Imaging Challenges & Pilot Survey

Steven T. Myers (NRAO)

1 VLASS – CALIM Oct 2016 Why a new VLA Sky Survey? A New VLA deserves a New Sky Survey https://science.nrao.edu/science/surveys/vlass • Expanded (bandwidth, time/freq resolution) capabilities of JVLA – 24x instanteous continuum bandwidth 84MHz  2048MHz • 20 years since previous pioneering VLA sky surveys – NVSS 45” 1.4GHz (84MHz) 450µJy/bm (2932hrs) 30Kdeg2 2Msrcs – FIRST 5” 1.4GHz (84MHz) 150µJy/bm (3200hrs) 10.6Kdeg2 95Ksrcs • Preparing the way for a new generation of surveys – ASKAP / MeerKAT / Apertif – SKA phase 1 – new O/IR surveys (i/zPTF, PanSTARRS,DES,LSST), X-ray (eROSITA)

2 VLASS – CALIM Oct 2016 VLASS Survey Science Group A Survey For the Whole Community Proposal Authors: Table 8: VLASS Proposal Contributors, Including VLASS White Paper Authors F. Abdalla, Jose Afonso, A. Amara, David Bacon, Julie Banfield, Tim Bastian, , Stefi A. Baum, Tony Beasley, Rainer Beck, Robert Becker, Michael Bell, Edo Berger, Rob Beswick, Sanjay Bhat- nagar, Mark Birkinshaw, V. Boehm, Geoff Bower, Niel W. Brandt, A. Brazier, , Michael Brotherton, Alex Brown, Michael L. Brown, Shea Brown, Ian Browne, Gianfranco Brunetti, Sarah Burke Spolaor, Ettore Carretti, Caitlin Casey, Sayan Chakraborti, Claire J. Chandler, Shami Chatterjee, Tracy Clarke, Julia Comerford, Jim Cordes, Bill Cotton, Fronefield Crawford, Daniele Dallacasa, Constantinos Science Proposal and Demetroullas, Susana E. Deustua, Mark Dickinson, Klaus Dolag, Sean Dougherty, Steve Drake, Alas- survey definion by tair Edge, Torsten Ensslin, Andy Fabian, Xiaohui Fan, Jamie Farnes, Luigina Feretti, Pedro Ferreira, Dale Frail, Bryan Gaensler, Simon Garrington, Joern Geisbuesch, Simona Giacintucci, Adam Ginsburg, Gabriele community Survey Giovannini, Eilat Glikman, Federica Govoni, Keith Grainge, Meghan Gray, Dave Green, Manuel Guedel, Science Group (SSG) Nicole E. Gugliucci, Chris Hales, Gregg Hallinan, Martin Hardcastle, Ian Harrison, Marijke Haverkorn, Martha Haynes, George Heald, Sue Ann Heatherly, Alan Heavens, Joe Helmboldt, C. Heymans, Ian Hey- following AAS 223 wood, Julie Hlavacek-Larrondo, Jackie Hodge, Michael Hogan, Assaf Horesh, C.-L. Hung, Zeljko Ivezic, workshop in Jan 2014 Neal Jackson, Matt Jarvis, B. Joachimi, Atish Kamble, David Kaplan, Namir Kassim, S. Kay, Amy Kim- ball, T.D. Kitching, Roland Kothes, Diederik Kruijssen, Shri Kulkarni, Mark Lacy, Cornelia Lang, Casey Law, Joe Lazio, J.P. Leahy, Jeff Linsky, Xin Liu, Britt Lundgren, R. Maartens, Antonio Mario Magalhaes, Minnie Mao, Sui Ann Mao, Maxim Markevitch, Walter Max-Moerbeck, Ian McGreer, Brian McNamara, Community review and Lance Miller, Elisabeth Mills, Kunal Mooley, Tony Mroczkowski, Eric J. Murphy, Matteo Murgia, Tom approval in March 2015 Muxlow, Steve Myers, Bob Nichol, Shane O’Sullivan, Niels Oppermann, Rachel Osten, Pat Palmer, P. Patel, Wendy Peters, Emil Polisensky, Ed Prather, J. Pritchard, Cormac Purcell, A. Raccanelli, Scott Ran- som, Urvashi Rao, Paul Ray, A. Refregier, Gordon Richards, Anita Richards, C. Riseley, Tim Robishaw, Anish Roshi, Larry Rudnick, Michael Rupen, Helen Russell, Elaine Sadler, M. Santos, Anna Scaife, B.M. SSG is acve – you can Schafer, Richard Schilizzi, Dominic Schnitzeler, Yue Shen, Kartik Sheth, Greg Sivakoff, Lorant Sjouw- join and parcipate erman, Ian Smail, Oleg Smirnov, Vernesa Smolcic, Alicia Soderberg, Dmitry Sokolov, Tim Spuck, Judy Stanley, J.-L. Starck, Jeroen Stil, John Stoke, Michael Strauss, Meng Su, Xiaohui Sun, R. Szepietowski, A.N. Taylor, Russ Taylor, Valentina Vacca, Reinout van Weeren, Tiziana Venturi, Andrew Walsh, Wei- Hao Wang, Dave Westpfahl, Robert Wharton, Rick White, Stephen White, L. Whittaker, Peter Williams, Kathryn Williamson, Tony Willis, Tom Wilson, Maik Wolleben, Nicholas Wrigley, Ashley Zauderer, J. Zuntz

3 Note: MembersVLASS of the – SSGCALIM are listed Oct in bold 2016.

Each working group was led by two co-chairs, with the co-chairs comprising the SSG Governing Council. The SSG Governing Council itself had two co-chairs (Stefi Baum and Eric Murphy). Con- tributions to the WG discussions were enabled through the NRAO Science Forum,13 with material also posted on the NRAO Public Wiki, 14 along with other methods of group communication such as Google Groups, as defined by the co-chairs of the individual WGs. In this way, contributions to the discussion on these WGs was expanded well beyond the original authorship of the WPs. The process by which the VLASS survey definition proceeded from this point is worth docu- menting, as it may serve to guide the development of future surveys. Initially, the three scientific WGs (Galactic, Extragalactic, and Transients/Variability) were asked by the SSG Council co-chairs to specify their “ideal” survey designs, supported by key science goals. A “virtual face-to-face” meeting was then used to assess areas of commonality between the elements of the proposed sur- veys (frequency band, array configuration) and to identify areas that needed further discussion (number of epochs, depth of each epoch, monolithic vs. tiered). At this stage, the focus of the

13https://science.nrao.edu/forums/viewforum.php?f=59 14https://safe.nrao.edu/wiki/bin/view/VLA/VLASS

60 What is the VLA Sky Survey?  Captures a set of snapshots of the radio sky unique in me & “space”  Enables : focused radio, mul-λ, stascal, me domain & polarizaon studies VLASS rms depth (1σ)  S-Band (2 – 4 GHz), B/BnA configuraons, 2MHz channels single epoch, 2.5” res.  Wide Bandwidth – connuum images + spectral cubes 1500MHz – 120 µJy/beam  Full Polarizaon – Improved RM Synthesis Imaging 128MHz – 410 µJy/beam 10MHz – 1.5 mJy/beam  Synopc – 3 epochs, 120µJy/beam per epoch, 32 month cadence  All-sky visible to VLA (δ>-40° ~34kdeg2) – including Galacc plane and bulge  High Time Resoluon 0.45sec  OTFM scanning at ~3.3’/s (~24deg2/hr)  High Angular Resoluon (2.5”)  locate hosts and locaon within hosts Density Total Tier (deg-2) Detecons  Start Sep 2017 – End 2025 All-Sky 290 9,700,000  5520 hr investment over ~7yr (920h/cycle)  <15% impact on PI me 10x FIRST yield, ~5x NVSS

Area Resoluon Rms Time Tier (deg2) (’’,robust) (µJy/bm) (hr) Epochs All-Sky 33,885 (δ > -40°) 2.5 69 5520 3 4 VLASS – CALIM Oct 2016 VLA Sky Survey Science VLASS Science Drivers • Key science from VLASS proposal (SSG): – “Hidden explosions” and other radio transients. – Faraday tomography of the magnetic sky. – AGN and galaxy evolution in concert with new optical/IR surveys. – Peering through our dusty Galaxy. – “Missing physics”, rare objects, opening new discovery space • What defines a sky survey? – Area, Survey Speed (SS) & “depth” – covering an area of sky in given time (e.g. VLASS ~24 deg2/hour)

∑wkΩB (ν k ) Ωtot ΩB  k 2 −2 SS = = =θ θrow ΩB = ΩB ≈ 0.5665θFWHM ∝ν ttot tint ∑wk – see Technical Implementation Plan: k https://www.authorea.com/users/4076/articles/8161

5 VLASS – CALIM Oct 2016 On-the-Fly (OTF) Mosaicking with JVLA Enabling new surveys with old telescopes • Scan telescopes across sky while taking array data (step phase centers) – Lose no data while scanning instead of 3-7s per step in standard mode – Efficient when dwell times on-sky are <25s (VLASS ~5s) • High scanning rates & survey speeds (VLASS: 23.83 deg2/hr) – limited by data rates (25MB/s GO) for fast correlation dump times – >100 deg2/hr possible at L-band (1-2GHz) – VLASS: 23.83 deg2/hr • 3.31’/s scan rate • 7.2’ row sep. ×× × ×

See hps://science.nrao.edu/facilies/vla/docs/manuals/obsguide/modes/mosaicking

6 VLASS – CALIM Oct 2016 OTFM: Considerations Primary Beam Smearing • Antenna motions over a visibility integration time cause the effective primary beam to be smeared along the motion axis. • For VLASS at 3.31’/s and 0.45s integrations, this is 1.49’ which is 14.4% of the 10.34’ primary beam FWHM at 3.95GHz, and 10.6% of the 14’ beam at 3GHz. This leads to a sensitivity loss of <1%.

7 VLASS – CALIM Oct 2016 OTFM: Considerations Mosaic Uniformity • For a given survey speed, one chooses an OTFM row separation which then prescribes a scan rate. The row separation should be chosen to limit non- uniformity. For VLASS this is 7.2’, giving 2%-5% expected non-uniformity.

for θrow=7.2’

3.56GHz 3GHz 3.95GHz

8 VLASS – CALIM Oct 2016 What will the VLASS produce? ~1PB Raw vis 523TB Data & Data Products Publically Available! Images 440TB • Basic Data Products (BDP): NRAO produced & verified Product Timescale Notes Raw Data immediate no proprietary period Calibrated Data 48 h. – 1 week same, served from archive Quick-Look Images 48 h. – 2 weeks (?) connuum only, simple QA Single-Epoch Images & Cubes 6 mos. (12 mos. pol beer quality assurance, includes (2048/128MHz) cubes) polarizaon, coarse cubes Single-Epoch Catalog w/SEI more object parameters Cumulave Images & Cubes 12 mos. (16 mos. pol produced aer each epoch, (2048/128/10MHz) cubes) increased depth aer first Cumulave Catalog w/CI more detailed • Enhanced Data Products (EDP): community provided – opportunity for NSF funding (MSIP), international partners (e.g. AU,CA,ZA)

9 VLASS – CALIM Oct 2016 Enhanced Data Products & Services Community led effort

• Transient Object Catalogs, Marshals, & Alerts Opportunities for • Multi-Wavelength Catalogs for VLASS sources seeking funding for • Rotation Measure Images and Catalogs EDP from NSF (MSIP) and international • Light Curves (IQU) sources (AU,CA,ZA) • A hosted VLASS Archive with Image and Catalog Service  e.g., as currently available by IPAC/IRSA allowing for VLASS data to be integrated with Spitzer/Planck/WISE/Euclid/ etc… • VLITE (NRL): – commensal 350MHz • commensal optical surveys? • many possibilities …

10 VLASS – CALIM Oct 2016 VLA Sky Survey Pilot (June – September 2016) Science Verification Field Selection (SSG: VLASS Memo #2) • Program: total 196 hours (+other testing), 3920deg2 total images – 2480 unique deg2 covered • include key deep fields • 240 deg2 in galactic plane/bulge – 720 deg2 repeat 3x • consistency & variability • Cepheus 10x (include test) – 2160 deg2 in SDSS/FIRST footprint • for transient finding – Deep Fields covered: • COSMOS, GOODS-N, CDFS, Elais- N1, Lockman Hole, Lonsdale SWIRE, HAtlas-Bootes, and SDSS Stripe-82

11 VLASS – CALIM Oct 2016 Pilot Survey Fields Science Verification & Test Targets • Tile size ~10° x 8° (4 hr SB)

12 VLASS – CALIM Oct 2016 Science Example Transients with Pilot Survey • Figure from Kunal Mooley, 2160 deg2 in FIRST footprint

Pilot

13 VLASS – CALIM Oct 2016 Pilot Survey Scheduling, Observing, Calibration Development, Testing, Deployment • Schedules generated using custom Python scripts (A. Kimball) – input Tile / mini-Tile boundaries, list of calibrators, output ScanList text file – read into OPT, check LST & slew, adjust Cal durations as necessary, submit • Calibrator Observing Tiles: 10° x 8° – primary calibrators 3C286 or 3C48 (based on LST) Mini-Tiles: 10° x 4° S82 Mini-Tiles: 10° x 2°.2 – gain calibrators: nearest from VLA calibrator list – used multiple PA observations of a gain calibrator for leakage • Lesson Learned: overhead Survey: 50% short blocks, 50% 8+ hour blocks (19% O.H.) – using 4 hour blocks, overhead >22% (depending on calibrator availability) – using Mini-Tiles near transit and RFI belt important but tricky to schedule • Calibration – uses modified version of CASA Integrated Pipeline (CIPL) at AOC/AWS • sufficient in most cases (RFI issues), needs speed improvements for QL (48hrs)

14 VLASS – CALIM Oct 2016 Imaging Challenges for the VLA Sky Survey Risk and Reward • Imaging is the highest residual (post-observation) technical risk to reaching the science goals of the VLASS – different classes of products need different levels of processing • Many of the imaging issues are not special to VLASS, but affect JVLA wide-band wide-field (mosaic) imaging in general – OTFM introduces some issues due the high over-sampling of the mosaic • OTFM “smearing” effects should be sub-dominant – need to characterize with/without using the POINTING table • Pilot data is in hand, and is suitable to test these issues • Needs to be in massively (data parallel) automated pipeline • Needs the latest (and future) improvements in CASA tclean 4.7+ – testing and verification is ongoing

15 VLASS – CALIM Oct 2016 Imaging Challenges for the VLA Sky Survey Risk and Reward

• Imaging is the highest residualQL Workflow (post-observation) Schematic: v.20-Jul-2106 technical (R. Hiriart risk) to reaching the science goals of the VLASS – different classes of products need different levels of processing • Many of the imaging issues are not special to VLASS, but affect JVLA wide-band wide-field (mosaic) imaging in general – OTFM introduces some issues due the high over-sampling of the mosaic • OTFM “smearing” effects should be sub-dominant – need to characterize with/without using the POINTING table • Pilot data is in hand, and is suitable to test these issues • Needs to be in massively (data parallel) automated pipeline • Needs the latest (and future) improvements in CASA tclean 4.7+ – testing and verification is ongoing

16 VLASS – CALIM Oct 2016 Challenge: MTMFS Wide-band Wide-field Imaging for OTF mosaics • In principle, must take into account (reliably over full sky): – system response over full bandwidth (e.g. the complex bandpass and system noise spectrum); – antenna primary beam responses (in the R and L polarizations) over the field-of-view (initially main beam only, ultimately into the near sidelobes), properly rotated on the sky; – the array PSF as a function of frequency, for extended sources, possibly changing (rotating) during the course of the mosaic observing; – the intrinsic source spectrum over this band (intensity, spectral index, spectral curvature, etc.); – amount of flagged channels (e.g. for RFI) per channel over the band; – distribution of visibilities (and weights) in the uv-plane after gridding. • This is what CASA tclean does. The issue is how much of the above is needed for the VLASS imaging products…

17 VLASS – CALIM Oct 2016 Imaging a Mosaic Gather the data, grid, make image

• Decide on area to make uniform sub-image, e.g. Lsub = 1° x 1° (~2-4 x PB)

• Include data taken within a distance Rfld that will contribute to desired area

• Make bigger mosaic image Limg to include all the data (2°.11 x 2°.11) !! Limg%=%Lsub%+%4%Rg%

Joint Mosaic grid uv data together using aperture function and deconvolve Rfld%=%½Lsub%+%Rg% Linear Mosaic image fields separately and combine using Primary Beam weights

Lsub%

18 VLASS – CALIM Oct 2016 Progress: Quick-Look Images! Results from test & pilot observations • 1st Pilot observation June 2016, 1°x1° sub-mosaic : Stripe-82 at 2100+0000

VLASS QuickLook Image 1°x1° subimage (full 2°x2°) 1” pixel size, 2.5” resolution (13Mpix) 416 phase-centers (2 integ) ~10GB vis. data MTMFS nterm=2

σI=134µJy/beam max I = 137mJy/beam (~1000:1) Imaging Time: ~10h (no masking) single-thread Joint Mosaic

19 VLASS – CALIM Oct 2016 1 9 Progress: Images! Results from test & pilot observations • Pilot observations June 2016, 1°x1° subtile Stripe-82 at 2100+0000

VLASS QuickLook Image 1°x1° subimage (1°.56x1°.56) 1” pixel size, 2.5” resolution 416 fields (2 vis per center) MTMFS nterm=2 autoboxing Imaging Time: 4h4m single-thread

(left) VLASS 2.5” resolution, full image; (right) VLASS uv-tapered to 7.5” resolution to enhance low surface-brightness extended emission

20 VLASS – CALIM Oct 2016 2 0 Progress: Images! Results from test & pilot observations • Pilot observations June 2016, 1°x1° subtile Stripe-82 at 2100+0000

VLASS QuickLook Image 1°x1° subimage (1°.56x1°.56) 1” pixel size, 2.5” resolution 416 fields (2 vis per center) MTMFS nterm=2 autoboxing Imaging Time: 4h4m single-thread

(left) NVSS at 45” resolution, full image; (right) VLASS uv-tapered to 7.5” resolution. Good correspondence for many point-like sources.

21 VLASS – CALIM Oct 2016 2 1 Progress: Images! Results from test & pilot observations • Pilot observations June 2016, 1x1 Stripe-82 at 2100+0000

(left) VLASS full 1° QL image; (right inset) VLASS zoom-in on extended source

22 VLASS – CALIM Oct 2016 2 2 Progress: Images! Results from test & pilot observations • Pilot observations June 2016, 1x1 Stripe-82 at 2100+0000 (left) VLASS full 1° QL image & zoom-in, 2.5’’; (right) FIRST cutout image, 5” resolution

23 VLASS – CALIM Oct 2016 2 3 Progress: Images! NVSS 45” Results from test & pilot observations • Pilot observations June 2016, 1x1 Stripe-82 at 2100+0000 (left) VLASS full 1° QL image & zoom-in, 2.5’’; (right) FIRST cutout image, 5” resolution

24 VLASS – CALIM Oct 2016 2 4 Progress: Images! Results from test & pilot observations • Pilot observations June 2016, 1°x1° subtile Stripe-82 at 2100+0000 Zoom: SDSS i-band (red) and g-band (green) on VLASS (blue). Galaxy ID at z=0.25, possibly a in a cluster.

(left) VLASS original 2.5” resolution; (right) SDSS image overlay on VLASS

25 VLASS – CALIM Oct 2016 2 5 Progress: Images! Results from test & pilot observations • Pilot observations June 2016, 1°x1° subtile Stripe-82 at 2100+0000 Zoom: SDSS i-band (red) and g-band (green) on VLASS (blue). Galaxy ID at z=0.25, possibly a in a cluster.

L. Rudnick (UMinn.)

(left) FIRST contours on SDSS; (right) SDSS image overlay on VLASS. Improved VLASS resolution allows us to better classify the radio source.

26 VLASS – CALIM Oct 2016 2 6 Progress: Images & Source Finding! Results from test & pilot observations • Pilot observations June 2016, 1°x1° subtile Stripe-82 at 2100+0000

(left) VLASS original 2.5” resolution; (right) PyBDSM Gaussian modelfit. The PyBSDM model reconstruction is quite good even for this source!

27 VLASS – CALIM Oct 2016 2 7 VLASS Imaging Key aspects regarding imaging performance • Uniformity of survey – 3 epochs, 33885 deg2 (-40° ≤ δ ≤ +90°) – true uniformity not possible due to variable RFI flagging & bright sources • more important to characterize (location dependent) sensitivity • Wide-band Continuum Mosaicking – QLI/SEI/CI products have different performance requirements & timescales • QLI “quick and dirty”; SEI/CI better quality, automasking & self-cal needed – Starting Epoch 2 (2020) the Cumulative ME images start to drive specs • Spectral Cubes – uses same tclean code, no wide-band issues, lower dynamic range • Polarimetry – primary science comes from coarse/fine spectral cubes (RM modeling) – wide-band imaging effects not important, lower dynamic range – wide-field polarization beam effects important (A-projection w/Mueller)

28 VLASS – CALIM Oct 2016 Challenge: Primary Beam & Sensitivity Impact of RFI & weights on Imaging (post-flagging) • Example: 1 degree field in Stripe-82 (center 2100+0000) – tclean “primary beam” image, uses statistical weights (statwt) from CIPL – range ±7% – stripes: RFI – hole near 100mJy source • statwt down-weighting • Issue: • variable spectral weighting over mosaic • must be taken into account in PB correction (I,α,…) • should be done inside tclean or task that has access to weighting

29 VLASS – CALIM Oct 2016 2 9 Take-Away: VLASS Pilot Lessons & Future Detailed Pilot Testing ongoing over next year • Observations complete & successful – scheduling was successful using Amy Kimball’s software – some regions badly affected by RFI (equatorial near transit) • Calibration and Image Processing – CIPL performance was assessed and improved, useable – QL proto-pipeline script gives plausible images • use CASA 4.7 (or 5.0) to get correct JVLA beams and other fixes – initial source finding and comparison with CNSS & FIRST plausible • PDR held in Sep 2016. CDR in Mar-June 2017 timeframe • Pilot Data Available – public in archive, some processing scripts will be made available – https://science.nrao.edu/science/surveys/vlass/the-vlass-pilot-survey

30 VLASS – CALIM Oct 2016 www.nrao.edu science.nrao.edu public.nrao.edu

The National Observatory is a facility of the National Science Foundation operated under cooperative agreement by Associated Universities, Inc.

31 VLASS – CALIM Oct 2016