Photometric Science Alerts from Gaia · FOLLOW-UP CALIBRATION SERVERCalibration Server for Gaia Science Alerts Photometric Follow-up access can be fully automatised ... BP/RP spectra:

Photometric Science Alerts from Gaia

Simon Hodgkin

Institute of Astronomy, CambridgeThursday, 12 September 13

Hodgkin, LSST@Europe, IoA, Sept 11th 2013

Acknowledgements• Giuseppe Altavilla• Vasily Belokurov• Josh Bloom• Ross Burgon

• Nadejda Blagorodnova• Heather Campbell• Gisella Clementini• Michel Dennefeld• Andrew Drake• Gerry Gilmore

• Jorge Fernandez Hernandez• Anna Hourihane• Peter Jonker

• Sergey Koposov

• Floor van Leeuwen

• Goska van Leeuwen• Ashish Mahabal• Francois Mignard• Timo Prusti

• Guy Rixon• Iain Steele• Rachel Street• Yiannis Tsapras• Massimo Turatto• Nic Walton• Sjoert van Velzen• Roy Williams

• Lukasz Wyrzykowski

• Abdullah Yoldas

2

Thursday, 12 September 13


Outline

3

• Gaia’s Observing Strategy• When do we switch on Alerts• Early phases and verification• Alert Publication



What is a Photometric Science Alert?

• appearance of a new source, or a change in flux which suggests we could learn something from prompt follow-up

• this does not include:

• periodic variable stars (better left to the end of the mission)*

• moving objects**

• this could include spectral variation (independent or combined with flux variation)

4

*there are cases where it may be more efficient to do something promptly, e.g. Zucker et al. transients** astrometric microlensing would be an exception



Scanning Law

5

• Two telescopes, one focal plane• Time to cross a CCD: 4.4s• Time to cross the FOV: 45s• Time between FOVs: 106.5m

• Time between scans: 6h• Field revisited every ~30 days• Each object measured ~70x• Densest coverage ~200 epochs

45o

45



Gaia Catalina Sky Survey

PTF LSST

deg2 day-1 ≈ 1230 1200 1000 5000

Avg Cadence ≈ 30 days 14 days 5 days 4 days

Limiting mag 20 (21?) 19.5 21 r=24.7

fsky all sky 0.6 0.2 <0.48

Gaia as a Transient Survey

van Velzen et. al, 2011, Blagordonova6



Photometry per transit

7

1% at G=19 (colours ~10%) <2 mmag precision for G<12



Timeline for Data Flow16h 8h visibility

backlog real time acquisition Gaia

transmission MOC

transmission SOC

0 24one operational dayd-1 d d+1

48

Initial Data Treatment

First Look

Figure courtesy Francois Mignard, updated by LW+STH

Madrid, Spain

G<16, Astrometry (50 mas)

Astrometry (100 μas)

ASAScience Alerts (Cambridge)

G<20, Astrometry (50 mas)



Year 1

9

internal verification



• Soszyński et al. 2012, AcA....62..219S

• Covers 4x1.4 square degrees

• Monitoring V (10s),I (100s) since 2010

• 1.6 million stars• 6789 variable stars: 132

Cepheids, 686 RR Lyr, 2819 long-period, 1377 EBs

• 2 SN 1a’s, and 9 SN candidates

Ecliptic Pole ScanningGSEP: OGLE observations

10

Image courtesy of G Pojmanski, ASAS



Year 1

11

internal verification

test against external transient surveys

exact date is a bit fuzzy



Year 1 coverage

12

N. Blagorodnova


Hodgkin, LSST@Europe, IoA, Sept 11th 2013 13

Year 1 coverage

N. Blagorodnova



Size of Verification Plan?• For a training set:

100s per broad class seems to be a reasonable estimate

• This is of order 500 ‘follow-ups’

• Which is >50 nights

• Try to do the bulk on 2m class telescopes

14

– 28 –

VarStar−Periodic

VarStar−Misc

VarStar−CV

SN/Nova

AGN−cnSN−TDE

A−cnSN−T SN/N V−CV V−M V−P

Pred

icte

d C

lass

True Class

N= 1117 456 29 15 336

0.01

0.01

0.97

0.02

0.01

0.93

0.04

0.31

0.41

0.03

0.24

0.2

0.07

0.73

0.9

0.02

0.08

Fig. 12.— Confusion matrix for robotclass random forest classification. Classes are aligned so that

entries along the diagonal corresponds to correct classification. Probabilities are normalized to sum

to unity for each column. Recovery rates are !90%, with very high purity, for the three dominant

classes. Classification accuracy su!ers for the two classes with small amounts of data (note: class

size is written along the bottom of the figure).

forest is that its component trees are de-correlated by sub-selecting a small random number of

features as splitting candidates in each non-terminal node of the tree. As a result, the average

of the de-correlated trees has highly decreased variance over each single tree. To handle missing

feature values—which arise due to incompleteness in the context features—we use the missForest

imputation method of Stekhoven & Buhlmann (2011), which estimates the value of each missing

feature via an iterative nonparametric approach to minimize imputation error.

For the PTF Type classification problem, we have 1573 Transient and 380VarStar sources15.

Using features derived at the time of discovery, we obtain a 3.8% overall error rate (all error rates

stated are found using 10-fold cross validation). For the 1422 sources with SDSS coverage, the error

rate is 1.7%, while for the other 531 sources with no SDSS coverage the error rate jumps to 9.4%.

15There is some ambiguity in the initial typing scheme in the boundary between VarStar and Transient: cata-

clysmic variables (CVs), for instance, could be considered in either category. However, for definiteness, we put CVs

in the VarStar category.

Automating Discovery and Classification of Transients and Variable Stars in the Synoptic Survey EraJ. S. Bloom1, J. W. Richards1,2, P. E. Nugent3,1, R. M. Quimby4, M. M. Kasliwal4, D. L. Starr1, D. Poznanski1,3, E. O. Ofek4, S. B. Cenko1, N. R. Butler1, S. R. Kulkarni4, A. Gal-Yam5, N. Law6

http://arxiv.org/abs/1106.5491v1





Pragmatic Approach• Understanding the Gaia data stream (especially

variables) early on leads to:

• decisions about depth

• threshold for triggers

• classification training (on ‘bright’ samples), then extension to fainter sources

• population studies

• Plan to complete verification within the first year of operations

15



Do we really need to pre-filter Alerts?

• Why not release the whole data stream?

• Resource limited - Alerts where not originally envisaged for Gaia.

• Significant delay in processing chain ~1-2 days (cf 60s for LSST).

• Can still analyse the first data release and come back with a science case

16



SNa discovery rates

http://www.cbat.eps.harvard.edu/lists/Supernovae.html

See also: Supernovae and Gaia, Altavilla et al. 2011, 2012Ap&SS.tmp...66A

GAIA-C5-TN-IOA-SHO-001-00In this note, we revisit the simulations of Belokurov and Evans (2003), and make adjustments to account for the current mission parameters (the biggest change from the published numbers arising from the scanning law which roughly halves the number of observations of each part of the sky).

6000 to G=19


http://adsabs.harvard.edu/abs/2012Ap%26SS.tmp...66A

http://adsabs.harvard.edu/abs/2012Ap%26SS.tmp...66A


BP/RP SN Spectral ClassificationPoster by N. Blagorodnova



How we are organising/communicating?

• Annual meetings (here, Bologna, Paris) • Coordinated spectroscopic verification: ~80

scientists, ~50 nights INT, ~100s hours Liverpool Telescope (ToO)

• Photometric follow-up: 47 x (7cm-2m) telescopes listed, 13 observatories are already doing tests.

• Educational and amateur involvement• Provide useful tools (e.g. calibration server)

19



5.6475•104 5.6480•104 5.6485•104 5.6490•104 5.6495•104 5.6500•104 5.6505•10MJD

16.5

16.0

15.5

15.0

14.5

14.0

Flux

V : APASS -0.6r : APASS -0.4i : APASS -0.2B : APASS

Calibration ServerFOLLOW-UP CALIBRATION SERVERfor Gaia Science Alerts Photometric Follow-up

access can be fully automatised

software developed by Sergey Koposov, IoA



Calibration Server

5.6475•104 5.6480•104 5.6485•104 5.6490•104 5.6495•104 5.6500•104 5.6505•10MJD

16.5

16.0

15.5

15.0

14.5

14.0

Flux

V : APASS -0.6r : APASS -0.4i : APASS -0.2B : APASS

Creation of Sergey Koposov & Lukasz Wyrzykowski



Alert Publication•Publication of Alerts to the entire community: no proprietary data.

•Web-based, email-based

•VOEvent - machine-readable format

• Skyalert.org - can host both alerts and follow-up data

22



We will publish• coordinates• magnitudes, lightcurves• spectra, colours• proper motions, parallaxes (when

available)• astrophysical parameters (when available)• features (random forest classifier)• classifier probabilities• xmatch results

23

Gaia



• Known variables typically excluded from a transient survey.

• We will monitor a pre-selected list of interesting objects.

• Flexible - add objects to list of alerts during the mission, e.g. external triggers.

• Planning for AO timed with Gaia Launch.

• Sensible limits for numbers of targets.

Watch List



Summary

• The alert stream is non-proprietary and will be (some of) the first data from Gaia Q3 2014.

• A lot of internal verification can be done early in 2014 using the EPSL data.

• We have planned an extensive follow-up programme, for classifying large numbers of transients.

25

http://www.ast.cam.ac.uk/ioa/wikis/gsawgwikiThursday, 12 September 13

http://www.ast.cam.ac.uk/ioa/wikis/gsawgwiki

http://www.ast.cam.ac.uk/ioa/wikis/gsawgwiki


Astrometry per transit

• OGA1: 50 milli arcsec (IDT)• OGA2: 100 micro arcsec

(24hr later)26



BP/RP spectra: classification

• two low-res fused-silica prisms

• BP 330-680nm @ 4-32 nm/pixel

• RP 640-1000nm @ 7-15 nm/pixel

27



Operational schemeNew

Transits

Candidate Alerts

GaiaClassifier VO

Historic data

• SDSS• 2MASS• NED• ASAS• OGLE• HST archives• etc etc

The most recent calibrations are used to calibrate new data.

New observations are compared against historic ones

Lightcurves, spectra and other Gaia data are used for initial classification

Fine-tuning the classification using archived other data

Alerts released to the community for the follow-up.

cal

Cross-match

AnomalyDetector

photometry astrometry

spectroscopy

WSDB

Detection is simple to start with, but plan to include matched filter algorithms



Detection: the simplest algorithmH

isto

ry

Mean

threshold

His

tory

False alert

29


Documents

Photometric Science Alerts from Gaia · FOLLOW-UP CALIBRATION SERVERCalibration Server for Gaia Science Alerts Photometric Follow-up access can be fully automatised ... BP/RP spectra: