Connecting the Follow-up Chain: TOMs and Broker Requirements Rachel - - PowerPoint PPT Presentation

Connecting the Follow-up Chain:

TOMs and Broker Requirements

Rachel Street, Federica Bianco TVS Co-chairs Melissa Graham TVS member and LSST liaison R.A. Street, M. Bowman, E.S. Saunders, T. Borosona, 2018 https://arxiv.org/pdf/1806.09557.pdf

Astronomy’s Discovery Chain

Target & Observation Manager Systems Discovery engines Broker services Observing facilities

TOMs everything we know about the LSST changing sky is in the LSST alert streams, but it is a responsiblity of the downstream user to identify the targets of interest E.Bellm, LSST 08/16/2018 Software-enabled observing programs

Target & Observation Manager Systems Broker services

TOMs Software-enabled observing programs

Programmatically requesting observations enables a lot of science, especially:

• Rapid reaction science
• Programs with a lot of targets
• Programs with lots of observations

But at high data rates, keeping track of targets,

• bservations and data can be a big workload

TOMs

• Storing target and observation information in a database offers many advantages
• Can be combined with DB-driven software tools to perform common tasks
• Powerful user interfaces

TOMs: Target and Observation Manager systems

…but modern surveys issue alerts of new discoveries within minutes. We need to harvest targets of interest easily and often and quickly

TOMs Linking TOMs and Brokers

To handle the data rate, TOMs need to programmatically interact with brokers …probably several brokers, ideally with a common language Once targets are harvested, astronomer’s own system/team can decide what to observe

• automate alert ingestion for desired streams, ingesting multiple brokers+catalogs
• automate target selection and follow up trigger based on user priorities
• collect data from follow-up and organize it for further analysis and follow-up
• enables selective data sharing

Problems that need solving that TOMs can solve

TOMs Examples of TOM Systems

Supernova Exchange

Credit: Global Supernova Project

Many TOM systems share common functions, regardless of their science focus…

TOMs

Microlensing TOM

Credit: RoboNet Microlensing Team

…yet many aspects are

• bviously tailored

to the scientific goals of the project

Examples of TOM Systems

TOMs

NEO Exchange

Credit: LCO Solar System Team

Building a TOM is a lot of work though

Examples of TOM Systems

TOMs The TOM Toolkit Project

• A general-purpose software toolkit
• Enable astronomers to easily build TOM systems that they

can customize to suit their needs

• Provide a professionally-developed codebase that will ensure

that the systems are capable of scaling to future programs. Software under development now by LCO’s in-house software engineers First version expected ~mid-2019 Alpha testers will be needed!

TOMs The TOM Toolkit Project

TOMs The TOM Toolkit Project

e.g.Transient Name Server, Astronomers Telegrams, the Minor Planet Center… LCO Network. Gemini, sending requests and receving status updates

TOMs The TOM Toolkit Project

Broker Survey

TVS Broker Requirements Task Force

Chair: R.A. Street Spokesperson: Markus Rabus Francisco Förster Buron, Suvi Gezari, Melissa Graham, Ashish Mahabal, Gautham Narayan, Keivan Stassun, Paula Szkody, Stephen Smartt, Ken Smith

Stimulate scientists from all fields of astronomy to think through how they will extract targets of interest from LSST, what information + data products they need at each stage, timescales of delivery and modes of interaction with brokers. Derive and document Scientific and Functional & Performance requirements

https://lsst-tvssc.github.io/broker_task_force_work_plan.html

Broker Survey Broker Requirements Surveys

User Requirements Survey Nature of the research Target selection criteria Mechanism to receive alerts Information content of alerts Cross-matched catalogs/other surveys Interface requirements:

• Time to access alerts
• Programmatic vs. human-interactive
• Search functions
• Search metrics/criteria

17 questions on: Broker Developer’s Survey 19 questions on: Time to readiness Limitations on users Software requirements for users Data provided Functionality of interface Time to access alerts Search capabilities of interface Interactions with other brokers Foreseen technical challenges Classification functions

(allow more descriptive answers)

Broker’s Clients Survey Broker Survey Responses

Surveys composed as Google Forms Widely circulated within the LSST Science Collaboration and Publicized outside it (open to everyone). Users survey: 61 responses Developers survey: 6 responses Results summarized in an informal report [living document]: https://github.com/LSST-TVSSC/broker-requirements-survey

Broker’s Clients Survey

User Community Represented

What is the subject area of your primary research?

• Scientific diversity

Broker Clients’ Questionnaire

Broker’s Clients Survey

• Scientific diversity
• Geographical distribution

User Community Represented

Broker Clients’ Questionnaire

Broker’s Clients Survey

• Scientific diversity
• Geographical distribution

Do you have LSST data-access rights?

• Scientific diversity
• Geographical distribution
• Almost all already involved

with LSST User Community Represented

Broker Clients’ Questionnaire

Broker’s Clients Survey Result: Alert Information Content

Will the LSST alert contain enough information for your purposes?

Broker Clients’ Questionnaire

Broker’s Clients Survey

Will the LSST alert contain enough information for your purposes? If not, what do you need?

Result: Alert Information Content Broker Clients’ Questionnaire

Broker’s Clients Survey What catalogs should alerts be cross-matched against?

(Essentially everything!)

+ 20 individual suggestions

Result: Alert Information Content Broker Clients’ Questionnaire

Broker’s Clients Survey

How quickly should alerts be disseminated? [Multiple answers/science drivers allowed] How often would you query a broker to search for new targets?

Result: Speed & frequency of access to alerts Broker Clients’ Questionnaire

Broker’s Clients Survey

32 distinct requests for target selection filters users would like to use to identify targets of interest to them.

Result: Target Selection Criteria Broker Clients’ Questionnaire

Broker’s Clients Survey

How do you prefer to access a broker? [Multiple options allowed]

Result: Interface Functionality Broker Clients’ Questionnaire

Broker’s Clients Survey

How do you prefer to access a broker? [Multiple options allowed]

Result: Interface Functionality Broker Clients’ Questionnaire

How would you prefer to receive alerts matching your criteria?

Broker Survey

Questions/answers were more descriptive. The goal was to provide a description of the status and intent of current development teams Note: 1 respondent focused on classifier algorithms rather than full-scale broker service.

Broker Developers’ Questionnaire

Broker Survey

Broker Operational Goals

• At least 3 brokers already have working software (ALeRCE, ANTARES, Lasair)
• Several are already ingesting ZTF alerts as a pathfinder;

Lasair is ingesting Pan-STARSS and ATLAS alerts

• Lasair has a public interface operational now: http://lasair.roe.ac.uk/candlist/
• ANTARES expects to have a limited-feature public interface available Fall 2018
• All aim to be fully operation for the start of the LSST main survey

Broker Developers’ Responses

Broker Survey

What interfaces do you plan to use to communicate with your users?

Broker interface

Broker Developers’ Responses

Broker Survey

All brokers plan to ingest alerts in close to realtime (within minutes) All expect to be publicly accessible, though Lasair may give access to a restricted set of value-added data products

Broker Added-Value Services

Value-Added Products include:

• The classification and anomaly score of objects based on their light curves
• Unique identifier, labels, timestamp, magnitudes, errors, and event classification
• User-customizable output VOEvent stream.
• ANTARES: LSST Alert package, computed light curve features, external catalog associations,

classification and score

• Lasair: Not yet decided.
• ALeRCE: To be finalized, will include: ID, class probabilities and uncertainties, position, time

and last magnitude and error, or light curve for certain classes; known IDs in external databases.

Broker Developers’ Responses

Broker Survey

Selecting targets of interest

In general, brokers expect users to request all targets within a class, rather than to submit custom queries. ANTARES expects to provide limited support for user-defined filters through authenticated interfaces Several expects to issue push notifications for alerts of interest Fully user-customized filters require users to send the query to the broker in advance and receive alerts whenever a new object matches that filter. This requires authenticated user accounts.

Broker Developers’ Responses

Broker Survey

• Complexity of machine learning algorithms and scaling their computational

requirements to match the LSST alert rate

• Cost of computing resources / maintaining funding
• Maintaining the integrity of the database in the event of data processing

mistakes or revisions

• Data provenance & duplication
• Early alert classification

Broker Concerns

Broker Developers’ Responses

Broker Survey User input required!

• Need more classifications/descriptions of astrophysical phenomena
• Need more labeled data for algorithm training
• Need alpha testers of the current services to provide feedback
• User generated election criteria (not prepared to ingest it in real time)

Broker Developers’ Responses

Questions? Comments? Contact Rachel Street on Slack @rachel3834 or Email: rstreet@lco.global