Radio Astronomy in LSST Era¹

By the middle of next decade, the Large Synoptic Survey Telescope (LSST) will be in the midst of a decade-long sky survey, en route to producing a deep, multicolor view of the sky and generating potentially a million alerts nightly. Radio wavelength observations will provide independent, complementary views of the sky, and they will be crucial for full exploitation of the LSST data products. A community workshop, hosted by the National Radio Astronomy Observatory in Charlottesville, VA (2013 May 6–8), was held to explore the science themes of time domain radio astronomy and the radio components of multi-wavelength sky surveys, with the aim of identifying emerging scientific and technical capabilities needed for conducting observations in these areas. The focus of this workshop was primarily on centimeter- and meter-wavelength observations (∼30 MHz–50 GHz). This document is intended to capture the key conclusions and recommendations of that community workshop. Section 2 (and Appendix A) summarizes the likely radio astronomy landscape by the middle of next decade, § 3 discusses the complementary nature of radio astronomical observations and LSST observations, § 4 describes radio wavelength surveys and their complementary nature to the LSST survey, and § 5 summarizes other topics of discussion at the workshop. Appendix B summarizes LSST data products, Appendix C describes potential future radio surveys, and Appendix D lists the workshop participants.

Observations at centimeter- to meter-wavelengths have been responsible for the discovery of many of the objects and much of the phenomena studied in modern astronomy and have resulted in the awarding of three Nobel Prizes in Physics. Discoveries include the cosmic microwave background (CMB), quasars, pulsars, indirect evidence for gravitational waves, astrophysical masers, cosmic magnetic fields, nonthermal emission mechanisms (specifically synchrotron radiation), and the jets from black holes and other objects. Moreover, the use of aperture synthesis techniques—also recognized by the Nobel committee—has enabled observations at radio wavelengths to reach unprecedented levels of imaging resolution and astrometric precision not attained in any other wavelength band, providing the fuel for further discovery.

Technological developments over the latter half of the 20th Century, often stimulated by commercial considerations, offer a path to substantial improvements in future radio astronomical instrumentation. Among the range of improvements are mass production of centimeter-wavelength antennas, fiber optics for the transmission of large volumes of data, high-speed digital signal processing hardware for the analysis of the signals, and computational improvements leading to massive processing and storage. Applying these new technologies to radio astronomy can open up an enormous expanded volume of discovery space. A vigorous international program of upgrading existing radio telescopes and constructing new ones has been stimulated by the potential of applying these technologies to radio astronomy, a process that is likely to continue at least through the current decade.

While it is difficult to capture the full range of international activity, it spans much of the available parameter space for radio wavelength observations. Various radio telescopes now allow access to frequencies near the ionospheric cutoff (∼20 MHz) and as high as the significant tropospheric oxygen absorption lines (∼50 GHz). Filled aperture radio telescopes are in operation providing significant surface brightness sensitivity, while interferometers, using the technique of very long baseline interferometry (VLBI), routinely obtain submilliarcsecond resolutions. With modern digital signal processing, submicrosecond time resolution has been obtained, and, in conjunction with data archives, decade-long light curves of some objects are being produced. Full polarization measurements of the electric field can also be obtained routinely. Even more significant expansions in one or more directions in parameter space will be enabled with future instruments, including the Hydrogen Epoch of Reionization Array (HERA; Backer et al. 2009); the Square Kilometre Array (SKA; Schilizzi et al. 2010; Dewdney et al. 2013); and a high frequency long-baseline standalone U.S. array, possibly a component of the SKA, grown from the Karl G. Jansky Very Large Array (JVLA) and Very Long Baseline Array (VLBA).¹⁵

Appendix A provides more quantitative information about radio telescopes likely to become operational before or during the LSST era. In the spirit of this workshop, these summaries of telescope capabilities are intended to provoke thinking about the kinds of radio wavelength observations that could be conducted both now and during the LSST era.

Time domain observations at radio wavelengths have a long history, with the initial theoretical prediction for the existence of neutron stars (Baade & Zwicky 1934) being verified by their discovery at a radio telescope that had a fortuitous combination of a wide field of view and high time resolution (Hewish et al. 1968). Since then, it has been recognized that variable and transient emissions—bursts, flares, pulses—mark compact sources or the locations of explosive or dynamic events. As such, radio transient sources offer insight into a variety of fundamental physical and astrophysical questions such as the cosmological star formation history (through observations of supernovae and gamma-ray bursts), the nature of strong gravity (through observations of radio pulsars), and mechanisms for efficient particle acceleration. Considerable recent excitement has been generated by the report of millisecond radio bursts having properties consistent with a cosmological population of sources (Thornton et al. 2013), suggesting that the radio sky may harbor other, yet-to-be-discovered populations of transient sources as well. The wide field of view and cadence of the LSST suggest that it may have a similar discovery potential as some of the early radio telescopes.

Table 1 summarizes various classes of known radio transients, many of which are likely to have LSST counterparts. Many of these classes have visible wavelength counterparts, and there are many classes of transients that are discovered first at shorter wavelengths before being observed at radio wavelengths. Consequently, fully exploiting the LSST data stream, particularly to include understanding new classes of sources that LSST might detect, is likely to require some amount of follow-up radio observations. Moreover, is possible that a new class of radio transients might be discovered for which LSST observations would be useful in characterizing the class or determining what the visible wavelength counterparts (if any) are.

The intrinsic transient radio emission or variability from astronomical objects can be broadly grouped into four classes, depending upon their location and emission mechanism. Radio transients can be located in the Galaxy or can be extragalactic, and their emission mechanisms can be either incoherent (e.g., synchrotron) or coherent (e.g., masers). For incoherent processes, particles radiate independently (luminosity L ∝ N, for N radiating particles), which tends to favor centimeter (and shorter) wavelengths by virtue of the Rayleigh-Jeans approximation. For coherent processes, the emission occurs in phase, which can result from stimulated emission, such as occurs in masers, or from collective processes in which the radiating particles are "bunched" within a wavelength (with L ∝ N²). The wavelength of maser emission depends upon the transitions (e.g., OH, H₂O), but collective processes tend to favor meter wavelengths because the volume within which particles can radiate in phase scales approximately as λ³. These classes are not necessarily exclusive, as some kinds of objects could potentially display incoherent and coherent emissions at different times.

Radio signals are also affected by their propagation through a plasma, e.g., interplanetary medium, interstellar medium, or intergalactic medium (Rickett 1990). These propagation effects tend to be strongly wavelength dependent, and are most significant for compact sources, but can produce considerable apparent time variability. In extreme cases, the propagation-induced variability can exceed 100%, producing brightness changes that far exceed any intrinsic variability.

3.1. Transients at Radio Wavelengths

3.2. Discussion

Radio continuum surveys provide an unobscured view of star formation and accretion activity in galaxies out to the highest redshifts as well as a potentially powerful means of constraining cosmological parameters. Nearly all discrete radio sources are extragalactic and are at cosmological distances z ≳ 1, so their distribution on the sky is nearly isotropic. Surveys covering large solid angles are needed to generate statistically useful samples of nearby (z ≲ 0.1) radio galaxies or intrinsically rare objects (e.g., the most luminous quasars), and to make sensitive cosmological tests (e.g., constraining dark energy via the Sachs-Wolfe effect). Radio surveys in the much smaller LSST "Deep Drilling Fields" (≈10 deg² each, Appendix B) will be sufficient for generating fair samples to study the evolution of AGNs and star-forming galaxies.

The power that radio continuum surveys have to constrain the cosmological model is becoming more apparent, mostly due to the high-redshift tail of the radio source populations, which are difficult to access in optical surveys. To be of maximum value, redshifts of the objects detected in radio continuum surveys are required, but redshifts greater than unity will be difficult to obtain even in the SKA era. The SKA will measure the 21 cm H I line toward gas-rich galaxies, but the weakness of this line means that individual galaxies will be difficult to detect in this line at z > 1, and it is unlikely to be useful for gas-poor galaxies. Thus, we will continue to rely on redshifts derived from optical and near-infrared instruments. The LSST's 6-band optical photometry will provide accurate photometric redshifts to z ∼ 1.5 (based on the 4000 Å break spectral feature) and to z > 2 (based on the Lyman break spectral feature); these photometric redshifts, particularly if combined with complementary near-infrared photometry, will provide much of the key data necessary to derive the evolution of activity in the Universe, as traced by radio emission. The possibility of obtaining a large number of photometric redshifts for the z < 2 radio source population with the LSST all-sky survey will allow the remaining unidentified sources to be used for probing the largest scales at the highest redshift (e.g., Raccanelli et al. 2012; Camera et al. 2012). Furthermore, LSST's multiple visits will probe photometric variability to relatively faint magnitudes, providing additional information to aid in determining which objects contain an active nucleus, and can help classify AGN in the deep radio continuum surveys.

The K-correction of optically thin (α ∼ -0.7, S_ν ∝ ν^α) synchrotron radio sources is so large that samples of z > 5 sources may still be relatively small in number, even with sensitive surveys at traditional frequencies (ν ≈ 1.4 GHz, but also see below). Toward specific targets, deep, targeted observations have been essential in obtaining precise positions, and the enhanced sensitivity of the JVLA improves this capability substantially. Thus, it is now possible to use ultra-deep 1.4 GHz observations of well-studied fields, like the Chandra Deep Field-North (CDF-N), which has both extensive spectroscopy (both visible wavelength and CO) and X-ray imaging data, to find substantial numbers of massive, star-forming galaxies to the highest redshifts. At higher frequencies (ν ≈ 5–10 GHz), the K-correction of free-free emission is more favorable. Surveys at these frequencies could have a high yield of intense star-forming galaxies at z > 5, for which the (rest-frame) free-free emission would be the direct result of the ionizing activity of massive stars.

Most radio sources stronger than S ∼ 1 mJy at 1.4 GHz are powered by supermassive black holes in AGNs, while star forming galaxies increasingly dominate the population of fainter sources. The radio emission from a star forming galaxy is roughly coextensive with the star formation region, so the median angular size of faint sources is no more than about 1''. The 1.4 GHz source counts converge rapidly below S ∼ 10 μJy, the flux density of a normal galaxy at z ∼ 1. Consequently, instrumental confusion declines sharply for surveys with better than 10'' (FWHM) angular resolution (Condon et al. 2012). "Natural" confusion will never limit sensitivity, unless there is a previously unrecognized population of sources at microJanksy flux densities.

In addition to discrete sources, at the sensitivity levels of future radio surveys, diffuse sources such as radio relics and halos in clusters of galaxies could be detected. Synchrotron aging within diffuse sources typically results in diffuse sources having steep radio spectra, necessitating a different survey strategy. Relatively lower resolution surveys at frequencies lower than the "standard" of 1.4 GHz will be needed to maintain surface brightness sensitivity. Extended or filamentary structures may also be traced by cross-correlating the radio surveys with the photometric galaxy density from the LSST survey products.

Radio surveys in the LSST era should be designed to match both (1) known source properties and (2) the planned LSST synoptic surveys covering the whole southern hemisphere, plus up to 30 LSST Deep Drilling Fields covering approximately 10 deg² each. Sensitive radio surveys should have detection limits well below the median brightness temperature of star forming galaxies, T_B ∼ 1 K at 1.4 GHz. For example, the 50 μJy beam^-1 (5σ) detection limit in the 10'' (FWHM) beam of the proposed Evolutionary Map of the Universe (EMU) survey of the southern sky (Norris et al. 2011) corresponds to T_B ∼ 0.3 K. The sky density of LSST galaxies with r < 27.7^m is so high (about one galaxy per 25 square arcseconds) that radio position errors of order 0''.2 (rms) in each coordinate will be needed to make complete and reliable position coincidence identifications. Thermal noise alone contributes errors of order 0''.1 FWHM for 5σ sources, so the deepest radio surveys in the Deep Drilling Fields may have to be made with beams as small as approximately 2'' FWHM.

SKA Precursor and pathfinder arrays (e.g., ASKAP, MeerKAT, WSRT/APERTIF, SKA Phase 1; see Appendix A) may attain survey speeds over portions of the radio spectrum (below 10 GHz, in most cases below 2 GHz) superior to those of the JVLA, but the availability of the JVLA opens the possibilities of starting long-term, large-area surveys now that will be complementary to both future radio surveys and LSST surveys. There was considerable discussion of the benefits of such JVLA surveys, such as earlier epochs for synoptic transient surveys, higher frequency measurements, follow-up for the lower frequency Precursor surveys, and radio counterparts for the visible and near-infrared wavelength surveys that exist or are currently underway (e.g., the SDSS, PanSTARRS, the Palomar Transient Factory, Catalina Real Time Survey, Spitzer Extragalactic Representative Volume Survey). Additionally, data processing and imaging techniques can be developed and tested that will be important for enabling future arrays to realize their full potential. Appendix C gives examples of potential surveys that could be performed in the near future with the JVLA and later with the SKA Phase 1, taking advantage of their new capabilities, in preparation for the LSST era.

In the next few years, and complementing the timescale for any JVLA surveys, the Low Frequency Array (LOFAR) will be conducting a series of tiered surveys to explore the radio continuum sky below 300 MHz. The key aims of these surveys are to trace the evolution of activity with the deepest tiers and to constrain the cosmological model through its large-area survey. The frequencies at which LOFAR operates also makes it highly efficient at discovering steep-spectrum radio sources, such as radio halos and relics associated with massive and likely merging clusters of galaxies, and high-redshift radio galaxies, which typically exhibit steep spectral indices (e.g., Chambers et al. 1996; De Breuck et al. 2000; Blundell et al. 1998; Cruz et al. 2007). (However, Roettgering et al. 1994 present a cautionary counterexample.) The low radio-frequencies of the LOFAR surveys should compensate for the large K-correction and could greatly increase the sample of candidate radio sources at z > 5 (and possibly into the epoch of re-ionization), though spectroscopic or reliable photometric redshifts will be required for conclusive identifications.

While the workshop was structured around two science themes, there were several topics discussed that were relevant to both. These topics included "co-observing," probabilistic detection and classification of sources, and the approach toward Deep Drilling Fields. Further, several questions were raised for which either insufficient time for full discussion existed or no consensus was reached.

"Co-observing" was described as a technique for potentially more efficient use of telescopes. The LSST may broadcast its observing schedule, allowing other observatories to know where the LSST will be pointing. In contrast to the current practice of coordinating observations among multiple, independent telescopes, "co-observing" effectively removes a degree of freedom in the coordination, because the LSST schedule can be treated as fixed and known. Radio telescopes could visit LSST fields with the appropriate lag to find certain classes of sources. However, while identified as a possible observing mode, a specific science case or cases that would benefit explicitly from this approach was not identified.

The use of machine learning or artificial intelligence techniques for source detection and classification is not new in astronomy (e.g., Weir et al. 1995; Rohde et al. 2006). There was widespread agreement that such techniques for the probabilistic detection and classification of sources will be essential in the LSST era, if not well before. It may not be possible to determine with complete confidence whether two objects at different wavelengths (e.g., visible and radio) that lie close together on the sky are in fact the same object, particularly if the flux density of one of the sources is near the detection threshold. Similarly, in the early stages of some variable sources, it may not be possible to distinguish between two or more classes of objects.

The locations of the Deep Drilling Fields yet to be selected were of considerable interest. Many workshop participants found the locations of the four selected Deep Drilling Fields¹⁷ (in the "extragalactic sky") to be well-motivated. However, there was concern expressed that the distribution of strong radio sources in or near a Deep Drilling Field may affect the dynamic range of the radio images that could be obtained, thereby limiting the full extent of multi-wavelength data that could be collected for them. (This issue may also affect the recently announced Hubble Frontier Fields)¹⁸ Further discussion topics included the possibility of additional Deep Drilling Fields that have yet to be selected, the need for at least one of these to be optimized for Galactic targets, and the process by which input into the selection of future Deep Drilling Fields could be provided.¹⁹

There was support expressed for interoperability between software tools, such as has been developed in the Virtual Observatory²⁰ (VO) context. Given the multi-wavelength nature of many projects, as stressed in this workshop, the ability to integrate results from multiple surveys or observations will remain essential. Further, for time domain observations, the capability to distribute notices of transients and obtain rapid response to those transients, if needed, requires communication protocols and inter-operability. More generally, the need for a better focus on topics such as long-term data management and archiving was identified. With data volumes continuing to rise, even well before LSST becomes operational, it is increasingly essential that projects take into account how their data can be discovered, accessed, and analyzed.

There was considerable support for the notion that there is groundwork that could be laid now, both in terms of time domain observations and surveys. While there exist mechanisms today for the rapid response of alerts, it is clear neither that these mechanisms have been optimized, nor whether all avenues for rapid response have been identified. Similarly, there will be radio telescopes capable of conducting deep observations or significant surveys well before the LSST becomes operational. Identifying the science case for such surveys, and then executing them, would be important groundwork for future comparisons with LSST images.

Once the LSST is operational, the resulting data set will render unusual events commonplace, and the rarest of events observable. The challenge for the community is to determine how to use the LSST and its major sister facilities to determine how to optimize the follow-up observations that will be necessary to identify, classify, and characterize those events which represent new physics; clearly, given the size of the transient event pool, a set of stringent standards governing access to complementary facilities, both ground and space-based, will be essential to maximize the scientific output of the LSST, and to balance the ongoing operational missions of existing major facilities with time-critical follow-ups.

Meeting this challenge effectively will require establishing and maintaining broad communications channels between the LSST project, its sister institutions, and within the community at a level that appears to have been relatively uncommon to date. Issues which should be addressed might include identifying and setting up appropriate information flows between LSST and the complementary facilities that could be called upon to provide follow-up or concurrent data, developing trigger standards for such observations, making sure that these criteria evolve to reflect ongoing experience as the LSST refines and ramps up its observing programs, and making certain that sister observatories develop and put in place processes to maintain the de facto community ownership of the LSST transient data pool. As an example specific to NRAO, how should observing time on the JVLA be allocated to follow up particularly intriguing events? Should time be granted based on standing trigger proposals (current NRAO practice), or be restricted to first-come/first-serve telescope time allocations only, or should NRAO consider setting aside a fraction of its observing time for follow-up of LSST transient events that satisfy certain community-established criteria, with the data going immediately into the public domain, or some combination of all of the above? While this example is specific to the NRAO and the JVLA, it may have a broader resonance with the entire nature of "target of opportunity" or "rapid response" proposals and the process of time allocation at telescopes.

Finally, there was considerable discussion of future support. In an era of large surveys, with large numbers of people involved, how can individual investigators obtain funding or be recognized for their individual contributions? At least in the U.S., there is unlikely to be any program dedicated to LSST data analysis analogous to a "guest observer" program.²¹ Among the approaches discussed, though no concrete resolution was achieved, was the question of whether it is possible to design a survey in a manner that would make it easier for researchers involved in the survey to obtain funding.

Tables 2, 3, and 4 highlight radio telescopes either currently operational or in construction and likely to be operational in the LSST era. These tables may not be complete, and, particularly for telescopes still under construction or in commissioning, various aspects of their capabilities may change. Dewdney et al. (2013) have conducted a similar comparison.

We stress that these values should be taken as illustrative of relative performance, but there are many factors that affect the ultimate performance for a given observing program. The observatories that operate most of the telescopes listed maintain up-to-date status documents that should be consulted for the most recent values.

LSST will be a large, wide-field ground-based optical telescope system designed to obtain multiple images covering the sky that is visible from Cerro Pachón in Northern Chile. The current baseline design, with an 8.4 m (6.7 m effective) primary mirror, a 9.6 deg² field of view, and a 3.2 Gigapixel camera, will allow about 10000 deg² of sky to be covered every three to four nights using pairs of 15 second exposures, with typical 5σ depth for point sources of r ∼ 24.5 (AB). The system is designed to yield high image quality as well as superb astrometric and photometric accuracy. The total survey area will include ∼30000 deg² with δ < +34.5°, and will be imaged multiple times in six bands, ugrizy, covering the wavelength range 320–1050 nm.

The project is scheduled to begin the regular survey operations at the start of next decade. About 90% of the observing time will be devoted to a deep-wide-fast survey mode ("Main Survey") which will uniformly observe an 18,000 deg² region about 1000 times (summed over all six bands) during the anticipated 10 years of operations, and yield a coadded map to r ∼ 27.5. These data will result in catalogs including over 38 billion stars and galaxies that will serve the majority of the primary science programs. The remaining 10% of the observing time will be allocated to special projects such as a Very Deep and Fast time domain survey.²²

LSST Data Management will perform, in an automated fashion, two types of image analyses resulting in two levels (classes) of data products:

• 1.  
  Analysis of difference images, with the goal of detecting and characterizing astrophysical phenomena revealed by their time-dependent nature. The detection of supernovae superimposed on bright extended galaxies is an example of this analysis. The processing will be done on a nightly or daily basis and result in Level 1 data products. Level 1 products will include difference images, catalogs of sources detected in difference images, astrophysical objects to which sources in the difference images are associated, and catalogs of orbits of identified Solar System objects. The catalogs will be entered into the Level 1 Database and made available in near real time. Notifications ("alerts") about differences between images will be issued using community-accepted standards. The primary results of analysis of difference images—discovered and characterized sources—will generally be broadcast as event alerts within 60 s of end-of-visit acquisition of a field, allowing for rapid follow up, lest interesting information about a source be lost.
• 2.  
  Analysis of direct images, with the goal of detecting and characterizing astrophysical objects. Detection of faint galaxies on deep coadds and their subsequent characterization is an example of this analysis. The results are Level 2 data products. These products, generated and released annually, will include the single-epoch images, deep coadds, catalogs of characterized objects (detected on deep coadds as well as individual visits), sources (detections and measurements on individual visits), and forced sources (constrained measurement of flux on individual visits using known source positions). It will also include fully reprocessed Level 1 data products. In contrast to the Level 1 catalog, which is updated in real-time, the data releases will be static and will not change after release. The analysis of science (direct) images is less time sensitive, and will be done as a part of annual data release process.
• 3.  
  Recognizing the diversity of astronomical community needs, as well as the need for specialized processing not part of the automatically generated Level 1 and 2 products, LSST plans to make its software and APIs available for community use, and to devote 10% of its data management system capabilities to enabling the creation, use, and federation of Level 3 (user-created) data products. Level 3 capabilities will help bridge the gap between LSST Data Products and the science envisioned for LSST, and enable science cases that greatly benefit from colocation of user processing and/or data within the LSST Archive Center.

C.1. Karl G. Jansky Very Large Array

C.2. Square Kilometer Array Phase 1

Table 6 lists the participants in the workshop.