Photometric Science Alerts From Gaia


Gaia is a European Space Agency (ESA) astrometry space mission, and a successor to the ESA Hipparcos mission. The main goal of the Gaia mission is to collect high-precision astrometric data (i.e. positions, parallaxes, and proper motions) for the brightest one billion objects in the sky. This data, complemented with G band, multi-epoch photometric and low resolution (lowers) spectroscopic data collected from the same observing platform, will allow astronomers to reconstruct the formation history, structure, and evolution of the Galaxy.

In addition, the Gaia satellite is an excellent transient discovery instrument, covering the whole sky (including the Galactic plane) for the next 5 years, at high spatial resolution (50 to 100 mas, similar to the Hubble space telescope (HST)) with precise photometry (1\(\%\) at G=19) and milliarcsecond astrometry (down to \(\sim\)20mag). Thus, Gaia provides a unique opportunity for the discovery of large numbers of transient and anomalous events, e.g. supernovae, black hole binaries and tidal disruption events. We discuss the validation of the alerts stream for the first six months of the Gaia observations, in particular noting how a significant ground based campaign involving photometric and spectroscopic followup of early Gaia alerts is now in place. We discuss the validation approach, and highlight in more detail the specific case of Type Ia supernova (SNe Ia) to be discovered by Gaia. The intense initial ground based validation campaign will ensure that the Gaia alerts stream for the remainder of the Gaia mission, are well classified.

What is a Photometric Science Alert?

A photometric science alert is the appearance of a new source, or a change in flux, which suggests we could learn something from prompt ground-based follow-up. This does not include: periodic variable stars (these sources may be better left to the end of the mission) and moving objects (however, astrometric microlensing would be an exception). The science alerts will be made public, within one to two days of Gaia detection, most of this time is due to downloading the data from the satellite.

Potential Triggers

Potential triggers for the the Gaia science alerts are objects of scientific interest which would benefit from fast ground based follow-up, as just discussed. Some examples of sources which maybe potential triggers include supernovae, super-luminous supernovae, tidal disruption events, cataclysmic variables, outbursts and eclipses from young stellar objects, X-ray binaries, microlensing events and other theoretical or unexpected phenomena. Figure \ref{fig:triggers} shows some of these potential triggers and the area of pars space they occupy for their brightness as a function of duration.

This shows the amplitude and duration of a range of potential triggers for the Gaia science alerts. \label{fig:triggers}

Gaia as a Transient Search Machine

Gaia is comparable to other transient search machines, such as the Catalina Sky Survey and the Palomar Transient Factory, as shown in Table \ref{tab:transient_machine}, which covers similar areas each day and similar limiting magnitude. The disadvantage of the Gaia survey is that the average cadence is only \(\sim\)30days whereas transient surveys usually have a cadence of approximately 3 to 5 days. However, there is also a shorter cadence of 106.5 mins from the two mirrors in the satellite, also sometimes a 253.5 mins cadence, and sometimes 3 or more observations are thus obtained (when close to the 45 degrees ecliptic latitude zones for example). This 106.5 mins cadence is a huge advantage and means that changes in brightness should be detected quickly. Also, Gaia will cover the whole sky (including the Galactic plane), which is a significant survey area increase over other transient searches. The Gaia transient alerts will also have high spatial resolution with precise photometry (1\(\%\) at G=19) and milliarcsecond astrometry (down to \(\sim\)20mag), lowres spectra for all objects brighter than \(\sim\)19mag and colours for fainter objects (see \citet{Jordi:2010} for details of the photometry and lowres spectra).

Patient Gaia Catalina Sky Survey Palomar Transient Factory
deg2 day-1 \(\sim\)1230 1500 1000
Avg Cadence \(\sim\)30 days 14 days 5 days
Limiting mag 20 (21?) 19.5 21
fsky all sky 0.6 0.2

Time line

The Gaia satellite was launched on the 19th December 2013, and has now successfully been placed into orbit around the second Lagrange point. Over the next few months the telescope will undergo system shake-down and ESA commissioning (Figure \ref{fig:timeline}). It is planned that in June the Gaia satellite will spend a month scanning the Ecliptic Poles internally verifying the data, and learning how to identify large amplitude variable stars (potential contaminants of the Gaia Science Alerts stream).

\label{fig:timeline} Current timeline for Gaia operations and data accumulation.

Then in July Gaia will switch to nominal scanning and history of the whole sky will begin to be accumulated. In Figure \ref{fig:scanning} we show the expected coverage of Gaia by the end of July and then the end of September 2014. This will give some history of each patch of sky in the Gaia passbands and allow detection of transient objects. We propose to begin Gaia Alerts Spectroscopic Follow-up in the last weeks of August and the first week of September.

Scanning law

The Gaia satellite consist of two telescopes, which are projected onto one focal plane. The time between the two fields of view being observed is 106.5 mins and then the time between subsequent scans is 6 hours. After these initial observations the field will be revisited every \(\sim\)10-30 days. Over the full mission each patch of sky will be measured, on average, approximately 70 times. The densest coverage is at 45 degrees to the ecliptic plane and this region is covered with approximately 200 epochs.

\label{fig:scanning} By 30 days 11.6\(\%\) of the sky has been observed at least 3 times by Gaia. By 90 days, 52.03\(\%\) of the sky has been observed at least 3 times by Gaia.

SNa discovery rates

Simulations, Belokurov (2003) and updated by Altavilla et al. (2012), predict Gaia will see \(\sim\)6000 SNe down to G=19 (3/day), and twice this to G=20. One SN per day will be brighter than 18th magnitude (see Fig \ref{fig:SN_rate}). For cataclysmic variables (CVs) the rate will be approximately similar, and Breedt, (priv. comm) predict Gaia should find 1000 new CVs. Blagorodnova (in prep) predict that Gaia will find of order 20 Tidal disruption event’s (TDE’s) per year. Young stellar objects outbursts will be less common and Gaia will probably only find a few per year.

\label{fig:SN_rate} Predicted SN detections with Gaia as a function of G-band magnitude.

Alert Publication

Alerts are expected to be discovered and published to the world within \(\sim\)24-48 hours of observation by the satellite. The Alert Stream will go live once Gaia has mapped at least 10\(\%\) of the sky, a minimum of 3 times, which takes approximately one month (see Fig \ref{fig:scanning}). Once the Gaia alert stream is fully operational all alerts will be made publicly available, and thus accessible for use by the community in their dedicated followup campaigns (see Section \ref{followup}). During the commissioning, initialisation and early operations phase of Gaia (January - August 2014) - there will be systematic validation of the Gaia alerts, whereby the operational system will be assessed before going ‘Live’. The science alerts will be available to the community in web-based and email-based formats and will be produced in Virtual Observatory Event (VOEvent) - machine-readable format.

Each Alert package will consist of: coordinates, magnitudes, light curves, spectra, colours, proper motions, parallaxes (when available), astrophysical parameters (pars) (when available), features (random forest classifier see Section \ref{class}), classifier probabilities, cross match results.


\label{class} Gaia is predicted to detect 44 million transits per day,which is \(\sim\)150 - 800 GByte/day of data. Within this huge volume of data we expect 100s -1000s of potential interesting astrophysical triggers per day (real variables/moving objects). This precludes visual classification of a rich data stream and thus automated methods which are fast, repeatable and tuneable are essential. The Gaia alerts classification pipeline uses random forest classification. The random forest will use all the information available, and its features will include: light curve photometry (gradient, amplitude, historic rms, magnitude, signal-to-noise ratio, transit rms), lowres spectra (flux v lambda, colours, spectral shape coefficients, spectral type), auxiliary information (neighbour star, shape pars, motion pars, coords, crowding, calibration offset, correlations, QC pars) and crossmatch environment (near known star mags, near known variable class, near galaxy, near galaxy redshift and circumnuclear).

To build up a sufficient sample of classification labels in order to train the random forest classifier (e.g. \citet{Ofek_Cenko_Butler_et_al__2012}) we aim to observe \(\sim\)500s homogenous high-quality spectra in the first year of the mission, spread across each broad class of transient phenomena (active galactic nuclei, core collapse SN, TDE, SN, Novae, CV and variable stars).

The light curve classification utilises the flux gradient of the transient object. The Gaia observations with 106.5 mins cadence are used to indicate the type of object. The lowers (BP/RP) spectra provide far more information to aid classification \citet{Blagorodnova_in_prep_2014} and provide robust class for most objects, at \(>\)19mag, when the classifier is fully trained on representative data. In addition, the transient object will be cross matched with archival catalogues, for example, Sloan Digital Sky Survey (SDSS), Two Micron All Sky Survey (2MASS), HST and Visible and Infrared Survey Telescope for Astronomy (VISTA). This will help remove known variable star contaminates and provide environmental information for the transient events, e.g. is there a host galaxy associated with the source and if so what is the type and magnitude.

Follow up

\label{followup} We are also co-ordinating a large program of photometric follow-up to improve the light curve sampling of Gaia transients. 47 x (7cm-2m) telescopes are listed as currently active ( and 13 observatories are already doing tests ( All make use of our photometric calibration server (a tool developed to maximise the usefulness of the photometric followup data) to place the disparate data onto the same system (Wyrzykowski et al. 2013 \(ATEL\#5245\)). Additionally, Las Cumbres Observatory Global Telescope Network (LCOGT) are expected to play a key role in the follow-up especially of \(\mu\)lensing and young star transients. We point out the strong synergies with external facilities operating at different wavelengths. We will be able to confirm and characterise e.g. Low Frequency Radio Array (LOFAR) transients, and we may also trigger prompt SWIFT follow-up for particularly interesting events.

There is also a large educational (mostly utilising the Faulkes telescopes) and amateur involvement planned in the followup of these transient events, to assist in compiling light curves and increase the public evolvement and interest.

We need a large sample of well-exposed (S/N\(\sim\)20\(-\)50), medium-dispersion (R\(\sim\)500\(-\)1000) spectra, over a wide range of classes and magnitudes, to build classification training sets, in order for our (Random Forest) machine learning algorithms (discussed in Section \ref{class}) to perform well for the Gaia spectra for the remainder of the mission. Therefore we aim to obtain 1.5-4m telescope time to build this training set. It is important to invest time at the beginning of the Gaia mission to understand and characterise the transients that will be discovered with Gaia, so that we can optimise the process, and ensure that the rest of the mission is as productive as possible. We also intend to archive and release our spectroscopic classifications promptly after processing each night’s observing.


The alert stream is non-proprietary and will be (some of) the first data from Gaia Summer 2014. We have planned an extensive follow-up program for classifying large numbers of transients: e.g. 10,000 SNe Ia over the whole sky. The alerts will be published one to two days after the event was initially detected (most of this time is due to the time taken for the data to be down linked from the satellite and processed). The alerts will be preliminarily classified using random forest classifiers based on the Gaia photometry and lowres spectra with additional cross match information from existing surveys. These classifications should improve after the first few months of ground based followup and retraining of the Bayesian classifiers. The alerts will be published in the VO format. For more information visit:


Material used in this work has been provided by the Coordination Unit 5 (CU5) of the Gaia Data Processing and Analysis Consortium (DPAC). They are gratefully acknowledged for their contribution.

HCC, STH, GG, NAW, LW are members of the Gaia Data Processing and Analysis Consortium (DPAC) and this work has been supported by the UK Space Agency. NB has been supported by the The Gaia Research for European Astronomy Training (GREAT-ITN) network, funded through the European Union Seventh Framework Programme ([FP7/2007-2013] under grant agreement n\(\,^{\circ}\) 264895.


  1. C. Jordi, M. Gebran, J. M. Carrasco, J. de Bruijne, H. Voss, C. Fabricius, J. Knude, A. Vallenari, R. Kohley, A. Mora. Gaia broad band photometry. Astronomy & Astrophysics 523, A48 EDP Sciences, 2010. Link

  2. V. Belokurov. Supernovae with super-Hipparcos. 341, 569-576 Springer-Verlag, 2003. Link

  3. Giuseppe Altavilla, Maria Teresa Botticella, Enrico Cappellaro, Massimo Turatto. Supernovae and Gaia. Astrophysics and Space Science 341, 163-178 Springer-Verlag, 2012. Link

  4. N. Blagorodnova. BANGST! Bayesian Analysis for Nearby Gaia Supernovae Transients. (in prep). Link

  5. J. S. Bloom, J. W. Richards, P. E. Nugent, R. M. Quimby, M. M. Kasliwal, D. L. Starr, D. Poznanski, E. O. Ofek, S. B. Cenko, N. R. Butler, et al.. Automating Discovery and Classification of Transients and Variable Stars in the Synoptic Survey Era. Publications of the Astronomical Society of the Pacific 124, 1175-1196 The University of Chicago Press, 2012. Link

[Someone else is editing this]

You are editing this file