LSST SAC phonecon Friday, September 19, 2014 Attending: SAC members: Renu Malhotra, Harry Ferguson, Lisa Hunter, Mansi Kasliwal, Rachel Mandelbaum, Bhuvnesh Jain, Amy Mainzer, Beth Willman, David Kirkby, Chris Hirata, Lucianne Walkowicz, Michael Strauss (Chair). Guests: Steve Kahn, Zeljko Ivezic, Andy Connolly. ******Discussion of the status of the Operations Simulator The main focus of our previous face-to-face meeting in Phoenix was a discussion of the plans for exploring options for the LSST cadence. We invited Andy Connolly (U. Washington), who leads the LSST simulations effort, to explain the status and future plans of the Operations Simulator (OpSim), which simulates the set of pointings that LSST will make over the 10-year survey. Connolly: OpSim was the first piece of simulation code that we wrote. It uses a so-called "greedy algorithm", whereby all possible pointings on the sky are ranked according to a set of rules (including slew time, how recently that area had been observed previous, airmass limits, etc), and the highest-ranked field is observed next. These ranks are then recalculated, and again the highest-ranked field is observed. This has been quite effective initially, and it gives a quite uniform survey at the end of 10 years. However, it is clearly not optimal with respect to temporal sampling. In talking to folks responsible for operations for various observatories, they all stressed the need for good metrics to evaluate how effective any given scheduler might be. These metrics should be science-based, and need to be invented, and vetted, by the scientific community. With this in mind, the LSST sims team developed the Metric Analysis Framework (MAF) for writing metrics and running them on the outputs of the OpSim. We now have a github repository for metrics (6 are in place thus far; we'd like more); it can be found at: https://github.com/LSST-nonproject/sims_maf_contrib Much thanks to the hard work from Phil Marshall, Lucianne Walkowicz and others in getting the github in place. The documentation of the existing metrics is pretty brief, and does need to be fleshed out further. One of the purposes of OpSim is as a tool to test and evaluate the scheduler, the code that LSST will use to actually schedule the telescope during operations. Indeed, OpSim is being upgraded this year to simulate the telemetry streams from Observatory Control. If we get this right, then the software can move smoothly to become the scheduler itself. For this year, we're going to continue supporting MAF. By the end of the year, we'll publicize and release a series of Tier 1 OpSim runs, which change some of the basic assumptions (such as airmass limits, u-band exposure times). Rolling cadence, whereby the area of sky imaged at a given time is smaller than currently ("everything available"), is further in the future. (There is a document describing ideas for rolling cadence at https://project.lsst.org/meetings/ocw/sites/lsst.org.meetings.ocw/files/OpSim%20Rolling%20Cadence%20Stratgey-ver1.3.pdf ) The list of the Tier 1 runs may be found here: https://confluence.lsstcorp.org/display/SIM/Cadence+Workshop+Simulated+Surveys More information on how one might install OpSim (still not user-friendly) and understand its configuration files may be found at http://www.noao.edu/lsst/opsim/docs/simulator/index.html and http://www.noao.edu/lsst/opsim/docs/simulator/configuration.html The next cadence workshop will take place next summer. A major goal before that time is to make the code easy to install, and to document the algorithms and the parameters that they use. However, this will not allow people to change the internal logic or algorithms easily. For example, the sharp boundaries currently between different programs in the observing cadence are set at the algorithm, not parameter level. Major algorithmic changes that are planned in the not-too-distant future include implementing rolling cadence (see above), and a more sophisticated algorithm than the "greedy algorithm", which can look ahead to select fields with a broader view of what is needed. The OpSim development team is quite small; progress is limited by manpower, so we don't have the resources to go beyond this. The OpSim team is planning a workshop bringing together folks from both industry and other observatories (Las Cumbres, HST, JPL) to talk about the scheduling problem and share insights. A full 10-year run of OpSim currently takes a large amount of compute power. There is a stretch goal of making the code parallelizable. If that happens, and the code is run at NERSC, this could speed things up quite a bit. There was a session at the cadence workshop to "think outside the box", but it was perhaps too early to do so, given that we're not ready to implement further algorithmic changes to OpSim. The SAC found this description of the status of OpSim enlightening, and urged the team to describe the timeline for further development and distribute this to the community, making it prominent on the LSST website. This document should include what the next tier of OpSim runs will be like. Indeed, the OpSim team should work with the LSST SAC when planning which simulations to run for Tier 2. Getting a good set of Tier 2 simulations in place is more important than making the OpSim code accessible to the community for minor tweaking. The important message is that the LSST cadence has not been set in stone; far from it, but it will take some time before we can explore significant xalternatives. In the meantime, people are encouraged to develop additional metrics. Steve expressed the opinion that while true mathematical optimization of the cadence is impossible, the number of free parameters that will end up mattering are relatively small; we ultimately want to have measures of short timescale repeat time, spatial homogeneity, and depth. Quantifying this in detail of course requires proper science-driven metrics. ******Update on the LSST Project (discussion led by Steve) While we were formally approved for construction on August 1, and the construction has formally begun, we are still working with the NSF to refine the budget, especially in contingency. We are working with a $473M "not-to-exceed" construction budget on the NSF side. The camera team is undergoing a major review in November. AURA has fiduciary responsibility for LSST; this is managed by the AURA Management Council for LSST (AMCL). While up until now, the AMCL and LSST Consortium Board have been quite close to one another, they are now becoming more distinct: the AMCL is focused on the finances, while LSSTC is increasingly focused on supporting the science. The LSSC has a new Executive Officer, Pat Eliason. She will be leading a discussion of what LSSTC can do, including supporting white paper/roadmap exercises, and raising funds to support science efforts within the community. Throughout all of this, the LSST SAC is the official representation of the scientific community to the Project. We ended the meeting with a brief summary of some relevant upcoming events. Harry Ferguson is leading a roadmap exercise for non-cosmology extragalactic science; a workshop will take place in Baltimore on October 6. And LSST is planning a booth for the upcoming Division of Planetary Meeting, in November in Tucson.