leveraging LSST Tony Tyson Director, LSST Project University of California, Davis 1 7.5 arcminutes DSS: digitized photographic plates 2 Sloan Digital Sky Survey 3 LSST -- almost 2800 galaxies i<25 mag ×200 for one DESpec FOV 4 LSST Observing Cadence Pairs of 15 second exposures (to 24.5 mag) per visit to a given position in the sky. Visit this position again within the hour with another pair of exposures. Number of 9.6 sq.deg FOV visits per night: 900 Deep-Drilling: 1 hour per night on a selected field. Continuous 15 sec exposures Detection of transients announced within 60 seconds. Expect ~1 million per night 5 Number of visits per field in Deep Wide Survey 6 Two planned LSST surveys MAIN SURVEY Deep Wide Survey: 20,000 square degrees to a uniform depth of u: 26.7 g: 27.4 r: 27.7 i: 26.9 z: 26.1 y: 24.9 DEEP DRILLING 10% of time: ~30 selected fields. 300 square degrees Continuous 15 sec exposures. 1hour/night 7 LSST Wide-Fast-Deep survey 4 billion galaxies with photometric redshifts 20 trillion photometric measurements of 20 billion objects 70PB database Immediate transient alerts LSST Science Book v2.0 written by LSST Collaboration 245 authors 598 pages Living document (on lsst.org) http://www.lsst.org/lsst/scibook 9 LSST Science Charts New Territory Probing Dark Matter And Dark Energy Mapping the Milky Way Finding Near Earth Asteroids 10 Three classes of useful LSST spectroscopy 1. Calibration samples for quantities that can be derived from photometric data: photometric redshifts for galaxies, photometric metallicity for stars 2. Supplemental data that cannot be obtained from LSST data: radial velocity, emission and absorption line strengths 3. Identification spectra for transient, weird and unusual objects (SNe, GRB followup, high-z quasars, brown dwarfs) These differ by the needed sample size, sample depth, required spectral resolution, and the time delay relative to imaging data. 11 The vast majority of LSST science uses LSST multi-band time domain photometry data alone Some exceptions: 1. Photo-z needs spec for r<24 mag over 10 sq.deg -> 10meter 2. TRANSIENTS • Rare bright needs 2-4meter • Faint (22-24mag) needs 10-30meter 3. Strong lensing: magnified source spectroscopy 4. SNe z<1.2 5. Stellar needs 10meter mostly hi-res 12 190 sq deg DESpec on Blanco* to 24.5 13 Alert Rate In ten minutes time the LSST transient pipeline is likely to issue ~10,000 alerts at 5σ. While most of these will be moving objects, perhaps several thousand will be flaring objects or bursts. Possibly new kinds of objects! Clearly any feasible spectroscopic followup at 23 mag will lag behind ~1 hour per hour. What is needed then is highly trusted event classification. FAST 14 DESpec coverage of LSST surveys ¼ of main 20,000 sq.deg to 22.5 mag at S/N=5 0.7% of gold sample of i<25 mag galaxies useful for correlation calibration of photo-z deep spectroscopy of deep drilling fields 20 fields, 20 hours each, 50 nights ~24 mag @ S/N=10 per 10Å bin transient host spectra co-observing spectroscopy of deep drilling fields 1 hour / night ~22 mag @ S/N=5 per 1Å bin transient spectra 15 CLASSIFICATION DATA PRODUCTS 16 Science at the Limit Much of the breakthrough science using surveys (imaging or spectroscopy) occurs at the limits of the surveys Noise, Sample incompleteness Subtle systematic errors Statistical studies must be corrected for these errors 17 large spectroscopic samples are useful Example: even an incomplete spectroscopic sample can help photo-z in two ways 1. angular cross correlation with faint photometric sample will calibrate photo-z statistically (Newman) 2. large spectroscopic samples can improve knowledge of evolution of galaxy SEDs. 18 There are currently 20 telescopes larger than 3m with spectrographs that can reach the LSST survey area. Even in the ELT era, wide field multiobject spectroscopy on 4-10m class telescopes will be useful Statistics matters. Calibration of the 70PB LSST database, and massively parallel follow-up of a million transients will be complementary to selected faint object spectroscopy 19 LSST science deliverables do not require followup spectroscopy But we can and should pursue a range of followup programs, from co-observing highly parallel spectroscopy, to individual object followup. 21 Assuring accurate classification Characterize the known clustering) Assign the new (classification) Discover the unknown (outlier detection) Tom Vestrand Benefits of very large data sets: • best statistical analysis of “typical” events • automated search for “rare” events 22 The dimension reduction problem: Finding correlations and “fundamental planes” of parameters • The Curse of High Dimensionality ! – Are there combinations (linear or non-linear functions) of observational parameters that correlate strongly with one another? – Are there basis sets that represent the full set of properties?