Astrophysical Modeling of Time-Domain Surveys

The goal of this work is to develop and apply algorithmic approaches for astrophysical modeling of time- domain surveys. Such approaches are necessary to exploit ongoing and future all-sky time-domain surveys. I focus on quantifying and characterizing source variability based on sparsely and irregularly sampled, non-simultaneous multi-band light curves, with an application to the Pan-STARRS1 (PS1) 3 pi survey: variability amplitudes and timescales are estimated via light curve structure functions. Using PS1 3 pi data on the SDSS "Stripe 82" area whose classification is available, a supervised machine-learning classifier is trained to identify QSOs and RR Lyrae based on their variability and mean colors. This leads to quite complete and pure variability-selected samples of QSO and RR Lyrae (away from the Galactic disk), that are unmatched in their combination of area, depth and fidelity. The sample entails: 4.8 x 10^4 likely RR Lyrae in the Galactic halo, and 3.7 x 10^6 likely QSO. The resulting map of RR Lyrae candidates across 3/4 of the sky reveals targets to 130 kpc, with distances precise to 3%. In particular, the sample leads to an unprecedented map of distance and width of Sagittarius stream, as traced by RR Lyrae. Furthermore, the role of PS1 3 pi as pilot survey for the upcoming LSST survey is discussed.