1. 1 FINAL DESIGN REVIEW | TUCSON, AZ | OCTOBER 21-25, 2013 Name of Meeting Location Date - Change in Slide Master The LSST Opserations Simulator A. Saha LSST OpSim lead October 21-25, 2013 FINAL DESIGN REVIEW October 21 - 25, 2013

2. 2 FINAL DESIGN REVIEW | TUCSON, AZ | OCTOBER 21-25, 2013 The Need for Scheduling Tools − To schedule observations with LSST in real time according to: Science Requirements and survey priorities Target position and availability The capabilities and constraints of the telescope Environmental parameters: sky conditions, sky brightness, etc. Scheduled and Unscheduled downtime Adapt to changes in goals and strategy over survey duration − A pair of tools: A “scheduler” that queues observations to the working telescope -Uses real time environment data A simulator that is logically identical to the scheduler, except that: -Uses simulations of environmental stimuli -Can simulate alternate scenarios − No drop shadows − Only use the font Calibri − Please do not alter template

3. 3 FINAL DESIGN REVIEW | TUCSON, AZ | OCTOBER 21-25, 2013 A Tool for Advanced Planning − Can the science goals in the SRD be met? An early achievement was to ascertain the minimal needed FOV area Determined constraints for readout and slew times − How can the survey be designed for the most compelling mix of science? What kinds of use cases are prohibitive? − How can performance margin be increased? − Learn how to construct scheduling tools for run-time operation − Communicate with science community re: survey cadence and coverage priorities

4. 4 FINAL DESIGN REVIEW | TUCSON, AZ | OCTOBER 21-25, 2013 Simulator Overview Parametrized Telescope/Instrument Model Site specific Observation Environment modelling: site, weather, sky-brightness Observational desiderata coded as “proposals” – generate time dependent “demand” Optimization: Maximize science programs simultaneously Output Observation History to Survey Database Supplementary analysis and reporting tools and metrics

5. 5 FINAL DESIGN REVIEW | TUCSON, AZ | OCTOBER 21-25, 2013 Salient Features − Sophisticated Telescope Model All movements tracked: mount, dome, optics, rotator, cable wraps, filter change 50 parameters configure speeds, accelerations, delays and limits Configurable table to determine sequence of movements for any given slew Simulation output can guide telescope engineering − Proposals Implement science programs Proposals run simultaneously Each proposal calculates demand dynamically Demand from all proposals aggregated - tuneable priorities Configurable parameters for defining target lists, filter and timing requirements, and other specifications

6. 6 FINAL DESIGN REVIEW | TUCSON, AZ | OCTOBER 21-25, 2013 Science Cases and Proposal Variety Can run any or all of the following types: –Deep Cosmology coverage total visits per field/filter combination – Transient proposals Simple to complex cadence with simple filter requirements –NEA proposal –Transients without color requirements – Transient SubSequence proposals Complex, independent, cadences for multiple filters Multi-color SN survey, multi-color KBO survey Universal Cadence – special case: Deep cosmology visits collected in 30 min pairs Designed to deliver several science goals simultaneously

7. 7 FINAL DESIGN REVIEW | TUCSON, AZ | OCTOBER 21-25, 2013 Simulation Runs − Well over 1000 simulations have been run over the years Simulator capabilities have grown over time Runs with different combinations of proposal types and parameter sets Informed engineering decisions: -Early: Telescope FOV, read-out time. Recent: Cable-wrap orientation − Primary survey: Universal Cadence for Wide Fast Deep survey for Cosmology and Milky Way studies, with ~30 min. pairs for Solar System objects. 18,000+ square degrees − Smaller area surveys devoted to: Deep supernovae and KBO – time sequences for SNe North ecliptic plane for NEOs Milky Way plane South celestial cap, including Magellanic Clouds

8. 8 FINAL DESIGN REVIEW | TUCSON, AZ | OCTOBER 21-25, 2013 A Fiducial Run (OpSim 3.61) − 10 year simulation: “existence proof” for an LSST survey − Observing start/stop at 12 degree twilight − CTIO 4m weather log as weather model − Scheduled downtime for maintenance − u filter in camera ~ 6 days per lunation − Five science proposals: WideFastDeep – Universal Cadence Galactic plane: collect 30 visits in each passband North ecliptic – Universal Cadence South Pole: collect 30 visits in each filter 6 fields of “deep drilling” for supernovae -100 day sequence – visit every 5 days in grizy

9. 9 FINAL DESIGN REVIEW | TUCSON, AZ | OCTOBER 21-25, 2013 OpSIm 3.61 visit distrbution on sky (SSTAR) The number of visits acquired for each field is plotted in Aitoff projection for each filter. All visits acquired by all observing modes are included in this plot.

10. 10 FINAL DESIGN REVIEW | TUCSON, AZ | OCTOBER 21-25, 2013 Single Visit limiting magnitudes (5-sigma) −The side box contains the values of the 25th, 50th (median), and 75th percentiles for each curve. The Simulator has limits for sky brightness and seeing conditions for each filter in each observing cadence. These limits result in the relatively tight distributions of limiting magnitudes for each filter

11. 11 FINAL DESIGN REVIEW | TUCSON, AZ | OCTOBER 21-25, 2013 Coadded depth for the WideFastDeep fields The distribution of fields with coadded depth in each filter. Only visits acquired by observing modes designed to meet the WFD number of visits are included. The inset box contains the values of the 25th, 50th (median), and 75th percentiles for each curve.

12. 12 FINAL DESIGN REVIEW | TUCSON, AZ | OCTOBER 21-25, 2013 Full survey coadded depth −The WFD design spec is used as a fiducial and the difference between it and the coadded depth for each field is plotted in Aitoff projection for each filter. Fields with positive values have a coadded depth deeper than the WFD zenith value. Visits acquired by all observing modes are included in this plot and are not limited only to observing designed to meet the WFD number of visits.

13. 13 FINAL DESIGN REVIEW | TUCSON, AZ | OCTOBER 21-25, 2013 Opsim3.61 is a stretch simulation −3.61 has 20000 square degrees for the Wide-Fast-Deep Survey (WFD) −3.61 has 1030 visits per WFD field over ten years −SRD design goals: 18000 square degrees for WFD 824 visits per WFD field −Simulator has investigated design vs stretch −OpSim 3.61 and a small number of variant project approved simulations are available to the whole LSST collaboration through a web-page: http://www.lsstcorp.org/opsim/home

14. 14 FINAL DESIGN REVIEW | TUCSON, AZ | OCTOBER 21-25, 2013 Necessary development with Version-3 −Look ahead capability (key for schedule optimizing algorithms) −Alternate optimization methods: beyond the “greedy” algorithm −Ability to “hot swap” proposals −Dithering capability −Simulate “failed observations” −Accommodate scripted cadences −Improve scattered light model −Develop operations scheduler requirements −Develop timeline for scheduler construction −Conduct wide range of cadence studies −Modularize code anticipating transformation to a scheduler that links to the OCS. −Make code exportable and supportable for experimentation by community

15. 15 FINAL DESIGN REVIEW | TUCSON, AZ | OCTOBER 21-25, 2013 Evaluation Tools: Beyond SSTAR − A large set of metrics have been developed to score the efficacy of simulations for a variety of science cases. Purpose: study of multiple simulations in a way that renders the complexity of the schedules comprehensible in a comparative way − Additional metrics have been sought from project and science collaborations − We have developed a general concept for a post-processing Framework, which would accept a list of merit functions, a list of simulations, evaluate those merits for those runs and produce a report for simulation comparison, or as material for further processing.

16. 16 FINAL DESIGN REVIEW | TUCSON, AZ | OCTOBER 21-25, 2013 Work schedule − Complete all development work by 2016 − Continue running simulations for engineering needs, as well as survey design − Complete design of scheduler-simulator pair by 2016 − Build final scheduler-simulator and deliver in 2019.

17. 17 FINAL DESIGN REVIEW | TUCSON, AZ | OCTOBER 21-25, 2013

18. 18 FINAL DESIGN REVIEW | TUCSON, AZ | OCTOBER 21-25, 2013 − Backup Slides

19. 19 FINAL DESIGN REVIEW | TUCSON, AZ | OCTOBER 21-25, 2013 Opsim 3.61 summary SSTAR uses LSST ETC to estimate depths from number of visits. 2,651,588 total visits, 20,000 square degrees: 75% in Universal (DeepWideFast) – u 24.2, g 24.7, r 24.5, i 23.9, z 23.1, y 21.9 – 656,687 pairs of griz with 15-60 minute separation – ~ 6 pairs per field per lunation 4,000 square degrees of northern ecliptic (12%) – 41,774 pairs of griz with 15-60 minute separation – ~ 2 pair per field per lunation 1,900 square degrees of Galactic Bulge/Plane (7%) – 30 visits in ugrizy each 1,300 square degrees of south celestial pole (6%) – 30 visits in ugrizy each 23 perfect deep 100 day supernova sequences, 170 incomplete Excellent period recovery for periodic variables

20. 20 FINAL DESIGN REVIEW | TUCSON, AZ | OCTOBER 21-25, 2013 Period%RecoveredPeriod Tolerance 0.1 day100%< 1x10 -5 1.0 day 50% 2x10 -5 10 day100% 2x10 -4 100 day100% 2x10 -3 1000 day70% 2x10 -2 WFD cadence recovers periods well A light curve is constructed from an average Delta Scuti light curve‘s Fourier coefficients with an arbitrary mean and amplitude. The measurement errors are drawn from a Gaussian distribution with a mean of amplitude/Signal to Noise of 7 and a phase from a flat distribution of phases. The curve is sampled using r-band observations from fields in the wide, fast, deep cadence in the baseline survey, and analyzed with the cleanest algorithm. Periods range from 0.1 to 1000 days. The period is considered recovered if found within the listed tolerance. The cadence does well except for a 1 day alias.

21. 21 FINAL DESIGN REVIEW | TUCSON, AZ | OCTOBER 21-25, 2013 Design vs Stretch goals for visits and area −Design visits and design area 2.93: 18000 square degrees, 824 visits −Design visits and stretch area 1.97: 20000 square degrees, 824 visits −Stretch visits and design area 3.69: 18000 square degrees, 1030 visits −Stretch visits and stretch area 3.61: 20000 square degrees, 1030 visits Current ‘fiducial’ runs are design/design

22. 22 FINAL DESIGN REVIEW | TUCSON, AZ | OCTOBER 21-25, 2013 The need for version 3 − Going beyond the greedy algorithm: explore methods that of will deliver the most efficient schedules to decide how final scheduler should be crafted. Any increase of margin is useful. − What type and level of schedule optimization is possible and needed? − What science drivers can be supported by a simulated schedule that will complete in 10 years? What cannot? This includes running a wide range of cadence models, and discussion of expectations and limitations with the community. − Modularize code anticipating transformation to a scheduler that links to the OCS. − Make code exportable and supportable for experimentation by community

23. 23 FINAL DESIGN REVIEW | TUCSON, AZ | OCTOBER 21-25, 2013 OCS Software Scheduler Design Concepts These concepts come from the Opsim software development:  Fully automatic operation  Dynamic adaptation to environment and observatory conditions  Sky field map, tiling regions  A target is a sky field/filter combination  Weather and sky conditions from telemetry  Sky brightness dynamically modeled for each sky field  Comprehensive observatory kinematic model for slew time optimizations  Fully configurable set of concurrent competing science programs  Some science programs optimize visits depth distribution  Some science programs optimize visits interval distribution  Some science programs push for time interval/filter sequences  Scripted visit sequences are also possible on a science program  Next targets selected according to a ranking schema  Target rank balances science value and time cost  Observations database  Scheduling parameters can be fine tuned to direct the survey

24. 24 FINAL DESIGN REVIEW | TUCSON, AZ | OCTOBER 21-25, 2013