38
OGLE 22 0.21 0.43
ZTF 991 0.12 0.44
ATLAS 1067 0.26 0.83
POSS 30 0.55 1.34
PTSS 25 0.62 2.82
ASAS-SN 272 1.17 3.40