LSST and VOEvent VOEvent Workshop Pasadena, CA April 13-14, 2005 Tim Axelrod University of Arizona
2 Overview of Talk Science drivers Quick look at LSST Data pipeline Characteristics of LSST transients LSST and VOEvent
3 LSST Science Drivers Characterize dark energy through –Weak lensing –Supernovae –Galaxy cluster statistics Explore transient and variable objects Census of solar system objects, especially PHO's 3D structure of the Milky Way
4 A Quick Look at LSST Aperture diameter: 8.4m Effective aperture: 6.7m FOV: 3.5 deg Filters: u(?), g, r, i, z, y Observing mode: pairs of 15 sec exposures, separated by 5 sec slew Single exposure depth: 24.5 Site: Baja or Chile On sky: 2013
5 LSST Optics
LSST Focalplane 3.5 gigapixel, 2 sec readout
Data Acquisition Image Processing Pipeline Detection Pipeline Association Pipeline Image Archive Source Catalog Object Catalog Alerts Deep Detection Pipeline Deep Object Catalog VO Compliant Interface LSST Data Pipeline
Data Pipeline Functions Image Processing Pipeline is responsible for producing –Calibrated science images Astrometric calibration (WCS) Photometric calibration –Subtracted images –Stacked images Detection Pipeline is responsible for producing –The Source Catalog, which contains parameters of all sources found in an image: location, brightness, shape Association Pipeline is responsible for associating sources found at different times and (sometimes) locations, producing –The Object Catalog, which contains parameters of all astronomical objects: lightcurves, colors, proper motions, … Object Classifier, design TBD, is responsible for periodically (re)classifying all objects in the Object Catalog
Spatial Sampling Output of LSST observing simulator Cerro Pachon, 475 days, real weather Weak lensing + supernovae + NEA search
Time Sampling 3 day peak from SN
Time Sampling – cont
Detectable Astrophysical Transients We are limited mostly by –Time sampling –Photometric accuracy (goal is 1%) We will not see (for example) –Low amplitude pulsating WD's (photometry) –Exoplanet transits (photometry and time sampling) –Microlensing caustic crossing events (time sampling) We will see –Many classes of periodic variables with amplitude > 1% –Many microlensing events –Novae –SNe, QSO's, … –As well as “middle of nowhere” transients (eg transients found by DLS)
LSST and VOEvent LSST brings up nothing new regarding the “who”, “when”, or “where” aspects of VOEvent Areas of interest: –Making the “what” useful –Limiting “false alarm” rates –Quantifying “importance” (related to false alarm probability?) –Partitioning of responsibility
Classification of Events The LSST data pipeline will attempt to classify variable objects based on –Position in CMD –Lightcurve shape –Motion, and orbital elements, if applicable The classifier will play a key role in identifying “events” –If the object is already in the catalog, an event occurs relative to the object's previous behavior (an event is not simply a change in flux) –Not so useful for new objects, but still possible to locate in CMD
How can a customer specify an interesting class of event? An “Event” is more than a change in flux –“Notify me of all Cepheids that change period by more than 5%” –“Notify me of all transients > 5σ with no corresponding catalogued object” –“Notify me of any newly discovered solar system object with a > 15AU and confidence > 0.9” We need a flexible semantics for event filters –SQL query on the object catalog is not quite enough(?) –Need to include temporal logic so that past behavior can be referenced(?)
Transient Rates Astrophysical rates - stars –Roughly 5% of stars are variable at the 1% level or more –A typical LSST image contains roughly 2.5e5 stars –Rate from typical images are 1e7 per night –An exceptional LSST image (LMC, bulge) contains up to 4e6 stars Astrophysical rates – extragalactic supernovae –SN rate about 1 / 200 yr / galaxy –Changing flux from each visible for at least 30 d –A typical LSST (unstacked) image contains roughly 4e5 galaxies –Rate is about 1e5 per night
Transient Rates - cont Noise rates –Every PSF patch is a potential transient location – about 8e8 of these –Each is measured once every 35 sec (2 * 15 sec exposures; 5 sec slew) –Assuming gaussian noise About 3e4 / sec at 3σ About 8 / sec at 5σ (3e5 / night) Rate reduced by significant factor if detection required in each 15 sec exposure separately
Dealing With High Event Rates LSST will detect transients at rate of O(1e5 – 1e6 / night) –No group of humans can look at these individually –No followup facility can look at more than a negligible fraction –We need to filter these by a large factor to make them useful Excluding known variable objects results in the biggest reduction – but still leaves large noise rate Noise rates can be reduced by simply increasing the detection threshold – but at the cost of missing real information We need to carefully consider use cases, and make use of simulations, to find a way forward
VOEvent Processing Architecture LSST Data Pipeline VOEvent DB Event Filter Event Filter Event Filter System Boundary?
Unresolved Issues Who will implement the VOEvent DB into which LSST feeds? What latency is needed in generating VOEvents? How best to incorporate links to extra information into VOEvents – eg object lightcurve or image postage stamp How can we incorporate concepts like “classification change” or “period change” into VOEvent? –This type of event depends on a baseline, which somehow must be part of the data How can we assign “importance” in a quantitative way?