# TDSS Target Selection

The Time-Domain Spectroscopic Survey (TDSS) is an SDSS-IV eBOSS subprogram to systematically obtain follow-up spectra of ~200,000 photometrically time-variable sources. Target selection for TDSS in SDSS-IV is detailed in Morganson et al. (2015) and MacLeod et al. (2018).

TDSS targets come from three types of observing programs:

1. Single-Epoch Spectroscopy (SES) program encompassing ~80% of all TDSS fibers and aimed at obtaining initial single-epoch discovery spectra of photometrically variable objects
2. Few-Epoch Spectroscopy (FES) programs consisting of various smaller projects involving multi-epoch spectroscopy of special select targets (particular subclasses of quasars or stars with anticipated spectral variability)
3. Repeat Quasar Spectroscopy (RQS) also involving multi-epoch spectroscopy (but in a less subclass-biased way than FES) of quasars across a broad range of redshift, luminosity, quasar subclass, etc.

## TDSS Single-Epoch Spectroscopy (SES) target selection

SES target selection, which comprises the bulk of TDSS TDSS targets, is based on a uniform algorithm. The SES algorithm is detailed in Morganson et al. (2015), and briefly summarized immediately below.

Targets for TDSS SES spectra in SDSS-IV are selected based on photometric gri light curves of point sources, constructed using a combination of SDSS Data Release 9 single-epoch imaging and Pan-STARRS1 (PS1) 3$\pi$ survey multi-epoch imaging.

Photometrically variable objects are selected based on their:

1. long-timescale (~10 years observed-frame) variability
2. shorter-timescale (~2 years observed-frame) variability
3. median PS1 magnitudes

Long-timescale variability is measured as the difference between the SDSS and median PS1 epoch magnitudes (where the SDSS magnitudes are color-corrected to match PS1 filters). Shorter-timescale variability is measured as the variance in the PS1 light curves.

The median PS1 magnitudes are included in the target selection procedure primarily to take into account the increase in variability that is required for fainter objects to enter the sample due to increased photometric uncertainties, and do not involve color information.

These three parameters for each object are input into a 3-dimensional kernel density estimator (KDE) trained on the SDSS Stripe 82 variable object and standard star catalogs. The KDE assigns a probability that each object is variable based on the efficiency E with which variable objects are selected in its region of parameter space using the training sets. Candidate variable objects with E above some threshold are then selected as “variable.” To ensure a uniform sky density of targets across the survey footprint of 10 deg$^{-2}$, this E threshold is calculated independently for different sky regions (each typically 4 deg$^2$) and is thus non-constant.

Many TDSS targets selected using the above procedure have previous SDSS spectra (~70,000 objects), or are also selected for spectroscopy in SDSS-IV by other eBOSS programs (e.g. the majority of TDSS quasars are also targeted by eBOSS). TDSS SES does not obtain new spectra of targets with previous spectra, while objects also targeted by eBOSS are `free’ and do not count towards the 10 deg$^{-2}$ sky density of TDSS fibers.

After targets are selected, an additional visual pre-screening is performed before spectra are obtained. This visual pre-screening of TDSS targets uses postage stamps of SDSS imaging to remove targets whose variability is suspect due to obvious photometry problems (e.g., lying on diffraction spikes of bright stars, lying within the isophotes of a spatially resolved galaxy, etc.). This pre-screening removes a non-negligible fraction (~1/3 in some cases) of objects to maximize the purity of the final TDSS sample.

TDSS fibers are also included as part of the SEQUELS survey in SDSS-III. Differences in the target selection between SDSS-IV and SDSS-III SEQUELS are small, and are detailed in Ruan et al. (2016).

## TDSS Few-Epoch Spectroscopy (FES) and Repeat Quasar Spectroscopy (RQS) target selection

TDSS FES and RQS targets encompass multi-epoch spectra of variable stars and quasars of a range of interesting subsets, with target selection for the various subsets detailed in MacLeod et al. (2018). But briefly:

Targets for FES programs are selected with a variety of methods to address specific science questions amongst both stars and quasars known or suspected to potentially show interesting spectroscopic variability. There are nine separate, small FES programs to study spectroscopic variability. These include, in approximate order of decreasing sample size: BAL quasars, the most photometrically variable (“hypervariable”) quasars, high-S/N normal broad-line quasars, quasars with double-peaked or very asymmetric BEL profiles, hypervariable stars (including the most highly variable classical pulsators), active ultracool (late-M and early-L) dwarf stars with Hα emission, dwarf carbon stars, white dwarf/M dwarf spectroscopic binaries with Hα emission, and binary supermassive black hole candidates from Mg ii broad line velocity shift analysis

For the RQS program, TDSS targets spectroscopically confirmed quasars (either point-source or with extended morphology) in the DR7-12 quasar catalogs, as well as newly discovered quasars in SDSS-IV. The subset of quasars with i<19.1 defines our highest priority RQS targets, accounting for the majority of targets (> 7 per sq. deg). Also included in the top RQS priority class are i<20.5 quasars with multiple existing spectra (except for the region in Stripe 82 where the density for such objects exceeds the TDSS fiber density allotment). Within Stripe 82, the SDSS-IV footprint, and a part of the North Galactic Cap, TDSS uses a variability selection to fill the remaining fibers, using SDSS, Stripe 82, and PS1 3π photometry. These lower-priority targets are defined by different cuts in the reduced χ2 for a model for which the quasar’s brightness level does not vary. For other areas, TDSS applies further magnitude cuts to limit the sample instead of a variability cut.