Optimization of the Observing Cadence for the Rubin Observatory Legacy Survey of Space and Time: a pioneering process of community-focused experimental design

Vera C. Rubin Observatory is a ground-based astronomical facility under construction, a joint project of the National Science Foundation and the U.S. Department of Energy, designed to conduct a multi-purpose 10-year optical survey of the southern hemisphere sky: the Legacy Survey of Space and Time. Significant flexibility in survey strategy remains within the constraints imposed by the core science goals of probing dark energy and dark matter, cataloging the Solar System, exploring the transient optical sky, and mapping the Milky Way. The survey's massive data throughput will be transformational for many other astrophysics domains and Rubin's data access policy sets the stage for a huge potential users' community. To ensure that the survey science potential is maximized while serving as broad a community as possible, Rubin Observatory has involved the scientific community at large in the process of setting and refining the details of the observing strategy. The motivation, history, and decision-making process of this strategy optimization are detailed in this paper, giving context to the science-driven proposals and recommendations for the survey strategy included in this Focus Issue.


INTRODUCTION
Vera C. Rubin Observatory is designing an astronomical survey, the Legacy Survey of Space and Time (LSST) that will be revolutionary in many ways. The amount of imaging data, and the combination of flux sensitivity, area, and the temporal sampling rate will be dramatically increased compared to precursor surveys at any waveband. Rubin Observatory will observe the southern hemisphere sky from El Peñón peak of Cerro Pachón, Chile, for 10 years, using the 8.36 m aperture Simonyi Survey Telescope and the LSST Camera with its unique 9.6 deg 2 field of view to collect over two million sky images. It will do so with multiple filters (ugrizy) and with exquisite sub-arcsecond image quality.
Rubin Observatory's enormous dataset, which will reach a size of ∼ 300 PB when the 10 year survey ends, will be released with a unique data policy. Every night, millions of LSST alert packets that identify astrophysical transient, variable, and moving objects will be released worldwide in real time with no restrictions. The LSST images and catalogs will be available in their entirety to all Rubin data rights holders: any US and Chilean scientist and members of International Groups that have agreements with the National Science Foundation (NSF) and/or the Department of Energy (DOE) or their managing partners the Association of Universities for Research in Astronomy (AURA) and SLAC National Accelerator Lab, respectively. 1 . These include catalogs available within 24 hours of observations, and annual releases of reprocessed images, deep coadded stacks, and associated catalogs 2 . These data will become available worldwide after two years from the original data release. This enormous data set, combined with an open data access policy, sets the stage for a broad and diverse user community and a commensurately huge opportunity for maximum scientific impact from the LSST.
LSST is designed to enable the pursuit of four main science themes: probing dark energy and dark matter, exploring the transient optical sky, mapping the Milky Way Galaxy, and building a catalog of Solar System objects over an order of magnitude larger than presently available. These science goals were chosen to drive the survey design such that a transformational survey addressing all these science themes will also be able to impact many other fields of astrophysics, including, for example, galaxy morphology and evolution, AGN, and quasar studies, etc. LSST science drivers and technical design details are summarized in the LSST overview paper (Ivezić et al. 2019).
Rubin Observatory is constructing a flexible scheduling system that can respond to the unexpected and be re-optimized as the survey progresses. LSST observations will be scheduled automatically, with the scheduling algorithm designed to maximize scientific return under a set of observing constraints. LSST is in fact an umbrella term to refers to a set of surveys that Rubin will perform in its first 10 years: a "main survey" (hereafter referred to as Wide-Fast-Deep, WFD), a series of pointings that will be observed with an intensified cadence leading to deeper coadded image stacks, the Deep Drilling Fields (DDFs), and additional "minisurveys" that cover specific sky regions such as the ecliptic plane, Galactic plane, and the Large and Small Magellanic Clouds, or that vary survey parameters such as the depth of a single visit.
Any implementation of LSST's 10-year observing strategy must meet the basic requirements described in the LSST Science Requirements Document (SRD 3 ) to ensure that the core science goals will be achieved: a footprint for the WFD of at least 18,000 deg 2 which must be uniformly covered to a median of 825 nominal 30-second visits per 9.6 deg 2 field, summed over all six filters. Additional constraints on the temporal distribution of these visits are derived from requirements on parallax and proper motion accuracy, as well as a requirement for two visits per night to enable discovery of main-belt asteroids using standard algorithms. Yet, the SRD intentionally places minimal quantitative constraints on the observing strategy, recognizing that science evolves and the LSST plan can be refined until first light and even beyond. To that end, the Rubin Observatory operations team plans to continuously monitor survey progress and, if needed, to modify the strategy to achieve the desired scientific goals (see Section 5). The SRD leaves significant flexibility in the detailed cadence of observations within the main survey footprint, including the distribution of visits within a year, the distribution of images between filters, and the definition of a "visit" itself (e.g., a single exposure or multiple exposures per visit). Furthermore, these constraints apply to the WFD only. Depending on the performance efficiency of the WFD, between 10% and 20% of the sky time will remain available for DDFs and minisurveys, whose design is not strongly prescribed by the SRD.
A nominal survey footprint showing the distribution of visits across the sky, including WFD, some DDFs, and some potential minisurveys, is shown in Figure 1. Even a core parameter such as the median time elapsed between observations of the sky in different nights can be varied significantly for the WFD within the constraints of the SRD. Figure 2 shows the distribution of inter-night gaps, a critical parameter to enable transient science, in two WFD survey simulations: a baseline proposal, i.e., a straightforward implementation of the survey design as described above, and a rolling cadence strategy, where the ∼ 825 visits for each sky pointing in the WFD are distributed unevenly over the 10-year LSST timeline, front-loading certain areas of the sky with higher density of observations early on, to later give way to intense observing of others. The rolling cadence is further discussed in subsection 4.1 and a more detailed description of the concept of rolling cadence and discussion of its trade-offs can be found in section 2.5 of LSST Science Collaboration et al. 2017 and 4.9 of Jones et al. 2020 as well as inline at this URL 4 .
With LSST likely to start in 2024-a delayed re-baselined schedule due to the COVID-19 pandemic---Rubin Observatory is undertaking the final planning for the initial observing strategy. To ensure that the survey science potential is maximized and that the survey design does indeed serve the broad scientific community that will have access to the data, Rubin Observatory has, to an unprecedented degree, involved the scientific community in the survey design itself. The general survey footprint must meet the SRD design specifications: the WFD covers 18, 000 deg 2 to a median of 825 visits. Other surveys, like the Deep Drilling Fields (DDFs, which appear in yellow due to the higher number of observations), North Ecliptic Spur (NES), Galactic Plane (GP, which contains a large region on the bulge side, but also a smaller footprint on the anti-bulge side), and South Celestial Pole (SCP) that are shown in this figure, cover regions outside of the WFD footprint and have more flexibility in the observing strategy, including the number of visits and their distribution among filters. The initial survey footprint was envisioned as in the left panel (map generated from OpSim baseline nexp2 v1.7.1 10yrs, see Section 2), but details of the visit distribution for optimal science return are still being determined, including if this general survey footprint should be modified. Expansion of coverage into additional low dust extinction area and higher coverage of a portion of the Galactic plane, for example, may result in a footprint conceptually similar to what is shown on the right (generated from OpSim footprint 6 1.7.1 10yrs). This figure is discussed in Section 1.
To place in context publications containing sciencedriven proposals and recommendations for the survey strategy that are included in this Focus Issue, we describe here the LSST survey cadence optimization process. The process has been conducted openly and in close contact with the survey stakeholders and the scientific community at large.
The ongoing pre-operations phase of cadence optimization has included three major contributions by the science community: • Early community engagement efforts, including the creation of software that enables the community to get involved in survey design. These are described in Section 2 and Section 3. This phase culminated in the Community Observing Strategy Evaluation Paper (LSST Science Collaboration et al. 2017, hereafter COSEP), briefly described in subsection 3.2; • Cadence White Papers, solicited in 2018, briefly described in Section 4; • Cadence Notes, solicited in 2020, and discussed in Section 5.
This Focus Issue provides the community with an opportunity to present the work that underlies many of these science-driven cadence recommendations. We highlight some lessons learned so far in this community-focused process for survey design in Section 6 and Section 7.

OPEN SOFTWARE TO ENABLE COMMUNITY ENGAGEMENT
To empower the community to make knowledgeable, science-driven recommendations, Rubin Observatory has released a series of simulated LSST pointing histories, using different survey strategies, and open-access software to generate quantitative analyses of these simulations. These software tools originated as part of the overall Rubin Observatory simulations effort (Connolly et al. 2014), and have evolved over time with considerable community input.
To generate potential realizations of the LSST surveys, Rubin developed a simulator that works with the LSST scheduler, collectively known colloquially as the Operations Simulator, or OpSim for short (Naghib et al. 2019;Delgado & Reuter 2016;Delgado et al. 2014), which can generate databases containing 10-year pointing histories, complete with weather, seeing, and sky brightness information (Yoachim et al. 2016). A short sample of a resulting simulated observing history is shown in Figure 3. The version of the scheduler used at the time of writing, generally referred to as the 'Feature Based Scheduler' or FBS, is the fifth generation . Distribution of median internight visit gaps (the time elapsed between visits to the same field in different nights) at a given location in the sky for two simulated LSST OpSim strategies: baseline nexp2 v1.7.1 10yrs (left) and rolling nm scale0.90 nslice2 fpw0.9 nrw1.0v1.7 10yrs (right, for more details on the simulations see Section 2). Each curve shows the median internight gaps across pointing positions distributed on a healpixel grid with N sides = 64 (corresponding to ∼ 0.85 sq. deg.). In the top row the distribution is shown for all observations, in the following rows for observations in each of the six Rubin LSST filters (as labeled). The distributions are smoothed via Kernel Density Estimation (Rosenblatt 1956;Parzen 1962); the location of the peak of the underlying distribution is indicated to the right of each curve. This figure shows that different OpSim runs, developed in compliance with the SRD requirements, can be very different even in core features. The figure is discussed in more detail in Section 1.
of scheduler codes developed for the LSST. 5 In general, this software is used by Rubin Observatory to produce and release databases of observations following recommendations from the community. 6 Sets of OpSim runs are released as "families" in which survey strategy parameters are varied along a specific simulation axis. For example: the wfd scale family explores allocating different fractions of the overall survey time to the WFD, the footprint family of simulations modifies the survey footprint according to different recommendations, and the filter dist family varies the distribution of visits in the survey between different filters. 7 The second open-access software package released by Rubin to facilitate community engagement in survey design is the Metric Analysis Framework (Jones et al. 2014, hereafter MAF ), which is intended to enable the calculation of metrics pertaining to the observations in a standardized and easily extensible manner. The MAF provides a simple API for calculating image properties (e.g., seeing, proper motion) over a variety of spatial scales as well as tools to predict the properties of photometric measurements for objects implanted in the survey simulations (including light curves). The MAF thus allows the user to create metrics associated with specific sci-5 https://community.lsst.org/t/from-opsim-v4-to-fbs-1-2/3856 6 https://community.lsst.org/tag/opsim 7 The name of the OpSim runs are composed of a root indicating the OpSim family (e.g., baseline or footprint), the specific parameters of the run within the family, the OpSim release (e.g., v1.7.1), and the duration of the simulation (e.g., 10yrs).
ence goals, and to evaluate them over existing survey simulations (the OpSim runs). That is: the user can test scientific cases and identify effective (or ineffective) cadences associated with them. Along with the API, many ready-to-run metrics, generally referred to as MAFmetrics (hereafter simply metrics), have been released. Furthermore, the most advanced metrics have been the result of collaborations with or direct contributions from the community 8 in the three phases of survey evaluation described below. The outputs of standard metrics on existing OpSim runs are provided to Rubin Observatory and made available online to the community to help guide survey strategy choices. 9

COMMUNITY ENGAGEMENT IN LSST CADENCE OPTIMIZATION -PRE-2018
Planning for LSST has been undertaken hand-in-hand with the community from the start of the project. For example, the choice of four main science themes was guided by the community-wide input assembled in the report of the Science Working Group of the LSST in 2004. 10 Similarly, the current design of the DDFs was driven by a set of eight science white papers 11 written 8 https://github.com/LSST-nonproject/sims maf contrib 9 The outputs of survey metrics are available at http:// astro-lsst-01.astro.washington.edu:8081 10  in 2011 by about 75 members of the LSST Deep-Drilling Interest Group. This led to the current selection of four DDF fields that maximize synergy with legacy data and ongoing surveys to enable a number of extragalactic science investigations. 12 The Rubin Observatory Science Advisory Committee 13 (SAC) has been providing valuable advice to Rubin on cadence issues since its inception in 2014.

The LSST Scientific Community and the Science Collaborations
Rubin Observatory and its construction and operation teams include 100's of people supported by investments of the NSF and DOE 14 to create the Observatory, and design and run the LSST, including preparing, distributing and supporting the usage of data and data products from the survey. Rubin's plan to maximize science from the LSST includes crucial support, engagement, and coordination with the scientific community. The community, in turn, needs funding support from all possible sources, including the US agencies and philanthropic organizations, to produce "User Generated Data Products" that drive the science analyses to the fullest extent embraced by Rubin's scientific vision and mission. User Generated Data Products come from the community and take the Prompt and Data Release Products produced by Rubin to the next level to ensure the mission of Rubin Observatory is fulfilled. These data products and analyses can only happen through a high level of planning, coordination, and effort from the scientific community.
Embracing these exciting challenges and promises of the LSST, the scientific community has organized into Science Collaborations (SCs). The SCs were formed to provide a forum for the community to interact with Rubin Observatory's construction team (then the LSST Project), and to make the scientific case to be presented to the 2010 Decadal Survey (LSST Science Collaboration et al. 2009). Today, the SCs are eight independent teams, self-governed and self-managed, that include over 1000 scientists from six continents 15 (Bianco et al. 2019). They work in close contact with Rubin to prepare to turn the LSST data into science and to help the Observatory make scientifically informed choices. Thus the SCs have been optimally placed to be core contributors of cadence recommendations throughout the cadence optimization process. For example, early survey strategy investigations using some of the initial OpSim simulations were driven by SCs, including Dark Energy Science Collaboration (DESC) analyses of potential dither patterns (Carroll et al. 2014;Awan et al. 2016) that motivated a random nightly shift of pointing centers to become the Feature Based Scheduler (FBS) default. Nonetheless, Rubin has sought input on cadence optimization from the entire scientific community, and suggestions have been welcomed from any group, both within and outside of the community of data rights holders.

Community Observing Strategy Evaluation Paper
Members of the LSST science community gathered in 2015 to help design an observing strategy that would maximize the scientific output of the survey. This community includes scientists primarily (but not exclusively) drawn from people engaged in the LSST SCs and LSST construction Project.
The core product delivered by this group is the COSEP, a document co-authored by over 100 scientists and currently over 300 pages long, that lists a compendium of ideas and results. It is designed as a living document with the aim to bring together the group of people who are thinking about the LSST observing strategy problem, and facilitate their collective discussion. 16 It provided a first venue for evaluations produced by the community through the systematic use of the MAF and enabled the evaluation of the survey throughput based on a set of simulations representing small variations of the baseline survey for a variety of science cases. In order to standardize various constraints derived from diverse science cases and enable comparisons, each section contains the explicit answers to ten questions (available in full in Section 7) designed to examine constraints and trade-offs between, for example: sky coverage and depth; uniformity and frequency of sampling (e.g., a rolling cadence); single-visit depth and number of visits; Galactic Plane coverage (spatial, temporal, or depth); DDF sampling and depth; fraction of observing time allocated to each band; pairing of filters; and any requirements placed on commissioning.
The COSEP is articulated in nine topical chapters that cover 25 specific science cases. Additional sections discuss minisurveys (e.g., covering specific sky locations and proposing modified exposure parameters), 16 The GitHub repository containing the living source for the COSEP is https://github.com/LSSTScienceCollaborations/ ObservingStrategy . The x-axis shows time away from midnight; the y-axis separates visits by the day of the month. The gray edges of the plot indicate twilight time, with observations continuing to −12 degree twilight in redder filters. Astronomical (−18 degree) twilight is shown with light gray, −12 degree twilight is dark gray, and periods where the Sun is above the horizon are in black. The time of lunar transit is shown as a wide yellow stripe; lunar rise and set times are shown by dotted yellow lines. Redder filters are used when the Moon is up and bluer filters when the Moon is down and the sky is dark. Each 30-second visit is separated from adjacent visits by thin white vertical lines indicating the time between exposures, which depends on the time it takes the telescope to slew to a new pointing and it is typically short (and often not visible in practice here). Thicker lines indicate longer than typical slew and setup times; for example, the time to change filters is about two minutes. Weather and telescope maintenance are also simulated: observing time lost to weather or telescope maintenance appears as larger white gaps on days 19 and 26. This figure is discussed in Section 2.
and plans for coordinating LSST observations with other missions (e.g., the Nancy Grace Roman Observatory, Spergel et al. 2015, then WFIRST). The SAC reviewed this work and made recommendations to Rubin that shaped the next phase of the survey strategy optimization, including: 1. The Rubin construction Project should implement, analyze, and optimize the rolling cadence idea (driven by supernovae, asteroids, short timescale variability).
2. The Rubin construction Project should execute a systematic effort to further improve the ultimate LSST cadence strategy, including optimization of the sky coverage, u-band depth, filter pairing, minisurveys, DDFs, etc.

THE 2018 CADENCE WHITE PAPERS SOLICITATION
In part as a result of the SAC recommendations, and with an overarching goal of maximizing the science impact of the LSST, in 2018 Rubin Observatory issued a call to the entire scientific community for white papers on survey strategy ideas. Although the COSEP is a living document that can continue to be updated, this call expanded the reach of the community engagement in the survey optimization process. The call 17 solicited discussions of science cases in which Rubin could have an impact and associated science-driven cadence suggestions for the main survey, the minisurveys, the DDFs, as well as the first Target of Opportunity strategy suggestions for Gravitational Wave counterpart detection with Rubin (Margutti et al. 2018). It was supported by a new and more extended set of OpSim runs, itself enabled by progress in the simulator development 18 (Section 2).
The aim of the Cadence White Papers call was to create a portfolio of survey ideas, to be vetted and prioritized by the SAC, that would become the basis for a larger and more comprehensive set of OpSim runs. The 46 submitted white papers 19 represent a wide swath of the astronomical community, and, together with the COSEP, shaped the next stage of the survey strategy evaluation. Most submissions arose within SCs (Figure 4,left) and many are the result of collaborative work across SCs, as demonstrated by the inter-connectivity of the authors' network ( Figure 4, right), but contributions were submitted as well by authors outside the SCs and interest groups related to other surveys (e.g., the Nancy Grace Roman Observatory, then WFIRST, Spergel et al. 2015 andEuclid, Capak et al. 2019). The contents of these white papers were distilled into several areas for investigation by the SAC in their 2019 report. 20 The ongoing cadence optimization process discussed in detail below is a direct result of those recommendations. The SAC also identified some vulnerabilities in the process: "many of the white papers only outline what an appropriate metric would be, and coding these up will require substantial effort.  how this work will be done;[...] we recommend that the OpSim team be given the resources to code the metrics suggested in the white papers [...]".

Survey Simulations and Cadence Optimization following the White Papers
Following the SAC response to the 2018 White Papers, the Rubin survey strategy team produced a series of simulations exploring the particular survey strategy options recommended for further investigation. These include investigations into the survey footprint, the amount of survey time devoted to WFD observations, various options for pairing visits within a night or adding more nightly visits, the exposure time per visit, enforcing seeing requirements at each point in the sky in certain filters, dithering options for the DDFs, and exploring the effect of adding specific minisurveys, such as short exposures, twilight Near Earth Object discovery visits, or additional visits at high airmass for better Differential Chromatic Refraction measurements, and more.
Sets of these new simulations were released in groups. For this call, a total of 173 OpSim runs were made available under OpSim releases 1.5 to 1.7.1. These simulations and their analysis with sets of standard metrics are described in detail in Jones et al. (2020, hereafter LSST document PSTN-051). 21 The metrics written to analyze the science return of these simulations through the MAF API, both those created by the Rubin survey strategy team and those created by the community, need to be ultimately combined 21 PSTN-051 is available at https://pstn-051.lsst.io. into a comprehensive view of an OpSim run that can enable comparative studies and, in the end, the definition of the survey design. This is a complex exercise in and of itself, which is at the heart of the last phase of the survey cadence optimization process described in Section 5. An example of a visualization of a subset of metrics indicating the effect of rolling cadence on a variety of science areas is shown in Figure 5.

THE 2020-2022 OPTIMIZATION PHASE
With the publication of PSTN-051, Rubin (including now both the construction and the early operations teams) has started the final phase of the LSST cadence optimization before the start of the 10-year survey with the creation of the Survey Cadence Optimization Committee.

The Survey Cadence Optimization Committee
The Survey Cadence Optimization Committee (SCOC) is an advisory body to the Rubin Observatory Operations Director . The committee was formed in 2020, and it will be a standing committee throughout the life of Rubin Observatory operations. The SCOC is responsible for optimizing the LSST cadence within the constraints imposed by the observing system, observing conditions, science drivers, and scientists invested in its mission and legacy. Its tasks are as follows: • Make specific recommendations for the initial survey strategy for the full 10-year survey and disseminate these recommendations via public reports and on-going engagement with the community.
• Make specific recommendations for "Early Science" observations, which might be carried out Figure 5. Illustration of a "radar plot" comparing science metrics from a variety of core science areas for five OpSim realizations (as indicated in the legend). Each corner of the hexagon represents a different metric, whose value is mapped to the distance from the center of the hexagon. Clockwise from the top, the metrics measure: the median number of visits per pointing in the WFD (N visits/pointing), the fraction of Near Earth Objects of magnitude H ≤ 22 detected (Fraction of NEO), the numbers of galaxies (N Galaxies) and stars (N stars) expected to be recovered, the expected number of supernovae type Ia (SNIa) observed sufficiently well to contribute to cosmological measurements (Lochner et al. 2018), and a collective metric for static probes of the extra-galactic sky (3x2pt FoM, combining Weak Lensing and Galaxy Clustering, Lochner et al. 2018). This set of OpSim runs varies the footprint (which impacts the science throughput for cosmological supernovae as measured by SNIa) and introduces a rolling cadence (which improves SNIa performance as a denser cadence enables better characterizations at relevant, short time scales and the distribution of visits in a more concentrated sky area leads to small efficiency improvements). This figure is discussed in subsection 4.1.
after the end of commissioning during the first months of operation of Rubin Observatory.
• Continue its activities throughout Operations by evaluating reports prepared by the Survey Evaluation Working Group, a project-internal group set up to measure the performance of the survey and scheduler, and make necessary recommendations for adjustments of the survey strategy.
The SCOC consists of ten voting members (12 total) 22 , named by the SAC. Most members are drawn from the science community and are not Rubin Observatory employees. The SCOC membership is expert and diverse, and its deliberations are transparent and inclusive. 23 To ensure continuous communication with the scientific community, the SCOC members also serve as designated liaisons with each SC. This direct line of communication between the SCs and SCOC allows the SCs to be kept informed about the ongoing simulation generation process, and promptly identify gaps in the landscape of simulations or issues with the implementation of metrics that measure their efficiency (see also Section 6).
The SCOC cadence recommendation process is itself split into two phases. The first phase, to be concluded by the end of calendar year 2021, aims to select a cadence family that will capture an overall strategy. By the end of calendar year 2022, the SCOC will fine-tune the selected cadences to come to a final recommendation for the initial cadence to begin execution when operations starts, though this recommendation will be continuously re-evaluated and revised if necessary after the start of operations, as discussed in Section 1. A detailed schedule of the SCOC activities and decision process, including workshops that put the SCOC in direct contact with the scientific community at large, can be found on the SCOC web pages. 24

Cadence Notes
The COSEP and the cadence white papers provided the necessary framework to identify needed simulations, as well as defining how further cadence recommendations should be structured to be maximally useful in the optimization process. To receive formal feedback from the SCs and other stakeholders about this new generation of simulated surveys, and with a more significant adoption of the MAF software by the community, the SCOC issued one further call 25 requesting community input, in the form of Cadence Notes, to explore and evaluate survey strategy options presented in the PSTN-051. Seven specific questions (listed in full in subsection A.3) were posed in the Cadence Notes.
Following the SAC recommendations, in this stage of the process the Rubin LSST survey strategy team was 22 See https://www.lsst.org/content/ charge-survey-cadence-optimization-committee-scoc 23 For the SCOC charge, membership, meeting minutes, recommendations and other documents, please see https://ls.st/scoc 24 https://www.lsst.org/sites/default/files/SCOC%20Handout.pdf 25 https://ls.st/cadencenotes in close communication with the SCs and the scientists working on Cadence Notes, offering, for example, regular office hours. The SCs self organized to collaboratively work notes and share expertise on the use of the MAF software. 26 The community submitted 39 Cadence Notes. 27 These Notes are under review by the SCOC and they represent a significant additional source of information for the SCOC in its deliberations towards the "phase 1" recommendation. The definition of the initial Rubin LSST observing strategy is yet to be finalized at the time of this writing. The remaining phases of the survey optimization process will be described in detail in the closing paper of this Focus Issue.

LESSONS LEARNED SO FAR
Over the last six years of intensive community involvement in optimizing LSST's observing strategy, many initial lessons have been learned. One key lesson has been that effective involvement of the broad community in LSST's survey strategy requires significant coordination across all elements of the Rubin ecosystem, including the construction Project, SAC, and SCs.
A multi-step process was required to (i) engage the most diverse possible user community, (ii) to iterate on initial community analysis and construction Project progress, and (iii) to be agile in response to evolving scientific opportunities. This process began with preconstruction community engagement on the choice of LSST DDFs. Since then, survey input from the community has been solicited in three phases: • through the creation of a single collaborative document, the COSEP, employing, for the first time, community analysis through the MAF and OpSim ; • through the submission of White Papers broadly including new cadence ideas and considerations; • and finally through the submission of Cadence Notes to specifically review a comprehensive set of OpSim runs, and fine-tune survey parameters that remain open to optimization. Table 1 summarizes the key differences between these three phases. We highlight five aspects of our community engagement process that have been important to its success to date: 1. Simulated Surveys and Open Analysis Software: -The release of open software that supports direct interaction with the LSST simulations (OpSim and MAF , see Section 2) was critical to enable contributions from the community. Rather than simply soliciting suggestions on preferred strategies, which would lead to a large set of options without a clear path to optimization, this software framework allowed the users to interact with the simulated survey and see the impact of cadence changes on a science case in conjunction with astronomical and technical constraints, as well as the impact that the same changes have on other science cases. The preparation and release of these software packages were key steps in the process of involving the community in survey strategy decisions.
The provision of software environments and support for using them was essential for including a wide range of astronomers in OpSim analysis and MAF -metrics development. In the White Paper phase, Docker images (Merkel 2014) were provided so that astronomers did not need to install and maintain the full set of interdependent software packages required to use MAF . In later phases, virtual analysis platforms such as the SciServer 28 and NOIRLab's Astro Data Lab 29 provided accounts to scientists working on Rubin optimization. The OpSim databases -which are ∼ 1 GB each -were made available on these platform alongside software environments, example codes, and compute resources for processing. Use of these platforms lowered the barrier to participation and enabled collaboration among scientists.
2. Leverage SC expertise -Developing, maintaining, releasing and supporting software and training the community in its usage are expensive, time-consuming activities. Future surveys that wish to engage the community in the survey design process as Rubin did should scope resources specifically for these dissemination and support activities. With limited resources available to the Rubin LSST survey strategy team, the SC environment became critical for sharing knowledge and knowhow and bringing this software to a large community of users by employing a sort of self-arranged "train-thetrainer" model (Pearce et al. 2012).
In order to respond to the opportunity to contribute to the survey design, Rubin Observatory, the SCs, and the LSST Corporation 30 (a non profit engaging in fundraising to support the scientific community in prepa- Optimizations in the rate of observing over time to better balance the time coverage of the minisurveys over the survey lifetime, ability to strongly prioritize observing closer to the meridian. Modularized survey algorithms to allow for more flexible reward calculations (beyond just slew-time and target maps), leading to more flexible observing strategies. Evolutions of the scheduler algorithm between simulation releases 1.5 and 1.7.1 allowed improved rolling cadence simulations, where visits are distributed on variable timescales in regions of sky over the 10-year survey. Table 1. Community contributions to Rubin LSST survey strategy design ration for LSST) have organized a number of workshops, hackathons, and meetings to enable knowledge transfer, exchange of expert opinions, and communication with the Observatory. The Rubin survey strategy team has participated in many of these gatherings making suggestions on metrics and metric implementation. Some of these activities were hosted within Rubin meetings (e.g., at the annual Project Community Workshop, the all-hands Rubin meeting), others within SC meetings, others yet in purposely organized events (e.g., the Heising-Simons Foundation sponsored a Cadence Hackathon supporting the participation of members of all SCs with travel grants and the development of Cadence White Papers with seed grants 31 ). The SCs particularly acknowledge the support of the LSST Corporation in securing and directing private funds toward activities that supported the recommendations presented to Rubin.
The SCs have been leveraging their internal network structure to coordinate the presentation of different scientific cases within and across SCs. Being in close communication with Rubin allowed the SCs to be central in this process and contribute scientific expertise and insight (see also Section 3 and Section 5). The majority of the 2018 White Papers and 2021 Cadence Notes were created within or in connection with one or more SCs (Figure 4). However, this exercise stretched the organizational capabilities of an un-funded organization that operates largely on a volunteer basis (Bianco et al. 2019).
Some SCs converged on a single document to present their recommendations with one voice (e.g., Dark Energy SC -DESC-, Solar System SC -SSSC-, Strong Lensing SC -SLSC-). Other SCs (e.g., the Transient and Variable Stars SC -TVS-, that covers topics related to Time Domain Astronomy in both the Galactic and Extragalactic environments, Stars, Milky Way, and Local Volume SC -SMWLV-, AGN SC) have responded to the calls for contributions with multiple papers and notes. In practice, the ability to present a single response, which requires internal reconciliation of competing science cases, has depended on the breadth of the science encapsulated within a single SC, as those with a greater diversity of science cases require more substantial management effort to enable that reconciliation. Coordination between SCs with overlapping interests also required additional work: for example, SMWLV and TVS collaborated closely on Galactic science topics and produced additional documents to the SCOC to summa-31 ls.st/hcca rize the SC contributions that were split among multiple notes. The SCs, except for DESC, 32 have no operational budget to support these efforts (see also Section 7 and Bianco et al. 2019).
3. Providing Templates -Expecting a large response, Rubin provided templates for the White Papers and Cadence Notes that ensured the responses would be concise, and limits to the number of pages were provided. This framework was helpful to limit the amount of work required of the committees reviewing the papers, and of the community itself. This Focus Issue, then, provides the opportunity to share the detailed work that underlies these papers and notes.
4. Reliable and Continuous Communication -Ensuring continuous communication with the entire scientific community was also crucial in this process: for example, new OpSim runs need to be distributed broadly and promptly for the community to be able to answer the most relevant questions regarding survey strategy contributions. In addition to providing direct lines of communication with the SCs (subsection 5.1), Rubin kept the community at large updated on discussion about new simulations and metrics or cadencerelated events via the Rubin Community public forum 33 , while metrics contributed by the community and vetted by the Rubin survey strategy team are collected on a dedicated GitHub repository. 34 As the process of cadence optimization moves through its final stages, more simulations will be produced and shared to respond to particular questions and aid in tuning the chosen survey strategy. Even during Rubin operations, additional simulations will be prepared for yearly evaluations of the ongoing survey and to determine best choices of survey strategy for the remainder of the LSST.
5. Aiming For Broad Community Input -Rubin's community approach to survey optimization was designed with the goal to enable all potential users to have a voice in LSST survey strategy planning. Scientists from a wide range of institutions participated in the process. The SCs, for example, include members affiliated with research-focused institutions (R1), teachingfocused institutions, Community Colleges, Labs, and virtual institutes. The geographic span of the SCs reaches five continents.
Enabling the publication of the work underlying the cadence recommendations in peer-reviewed journals is 32 DESC efforts on observing strategy are partially supported under DOE Contract DE-AC02-76SF00515 33 See for example https://community.lsst.org/t/july-2019-update/ 3760 34 https://github.com/LSST-nonproject/sims maf contrib an important step to ensure that individuals outside of the Rubin Observatories employees group can get academic recognition for the time they invest in the process. This Focus Issue collects work related to the optimization of the Rubin LSST observing strategy, as produced by the scientific community throughout the multiple phases of the cadence optimization, providing an opportunity to present the work that underlies 2021 Cadence Notes presented to Rubin Observatory, but also including earlier work, such as the 2018 White Papers.

A LOOK AHEAD
Although our community-driven optimization of Rubin's LSST cadence has delivered initial lessons and the science analyses described in this ApJS Focus Issue, challenges remain. Community engagement in optimizing Rubin's WFD survey has increased Rubin LSST science potential and supported community preparation to conduct science, but it has also been a time and labor intensive endeavor. It has taken six years of active community engagement in survey optimization to reach this point of our work. We aim for this Focus Issue to support future survey-driven missions in their aim to maximize survey science through community engagement in survey optimization.
A significant challenge is that any community involvement plan such as this one remains selective unless participation support is provided. The learning curve to master the cadence analysis remained significant through all three engagement phases. In many cases, this work did not fall under the specification of research grants and was done as service, supported by funds with unusual flexibility (e.g., start-ups and sabbaticals), or even performed outside of working hours. It was difficult for most members of the community to dedicate the time required to conduct these activities at a meaningful level, because not everyone is in a position to dedicate substantial unfunded time and effort to this work owing to workload and funding constraints. Very few are able to do so for extended periods of time. Likewise, the work of the committees in the evaluation of survey strategy input from the community is a time consuming activity. This effort was also uncompensated and performed as service to the scientific community. This framework biases the contributions to the cadence and the evaluation committee membership toward certain job profiles and seniority levels, particularly toward well-funded scientists with significant job flexibility, potentially also biasing the domain expertise to well-funded areas of astrophysics.
The closing paper of this Focus Issue will detail the cadence choice planned for the start of Rubin LSST and the process and considerations that led to its finalization. This opening paper has outlined the initial lessons learned from our community engagement to date. This is not the end of the story. The optimization of the observing strategy in an on-going process with the goal of maximizing science through fine-tuning the LSST survey strategy. In the months and years to come, we expect to learn new things about the most effective way to incorporate community input into this survey optimization. Facility: Rubin Observatory. We used the following Python packages: • numpy (Harris et al. 2020) • maplotlib (Hunter 2007) • seaborn (Waskom et al. 2017) To standardize the community cadence contributions, the COSEP (Section 3), White Papers (Section 4), and Cadence Notes (Section 5) templates included targeted questions aimed at identifying trade-offs and constraints for every science case and strategy suggestion offered by the community. A complete list of these questions follows here.
A.1. Full list of questions included in the COSEP 1. Does the science case place any constraints on the tradeoff between the sky coverage and coadded depth? For example, should the sky coverage be maximized (to ∼ 30, 000 deg 2 , as e.g., in Pan-STARRS) or the number of detected galaxies (the current baseline of 18,000 deg 2 )?
2. Does the science case place any constraints on the trade-off between uniformity of sampling and frequency of sampling? For example, a "rolling cadence" can provide enhanced sample rates over a part of the survey or the entire survey for a designated time at the cost of reduced sample rate the rest of the time (while maintaining the nominal total visit counts).

4.
Are there any science drivers that would strongly argue for, or against, further changes in observing time allocation per band (e.g., skewed much more towards the blue or the red side of the spectrum)? If available, please mention specific simulated cadences, and specific metrics, that support your answer.

5.
Are there any science drivers that would strongly argue for, or against, obtained two visits in a pair in the same (or different) filter? Or the benefits or drawbacks of dedicating a portion of each night to obtaining a third (triplet) visit? If available, please mention specific simulated cadences, and specific metrics, that support your answer.
6. Are there any science drivers that would strongly argue for, or against, the rolling cadence scenario? Or for or against varying the season length? Or for or against the AltSched N/S nightly pattern of visits? If available, please mention specific simulated cadences, and specific metrics, that support your answer.
7. Are there any science drivers pushing for or against particular dithering patterns (either rotational dithers or translational dithers?) If available, please mention specific simulated cadences, and specific metrics, that support your answer.

B. GLOSSARY OF ACRONYMS AND ABBREVIATIONS
A table of acronyms and shorthands used in this paper and in general in LSST cadence-related work follows, to enhance the readability of this work and works within this Focus Issue, more definitions are available on the Rubin website. 35 Table 3.