Face-to-Face meeting of the LSST Science Advisory Committee (SAC) Princeton University, April 7, 2014 Members of the SAC attending: Harry Ferguson, Niel Brandt, Ricardo Munoz, Lucianne Walkowicz, Rachel Mandelbaum, Renu Malhotra, Jason Kalirai, Mansi Kasliwal, Bhuvnesh Jain, Chris Hirata, Lisa Hunter (by phone), David Kirkby (by phone), Michael Strauss (chair) Regrets: Beth Willman, Dante Minitti LSST personnel: Steve Kahn (Director), Zeljko Ivezic (Project Scientist), Tony Tyson (Chief Scientist), Robert Lupton (Software Applications lead), Chuck Claver (Systems Scientist, by phone), Suzanne Jacoby (E/PO Subsystem manager, by phone), Victor Krabbendam (Project Manager, by phone) Outline and highlights: **Introductions and general remarks **Summary of project status **The Role of the SAC as a voice of the community; how it relates to other bodies within LSST **LSST Cadence: Current status; new approaches. **LSST pipeline outputs **Commissioning plans, and overlap with science **Diversity issues in the LSST workforce **********Introductions, interests and general remarks: We went around the room; each person introduced themselves and described some of their interests and concerns. Steve Kahn has been involved in LSST since he moved to Stanford in 2003. He has a background in X-ray astronomy, but his scientific interests are now focussed on cosmology, especially weak lensing. He has been LSST Director since July 1, 2013. Lucianne Walkowicz (Princeton) is co-chair of the Transients/Variable stars science collaboration. She is specifically interested in transients and variable stars, and the question of follow-up. Jason Kalirai (STScI) is interested in aspects of stellar astrophysics and stellar populations. He is worried about the NSF funding paradigm (largely driven by individual investigator proposals), and how it will work to support LSST science. He is Project Scientist for JSWT, as well as WFIRST/AFTA, and wants to explore how to develop synergies between them and LSST. Rachel Mandelbaum (CMU) has expertise in weak lensing, as a tool to understand the distribution of dark matter. She is interested in the mitigation of systematic effects in weak lensing, and wants to encourage increased interaction between the data management team and the scientific community. Renu Malhotra (Steward) is interested in solar system dynamics and planetary migration. She gave a presentation (Malhotra.pdf) describing the opportunities for Solar System science with LSST, including asteroids, outer satellites of the giant planets, comets, Kuiper Belt Objects, with emphasis on the azimuthal distribution of all these populations. Chris Hirata (OSU) is interested in cosmology as done by wide-field surveys, and is working on methods for calculating galaxy N-point correlation functions. Bhuvnesh Jain (U. Penn) is interested in cosmological problems, including weak lensing and testing modified gravity theories. He is spokesperson for the Dark Energy Science Collaboration (DESC), and is also involved in the CTIO Dark Energy Survey (DES). The DESC formed two years ago, and has about 250 members. He sees this as a paradigm for a new spirit of collaboration in the scientific community. He is also interesting in improving ties and communication between the scientific community and LSST Data Management team. Mansi Kasliwal (Carnegie) has been involved in the Palomar Transient Factory and its successor, the Zwicky Transient Factory, and is interested in time domain science. She is particularly interested in exploring alternative cadences more conducive to transient discovery than the universal cadence. Harry Ferguson (STScI) is chair of the Galaxies Science Collaboration, and is a member of the LSST Corporation Board committee on enabling science. He has led several of the Hubble Deep Fields, including CANDELS. He is particularly interested in coordinating follow-up to LSST discoveries, as well as parallel multi-wavelength surveys to complement LSST. Niel Brandt (Penn State) is chair of the AGN Science Collaboration, and chairs the working group on LSST Deep Fields. He is interested in deep X-ray surveys. Ricardo Munoz (U. Chile) represents the Chilean astronomical community. He is interested in Galactic structure and ultrafaint galaxies. Michael Wood-Vasey (Pittsburgh) recently stepped down as the co-chair of the Supernova Science Collaboration; he also leads the DESC supernova effort. He is the SDSS-III spokesperson, and thus knows that communication in a large collaboration is challenging. His scientific interested are in SN Ias in particular, and surveys and image processing in general. Lisa Hunter (UC Santa Cruz) is the director of the Institute for Science and Engineer Educators, and is interested in workforce development issues, and questions of diversity in both the workforce and the user community. She points out that these issues are distinct from, although related to, those of education and public outreach. She has worked on these issues with various telescopes in Hawaii. David Kirkby (UC Irvine) comes from a particle physics background, but is now increasingly involved in large-scale structure studies, including Ly alpha forest studies with BOSS, eBOSS, and DESI and weak lensing with DESC. Chuck Claver (LSST) is LSST Systems Scientist; he's been involved in the project since 1998. He has scientific interests in stellar populations and astrometry, and the technical challenges of PSF determination in the context of weak lensing. Tony Tyson (UC Davis) is the father of LSST. He poses two questions to this committee: -What metrics should we use to monitor the LSST data quality as it comes in? -How can we best organize scientists around specific scientific problems? Robert Lupton (Princeton) is leading the team that is developing the software algorithms to measure the properties of objects detected in the imaging data. He is interested in image processing, and the challenge of developing the community's ability to become familiar with and use the LSST data and software tools. ****************Status of Project First, some vocabulary: The LSST Project Office (LSSTPO) is the body formally charged with the LSST construction effort. It is overseen by AURA, via the AURA Management Committee for LSST (AMCL). The LSST Consortium (LSSTC) is the group of universities that has led the project until now; it now sees its role as leading the science enabling effort with LSST. With the start of construction, the distinction between LSSTC and LSSTPO is important. Victor Krabbendam gave a presentation summarizing the status of the project; see Krabbendam.pdf . The project has passed the NSF Final Design Review; the current NSF budgeted cost is $473 million (in then-year dollars). This represents a $15M cut from the original plans; that money will be saved by tightening up the schedule for commissioning. Engineering first light will be in 2019. The full camera will not be ready then, so a single-raft (9-CCD) camera will be built to help with commissioning. This will be a powerful survey instrument in itself, with half the etendue of the DES camera. This is a time of rapid hiring, as we ramp up to the start of construction (officially starting July 1, 2014). The LSST Project Office is likely to be invited to put in a bid (perhaps due in 2017) to operate the telescope once construction and commissioning are complete. The current estimated annual cost for operations is $36 Million per year, in 2011 dollars. These funds will come from a combination of NSF, DOE (who together will support roughly 2/3 of the costs), and buy-in from a variety of international affiliates. We already have letters of intent from a large number of institutions representing 17+ countries, and we're in the process of turning these into formal Memoranda of Agreement (MOA). DOE operations money will go at least partly through SLAC; they are interested in Dark Energy science, and will support scientific investigations for this, at the tune of perhaps $6M/year in *addition* to the operations costs above. NSF, on the other hand, is purely proposal-driven, and they seem unlikely to budge from this policy, at least for the time being. That is, LSST science proposals will have to compete with all other proposals asking for funding. There is interest in NSF Physics, as well as NSF Astronomy, to support LSST science projects. DOE science funding is over-subscribed by a factor 2-3, quite a bit better than NSF astronomy. We discussed the possibility of raising funds for science from private foundations: the LSST Corporation is the obvious mechanism to make this happen. For more details on all of this, see Victor's presentation (Krabbendam.pdf) ********Steve Kahn: Presentation on the SAC's role See Kahn.pdf The SAC is the formal vehicle for interactions (in both directions) between the LSST Project and the scientific community. As such, our deliberations must be public and visible. The SAC members will have staggered 3-year terms (some people will have to have longer terms in the beginning) and the intent is to continue the SAC into LSST operations. We are representatives of the community and ambassadors of the project, and the list of our names will be public: we should encourage people from the community to contact any one of us if they have concerns, suggestions, or advice. We are not an oversight board; we advise the Director, but if we have an issue we're unhappy about, we have the power to demand a response from the project. Our discussions will fall into several categories: -Information and updates about the project; -Requests from the project for our (and the community's) scientific opinion and feedback on various choices the LSSTPO makes during construction; -Concerns, issues, and questions that we (or the community) bring up; -Requests from the project for our (and the community's) opinion and feedback on policy issues. The LSST represents the culmination, as it were, of a cultural shift in the nature of optical astronomy, with increasing emphasis on large collaborations and working with large datasets. Moreover, the increase in sample sizes that LSST will enable will allow qualitatively new science opportunities: "More is Different". There is another body, the LSST Project Science Team (PST), chaired by Zeljko Ivezic and consisting of people paid by the project (including various of the subsystem leads), whose job it is to respond to specific technical decisions for the project and understand their scientific import. The SAC is related to, but distinct from, the LSST Science Collaborations (of which there are currently 11). The Science Collaborations were started in about 2005, and were initially fairly tightly controlled by LSST. But as we make the transition into construction, the science collaborations are now becoming self-governing, and setting their own membership rules (with the only restrictions being that those rules should be well-documented, and that membership is restricted to those with LSST data rights, i.e., currently members of the US and Chilean scientific community, as well as specific individuals mentioned in the small number of MOA that have been signed to date). *********Zeljko Ivezic: The LSST Cadence See Ivezic.pdf The specific design of LSST: the size of the primary mirror, the field of view, the exposure times (all observations are taken as a pair of 15-second exposures, termed a visit), the number of filters, and the number of visits over the course of ten years, follow from the science requirements of the project, and are quite tightly constrained. This is described in some detail in http://adsabs.harvard.edu/abs/2008arXiv0805.2366I, and in the presentation itself. The science requirements flow down from the core science goals of the project, and in turn are translated into system requirements. The survey footprint is defined by that region with airmass < 1.4 from the observatory site, with extensions to the Magellanic Clouds, and the northern reaches of the ecliptic plane; the main survey covers 18,000 deg^2. The operations simulator (OpSim) was used to demonstrate that we can meet our science requirements, and spends most of the time on the so-called Universal Cadence, whereby essentialy all the available sky at a given time is observed over the course of three nights in a given filter, before repeating (perhaps in another filter). It is now time to explore alternative cadences. In particular, for time-domain science, more rapid cadences over smaller areas of sky are definitely worth exploring. Improved dithering of fields can greatly improve the uniformity of the resulting imaging data. We also need not use a single cadence pattern for the entire 10 years of the survey. OpSim is being refined and sped up to allow exploration of these questions; LSST is sponsoring a workshop to discuss the details in the second week of August in Phoenix. One challenge in this context is to develop quantitative metrics for different science cases, so that, for example, we can directly compare two cadence models for their ability to measure stellar parallax over a ten-year survey. This led to a detailed discussion of the decision process for finalizing the cadence: who has the authority to decide what we're going to do? How will different science drivers be balanced? How can specialized observations (such as deep drilling fields, or targets of opportunities from gravitational wave detections), be incorporated? Do we need to stay with the same cadence algorithm for the entire 10-year survey? How can we get buy-in from the community on all of this? And who is going to do all the work of exploring the (infinite-dimensional) space of different possible cadences? These are going to be key questions as we go forward. ********Robert Lupton: Outputs of LSST pipelines See Lupton.pdf Each visit LSST makes will be subtracted from a template image of that area of sky; objects that differ significantly (>5 sigma) will be released immediately (within 60 seconds of closing the shutter) to the world as "Level 1" data products, allowing transients and variables to be identified (we expect several million such alerts to be issued each night). Level 2 data products are produced once per year and include a stack of all data taken to date; this gives a detailed picture of the static sky. Level 3 refers to software that the scientific community will write, which will run on LSST computers; a real challenge will be educating the community about the LSST software system so that they can write code that can integrated into the code base and run as Level 3. An example of Level 3 software will be sophisticated event brokers that sorts through all the Level 1 alerts and categorizes them into classes: asteroids, variable stars of various sorts, supernovae, AGN, and (perhaps most interesting) unclassifiable. The detailed planned outputs of the pipelines are described in Robert's presentation; see http://ls.st/pt5 for all the details. We discussed a number of details, such as the optimal processing of multiple images of a given area of sky (not from the coadds, but by fitting a global nodel to each image separately, each with its own PSF), how the algorithms will get documented, and how artificial objects can be added to the data stream to assess the completeness of various categories of objects. *******Chuck Claver: LSST Commissioning plans See Claver.pdf The commissioning plans are ultimately tied to the system requirements: each of these requirements has one or more corresponding tests that we need to pass. The camera will be ready for installation somewhat after the telescope will be ready, and thus we will carry out early commissioning observations with a single-raft (40 arcmin) camera. All aspects of the LSST system need to be commissioned: the telescope, camera, data systems, image-processing, database, user interface.... There will be a tightly run commissioning team, with an emphasis on keeping the number of people on the mountain at any given time at the minimum necessary to operate the telescope. Commissioning comes in two parts, engineering and science, and the division between the two will not necessarily be very clean. But even the one-raft commissioning camera will have half the etendue of DECam, and the data obtained during commissioning will be very valuable indeed. This leads to all sorts of questions: how can scientists get involved in commissioning? Can they do science with commissioning data? How will the data be distributed (before the pipelines, databases, and user interfaces are fully working)? Can we release data that are not survey quality, and indeed not fully characterized? How will we respond to the multiple questions that will inevitably arise, and manage expectations about the inevitable problems that people will find? What if commissioning goes better than planned; could we imagine starting the survey early? Could we imagine in this case having another few months of commissioning after, say, 3 years of observations? As the system nears operational readiness, we intend to spend a few months in full survey mode to demonstrate that is working well. For context, once the full camera is in place, LSST will do the equivalent of the full DES survey in about 1.5 months. Where should we point? With what cadence? There was a lot of interest in going to full 10-year depth over a limited area of sky, or working in movie mode (repeatedly observing at a given pointing in a single filter) for several hours. We could imagine putting out a call to the community for specific commissioning observing ideas (it will be interesting to decide how to vet such proposals!). This led to a discussion of coordination of LSST with other resources. By late 2022 (when the LSST survey proper is scheduled to start), JWST will already be in its 5th proposal cycle; commissioning data will be particularly valuable for selecting targets for that. More generally, synergies between LSST and WFIRST, Euclid, and other facilities need to be exploited. We ended this discussion with a discussion of whether there should be any restrictions on releasing commissioning data, and some of the bad feelings other facilities (Spitzer, Kepler) engendered when they made their early data available only to limited number of people. Chuck plans a detailed review of the commissioning plans in 2016; all these questions will be the continuing subject of discussion by this group between now and then. ********Suzanne Jacoby: Diversity issues in the LSST workforce and among the user community. See Jacoby.pdf LSST has a problem: while the membership of the science collaborations has a reasonable gender balance, last year's Joint Technical Meeting included only 6% women. But as the project ramps up to construction, it will be hiring 40 or more people over the next few years at a variety of institutions, and thus has an opportunity to at least partly mitigate this imbalance. We will have to coordinate hiring across the project, on such things at recruitment. On longer timescales, we can play a leadership role in training the next generation of people with the engineering, software, and scientific skills that LSST will need. We discussed the possibility of professional development workshops (perhaps following the model of the European Training Network, in which a cohort of students takes part in multiple workshops spread out over months or years), and ways to mentor undergraduate and graduate students who are interested in these areas. The big collaboration aspect of the project has the potential to be very interesting to undergraduates. A related question is to make sure that desireable career paths exist for those who work, e.g., in astronomical software: could LSST enter into partnerships with universities in which top people are hired as faculty, and supported in part by LSST? We should look for ways to further broaden the appeal of LSST to the general public using modern social media: JWST has over 100,000 twitter followers, for example. We finished the meeting with a brief discussion of communication mechanisms we can use to inform the community about the activities of the SAC, such as LSST E-News and the AAS Newsletter.