Productivity and Impact of Radio Telescopes
ABSTRACT
In 2001, 836 papers appearing in 15 journals reported and/or analyzed data collected with ground‐based radio, millimeter, and submillimeter telescopes, plus the HALCA, COBE, and SWAS satellites and a few balloon‐borne detectors. More than 80 telescopes were represented, including 36 that were each responsible for five or more papers. These papers were cited 11,332 times in 2002, 2003, and 2004, for a mean rate of 13.56 citations per paper, or 4.52 citations per paper per year (sometimes called impact or impact factor, and compared to 5.40 citations per paper per year for optical astronomy papers in the same period and 6.42 for space‐based papers). We examine here the distributions of papers, citations, and impact factors among subject areas and telescopes and make some comparisons with the 2100 optical and infrared and 1200 space‐based papers published and cited in the same years. The single largest item in the optical inventory was, naturally, the Hubble Space Telescope, with 16% of the papers and 19% of the citations. Radio astronomy houses an even more dominant entity, the Very Large Array (VLA), responsible for 22% of the papers and 27% of the citations. The VLA is, therefore, proportionately even more influential in world radio astronomy than HST is in world optical astronomy. A third paper in this series looks at papers and citations in the area of infrared, X‐ray, and gamma‐ray space‐based astronomy and planetary missions. Of the “radio” papers, 149 were also optical papers and 76 were also “space” papers, in the sense of reporting or analyzing data in both bands. Their impact factors were 5.71 and 7.51 citations per paper per year, respectively, slightly above the averages for the individual bands. Thus, slightly more than half of observational astronomy is still optical astronomy, but multiwavelength papers are somewhat more influential than average. No radio+optical paper went completely uncited during the triennium.
Received 2005 October 18; accepted 2006 April 18; published 2006 June 14
1. INTRODUCTION
In 2005, we reported data concerning about 2100 papers, published in 18 journals, which had made use of data obtained with ground‐based optical and infrared telescopes plus the Hubble Space Telescope (Trimble et al. 2005, hereafter Paper I). About 250 telescopes were represented, including 25 with primary‐mirror diameters of 3 m or larger. In the next 2 years, 2002 and 2003, those papers were cited 24,334 times, for a mean rate of 11.56 citations per paper, or 5.78 citations per paper per year (sometimes called impact or impact factor). HST was responsible for the single largest number of papers (346.3) and had a somewhat larger than average citation rate (13.71 citations per paper). Keck was second, with 104.5 papers and 20.86 citations per paper. Remarkably, telescopes with diameters less than 1 m contributed 107 papers, but only 5.02 citations per paper. See Paper I for further details, including citation rates as a function of subject matter, distribution of papers among the journals, and details of the method adopted.
In spring 2005, we decided to extend the analysis to papers reporting observational data from radio (including millimeter and submillimeter) telescopes and from space facilities operating at infrared, X‐ray, and gamma‐ray wavelengths, using as nearly as possible the same methods. Because ground‐based radio and optical telescopes generally have lifetimes of many years, data concerning them should be comparable. This cannot be true for the space‐based sample, because of the short lifetimes of most missions. As far as we know, no similar examination of radio astronomy (or space‐based) papers and citations has been carried out. Thus, no historical trends can be sought, and the present work is a sort of “first epoch” study, with the equivalent of proper motions to be determined 5–10 years into the future. In addition, the optical citation numbers have been updated to represent the 3 years of 2002–2005.
We present here the distributions of radio‐observation papers as a function of the telescope(s) used and subject matter, make some comparisons with the optical sample, and look very briefly at the space‐based sample (Trimble et al. 2006, hereafter Paper III).
2. METHODS
As noted in Paper I, an analysis of this sort cannot be done in real time. A facility must be debugged, significant numbers of observations accumulated, and papers written, refereed, and published. Only then can the papers be read and cited; and peak citation rates normally occur a year or two after publication. Our data were collected from May to September 2005, and it would thus have been possible to start over and use papers published in 2002 and cited in 2003 and 2004. The first author simply quailed at the task. At a minimum of 3 minutes per paper, examining 6000 papers, identifying the 4000 observational ones, and recording appropriate information about each takes more than 300 hr. Thus, what we have here is all the observational papers published in 2001 and citations to them from 2002 to 2004, compiled by the second author from the Science Citation Index (SCI)/Web of Science. This appears to be somewhat more complete than the citation data in the Astrophysics Data System, at least in the sense that fewer papers turn up with no citations at all.
The radio situation is generally analogous to the optical one. Some telescopes were just coming on line (including the new Green Bank Telescope), and others were probably past their prime or no longer in existence, although data from them were still being used (the Cambridge 3C survey at 178 MHz, to take an extreme example). The numbers of papers from these classes are likely to be smaller than their lifetime annual averages, but the citations per paper may be representative. In the optical case, Gemini, the Hobby‐Eberly Telescope, and the Italian TNG (Telescopio Nazionale Galileo) were just beginning operation, while the Mount Wilson 100 inch and the Byurakan 2 m were no longer producing new data.
This time around, in April–May 2005, V. T. went page by page through the same issues of the same 18 journals (plus a few that had turned up since summer 2004) and identified all the papers that reported or analyzed data from any ground‐based radio, millimeter, or submillimeter telescopes, plus the HALCA (Japanese interferometry), COBE (microwave background), and SWAS (submillimeter) satellites, and a few balloon‐borne millimeter and submillimeter telescopes, like BOOMERANG. P. Z. looked up all the citation numbers during summer 2005. This is probably the place to confess that the James Clerk Maxwell Telescope (JCMT) accidentally ended up in the optical sample last time. It has been properly moved to submillimeter this time, and we were careful not to count its papers or citations twice!
The rule, as in Paper I, was that it had to be possible to discern from the published paper itself which telescopes were involved. Thus, uses of the FIRST survey were credited to the Very Large Array (VLA), but papers that drew a sample of active galactic nuclei (AGNs) from several catalogs and mentioned only the catalogs were not included. The resulting list consisted of 836 papers, 149 of which had also been in the previous, optical sample. For consistency, they and their citations were here credited entirely to the radio telescopes used, since they had been credited entirely to the optical telescopes in Paper I.
The journals and paper yields were ApJ (205), A&A (including Letters, 195), MNRAS (116), ApJ Letters (110), AJ (85), PASJ (38), ApJS (23), Astronomy Reports (16), Icarus (14), Nature (10), Science (8), J. Astrophys. Astron. (8), PASP (4), Astronomy Letters (3), Ap&SS (1), and none in Astron. Nachr., JRASC, or Acta Astronomica.
For each paper, the following information was recorded: name of first author, number of additional authors, volume and page number, total number of pages, subject, and identity of all the telescopes contributing data to the paper, in the order they were mentioned by the authors. Subjects were chosen from the same list of about 25 used for the optical sample, although a “radio first” approach might well have led to a somewhat different assortment. The largest number of individual telescopes or dishes used in any one paper was 36, and at this point the decision was made to count the Very Long Baseline Array (VLBA) and the European VLBI Network (EVN) as single entities and to accept that there would be a class of “strays” that had contributed only small fractions of the data to one or a few papers. For a few (but only a few) papers, it was not possible to determine which telescopes were used. These appear only in various totals.
As in the optical case, the subject assignment was based on what the authors said they had in mind. Observations of quasars, for instance, could have been aimed toward establishing a more precise coordinate system (the “service” class), understanding the core‐jet structure (AGNs), or tracing out 21 cm absorption along the line of sight (very large scale structure or cosmology). The service areas included catalogs, surveys, and instrument calibrations, as well as astrometry.
Again as in Paper I, all radio, millimeter, and submillimeter telescopes contributing to the data in a given paper received equal credit for the paper and for the citations to it, except that citations were credited only in integers. Thus, the last‐mentioned of seven telescopes contributing to a 13 citation paper received credit for one citation, each of the others getting two. This was not really very common. And optical, X‐ray, or gamma‐ray telescopes contributing to the same papers were simply ignored, just as the radio ones were in Paper I. We suspect that the gravest injustice done by this equal‐credit system is to HALCA, the Japanese VLBI satellite. It was typically used in combination with two to very many ground‐based telescopes, but because of the enormous increase in baseline it made possible, it probably contributed more than 1/N of the weight to most results.
Which telescopes contributed to the 2001 literature, and which to keep track of? In the end, the bean‐counting approach won, and although 80 plus telescopes were mentioned at least once, we report only the 36 individual facilities that contributed five or more papers, plus a number of collectivities. The unlisted include some fairly well known places (like Hartebeesthoek) and others less so (like Qinghai).
How accurate are the data? Several kinds of problems are possible. Rereading a paper, one sometimes finds some additional observational data (usually near the end!) from additional telescopes not caught the first time and not mentioned in the abstract or methods sections. A possible final summary paper (to be submitted elsewhere) will correct the optical data in this respect. Some subject assignments are a bit arbitrary, on the edge between young stellar objects (YSOs) and ordinary stars, for instance. And there are occasional anomalies in the SCI database. About 10 papers (of the original 2100 of Paper I) actually showed fewer citations when examined in summer 2005 than they had the previous year. One paper’s citation numbers had jumped from three in 2 years to 53 in 3 years and may have represented an injustice in Paper I to one small but well‐used optical telescope. Error bars of a few percent should therefore be associated with all the numbers.
3. RESULTS
In Papers I and III, these were grouped under headings reflecting common misconceptions about the astronomical literature, and this section was originally structured the same way to facilitate comparisons of the results. In response to requests from the referees, it has been reorganized, and some of the conclusions and predictions have been removed. Copies of the unexpurgated version are available from the first author and can be sent, on request, in plain brown wrappers.
3.1. The Most and Least Cited Papers
We had not expected radio astronomy to have any one telescope responsible for as large a fraction of the papers as HST is in the case of optical data. Nor did we expect any one paper to stand out in quite the way that the HST Key Project determination of the Hubble constant did (443 citations in 2 years; 632 citations in 3 years). The latter expectations proved to be correct, the former wrong (§ 3.4). The optical standout is Freedman et al. (2001), which was miscited in Paper I.
Table 1 lists the 10 most cited radio papers, giving the numbers of citations in 3 years, the telescopes used, the journals of publication, and the subjects. Eight different telescopes were involved (compared to 10 for the 10 most cited optical papers), and two papers are in both sets, highly cited multiwavelength studies. The most cited radio papers appeared in only three journals and concerned seven different subjects (four journals and four subjects for the optical set). The 10 next most cited papers bring in some additional telescopes—the Institut de Radioastronomie Millimétrique (IRAM), BOOMERANG, Dwingeloo, the 30 m at Villa Elisa in Argentina, the Owens Valley Radio Observatory (OVRO), and the Multi‐Element Radio‐linked Interferometer Network (MERLIN)—a couple more subjects (pulsars, interstellar medium), and two additional journals (A&A and ApJS).
And some papers are never cited at all: 133 (6.3%) of the 2100 optical papers in the first 2 years, dropping to 63 (3.0%) with 3 years of citation data. The radio rate is very similar: 28 of 836 papers, or 3.3%, had no citations in the SCI database in the 3 years following 2001 publication. And none of the multiwavelength (radio+optical) papers ended up completely uncited. There are also no zeros among the radio+space papers, and only a very few among optical+space papers. Perhaps it is just a matter of having two communities who might want to mention the work.
The radio zeros are not distributed uniformly over either the journals or the telescopes, and surprises are few. The mix of topics is much the same set of active galaxies, normal galaxies, interstellar material, and star formation/YSOs that dominates the cited papers, but the uncited appear more often in journals published in less prosperous countries, and they often report data collected at telescopes in difficult places (Russia, Mauritius, and India, for instance). But neither the high‐profile journals (ApJ, A&A, MNRAS) nor the most productive telescopes (VLA, BIMA, VLBA) are completely exempt.
3.2. Trends
On this point, no information is available. Paper I noted that the most productive 3–5 m optical telescopes of a decade before had all yielded fewer papers in the new compilation: not just a smaller percent of the papers, but a smaller number, by 10%–40% per telescope. In contrast, the large ground‐based infrared telescopes had produced more papers and more citations in 2001–2003 than they had in 1990–1993. Possible reasons were suggested, and a prediction or two made. We are not aware of any previous survey of papers and citations from radio, millimeter, and submillimeter telescopes with which the present data could be compared.
That is not to say that such a survey could not exist. It would currently be very difficult to assemble the set of papers published in 2011 and cited in 2012–2014, but there is nothing to prevent one of us (or even one of you) from going back and looking at the radio papers from 1991 and their subsequent citation rates. The number of papers would probably be only about 500.
3.3. Distribution by Subject Matter
Table 2 shows the numbers of papers and citations to them, divided by subdiscipline, and using the same subject headings as in Paper I, except that several topics that appeal to optical astronomers are so rarely studied in the radio regime that we merged a few of them in the table. These include star clusters, brown dwarfs, white dwarfs, and all sorts of binary stars. Not surprisingly, interstellar material, active galaxies, and star formation are larger presences in the radio literature than in the optical.
On average, optical papers are cited more often than radio papers. (The numbers in Table 2 reflect 3 years of citations, not 2 as in Paper I). High‐profile topics in one tend to be high‐profile in the other (cosmology, gamma‐ray bursts, extrasolar‐system planets), and similarly for some low‐profile ones (binary stars, solar system). Supernovae and their remnants and the Milky Way garner more citations per paper at optical wavelengths than radio, and there are a few other differences. Most of the numbers are not statistically significant in the last digit, but the large differences among subjects are real and can been seen in some very different data compilations, centered around individual astronomers rather than facilities (Trimble 1985).
We have been assured by a number of colleagues that this is really just a matter of community size. Lots of astronomers working on, say, active galaxies means lots of citations per paper. But the numbers in Table 2 strongly suggest that size is not the whole story. Gamma‐ray bursts and exoplanets appear in very few, but very highly cited, papers, while the largest radio subdiscipline, the interstellar medium, consists of papers cited, on average, less often than the whole field (10.9 vs. 16.2 citations per paper).
3.4. Distribution by Facility
Table 3, at long last, reveals the numbers of papers, citations, and citations per paper for the 36 ground‐based radio telescopes (including millimeter and submillimeter facilities), balloon‐borne detectors, and space missions that contributed at least five papers each (in our equal‐credit‐to‐all‐telescopes in a paper rubric), plus groups that include the other 44 less productive telescopes (grouped as “other European” and so forth). The ordering is, first, interferometers and their component dishes when used separately, plus other single radio dishes and arrays, second, facilities used primarily for cosmology and study of the cosmic microwave background, and third, millimeter and submillimeter devices. Each is referred to by the name found most often in the papers read for this project (frequently a location), occasionally abbreviated to fit in the column.
Much can be said about the potential unfairness of such numbers. Green Bank (NRAO) was in the process of commissioning a new dish to replace the one that had collapsed, and several other facilities were just coming on line, had already fallen from space, or experienced various other indignities. And it is probable that HALCA contributed more strength to the projects in which it was used than is apparent from giving it equal credit with other participating telescopes, because it contributed the longest baselines. Papers and citations have, however, been apportioned in the same way, so that the relatively low rate of citations per paper for HALCA is real.
With all due reservations, one can nevertheless conclude several things. First, telescopes on well‐supported sites, maintained by countries with a long tradition of radio astronomy, tend to do best. Second, the VLA dominated world radio astronomy early in this century by an even wider margin than HST dominated optical astronomy. The VLA was responsible for 22% of the papers published in 2001 and 27% of the triennium citations, compared to 16% of the papers and 19% of the optical citations for HST.
Another point to be noted is that most of the millimeter and submillimeter facilities rank above average in citations per paper. An exception is the 12 m in Arizona, formerly owned and operated by NRAO, then closed down, and then reopened under different management, much of this during the time when data for 2001 publications would have been being collected. The James Clerk Maxwell Telescope and IRAM facilities are responsible for a sizable numbers of papers, while US observing at shorter wavelengths was largely in the hands of universities—Caltech (OVRO), the Five College Radio Astronomy Observatory of Massachusetts (FCRAO, recently closed), and the Berkeley‐Illinois‐Maryland consortium (BIMA). OVRO and BIMA were very nearly equal in both numbers of papers and citation rates, support for the wisdom of their ongoing unification of dishes, budgets, and programs. With the coming of ALMA (the Atacama Large Millimeter Array, an international collaboration), US involvement in this wave band will move back into public hands.
4. CONCLUSIONS
In 2001, 836 papers published in 15 refereed journals reported or analyzed data from radio, submillimeter, and millimeter telescopes on the ground, in the air, or in space. These were cited 11,332 times in 2002–2004 in journals that form part of the Science Citation Index/Web of Science database, for an average of 13.56 citations per paper, or 4.52 citations per paper per year (a bit smaller than the optical number for the same periods, 5.40 citations per paper per year; the space average is 6.42).
Some topics included many more highly cited papers than others, cosmology, gamma‐ray bursts, and exoplanets being popular in all wave bands. No one radio paper outweighed all the rest, in the way that the HST Key Project determination of the Hubble constant did among the optical papers, but the Very Large Array dominated radio astronomy even more thoroughly than the Hubble Space Telescope did optical astronomy. No multiwavelength paper went completely uncited, and the multiwavelength averages are a bit higher than those in any one wave band.
Although there seem to be no previous studies of radio astronomy papers with which this can be compared, it is tempting to suggest trends that might appear in the future, based on optical data and on overviews such as the decadal reports. All 4 m class optical telescopes have become less productive since the deployment of several 8 m mirrors, and one might expect ALMA to have a similar effect on smaller short‐wavelength facilities. The highly cited topics will probably become more so and the less‐cited topics still more obscure, because of the tendency of funding to follow the crowd.
We hope to come back in a decade or so and “do the numbers” again, but it is perhaps worth noting now that a single study in which all papers and citations are divided among all the observing facilities used, no matter what the wavelength, may not have the expected effect. The total numbers of papers and citations of course stay the same; they are simply more widely spread. Thus, the total number of papers for each and every telescope ever used in multiwavelength studies will come down, although citations per paper will not change. Because, for instance, the 149 radio+optical papers make up 18% of the radio total but only 7% of the optical total, the numbers of papers per radio telescope will necessarily come down further.
The authors are grateful to Major Dawn and Colonel Jim Deshafy of the US Air Force Reserve, who, by remarkable chance, brought them together. We are indebted to James Ulvestad for authorizing, and to NRAO for providing, the page charges. The National Radio Astronomy Observatory is a facility of the National Science Foundation, operated under a cooperative agreement by Associated Universities, Inc. Among the colleagues who read and commented on the first draft of this study, special thanks to Richard Hills and Richard Wielebinski for thoughtful insights and for catching some errors that had eluded us.
REFERENCES
- Freedman, W. L., et al. 2001, ApJ, 553, 47
- Trimble, V. 1985, QJRAS, 26, 40
- Trimble, V., Zaich, P., & Bosler, T. 2005, PASP 117, 111 (Paper I)
- ———. 2006, PASP, 118, 651 (Paper III)