1: %\documentclass[12pt,preprint]{aastex}
2: \documentstyle[12pt,aaspp4]{article}
3:
4: \def\etal{et\,al.\,}
5:
6: \begin{document}
7:
8:
9: \title{Ten Year Review of Queue Scheduling of the Hobby-Eberly Telescope}
10:
11: %----------------Place AUTHOR.TEX after this line------------------------------
12: \author{
13: Matthew~Shetrone,\altaffilmark{1}
14: Mark~E.~Cornell,\altaffilmark{2}
15: James~R.~Fowler,\altaffilmark{1}
16: Niall~Gaffney,\altaffilmark{3}
17: Benjamin~Laws,\altaffilmark{4}
18: Jeff~Mader,\altaffilmark{5}
19: Cloud~Mason,\altaffilmark{2}
20: Stephen~Odewahn,\altaffilmark{1}
21: Brian~Roman,\altaffilmark{1}
22: Sergey~Rostopchin,\altaffilmark{1}
23: Donald~P.~Schneider,\altaffilmark{6}
24: James~Umbarger,\altaffilmark{2}
25: and
26: Amy~Westfall\altaffilmark{7}
27: }
28:
29: \altaffiltext{1}{
30: McDonald Observatory, University of Texas,
31: Fort Davis, TX.
32: }
33: \altaffiltext{2}{
34: McDonald Observatory and Department of Astronomy,
35: University of Texas, Austin, TX.
36: }
37: \altaffiltext{3}{
38: Space Telescope Science Institute,
39: Baltimore, MD.
40: }
41: \altaffiltext{4}{
42: Population Research Center, University of Texas, Austin, TX.
43: }
44: \altaffiltext{5}{
45: W. M. Keck Observatory, Kamuela, HI.
46: }
47: \altaffiltext{6}{
48: Department of Astronomy and Astrophysics,
49: Pennsylvania State University,
50: University Park, PA.
51: }
52: \altaffiltext{7}{
53: Information Services, Texas Tech Library
54: Texas Tech University,
55: Lubbock, Texas.
56: }
57:
58:
59:
60:
61: %----------------Place AUTHOR.TEX above this line------------------------------
62:
63: \begin{abstract}
64:
65: This paper presents a summary of the first 10 years of operating the Hobby-Eberly Telescope (HET)
66: in queue mode. The scheduling can be quite complex but has worked effectively for obtaining
67: the most science possible with this uniquely designed telescope. The queue must handle dozens
68: of separate scientific programs, the involvement of a number of institutions with individual
69: Telescope Allocation Committees as well as engineering and instrument commissioning.
70: We have continuously revised our queue operations as we have learned from experience.
71: The flexibility of the queue and the simultaneous availability of three instruments,
72: along with a staff trained for all aspects of telescope and instrumentation operation,
73: have allowed optimum use to be made of variable weather conditions and have
74: proven to be
75: especially effective at accommodating targets of opportunity and engineering tasks.
76: In this paper we review the methodology of the HET queue along
77: with its strengths and weaknesses.
78:
79:
80: \end{abstract}
81:
82: \keywords{telescopes, techniques: spectroscopic}
83:
84:
85: \section{Introduction}
86:
87: The past three decades have witnessed growing interest in moving from the
88: traditional observatory operation model, in which a single observer (or program)
89: is assigned a series of dedicated nights to a queue-scheduled mode, where
90: an observatory team obtains observations on a variety of programs submitted
91: by investigators who no longer travel to the telescope themselves.
92: The queue model of observing has been implemented by many major observatories;
93: some of the early adopters, such as {\it Hubble Space Telescope},
94: {\it ROSAT}, and the
95: Very Large Array, have become
96: standards for comparison. The great promise of the queue model of
97: observing is the increase in science productivity of an observatory that
98: arises from the flexibility of choosing from a variety of options at a given
99: time in response to changing conditions and priorities.
100:
101: This optimization may take many forms,
102: such as observing the highest ranked targets and programs as assigned by an
103: allocation committee,
104: matching the observing conditions to the constraints of the
105: observing programs, permitting rapid access for targets of
106: opportunity (a particularly difficult challenge
107: for dedicated night observing),
108: and allowing for time-constrained observations.
109: An additional benefit of the queue
110: is that experts with the telescope and instrumentation conduct the
111: observations; this allows the Principle Investigator to avoid the time and
112: expense of traveling to remote sites and training on unfamiliar
113: systems, and can lead to better quality data obtained by those who are
114: more experienced with the details of the telescope and instruments.
115:
116: However, there are disadvantages to implementing queue scheduling. The
117: operational budget of a queue-scheduled telescope, unless it is automated or involves
118: instrumentation that is very simple to operate and maintain,
119: is considerably larger than
120: that of the standard dedicated night model.
121: There are also widely recognized sociological disadvantages to queue operations,
122: which were summarized by Boroson~(1996):
123: \begin{quote}
124: "The arguments from the communities who today use these telescopes or their
125: predecessors against flexible scheduling include 1) doubts about the
126: quality of data taken by observers other than themselves,
127: 2) the loss of their ability to make discoveries by quick follow-up
128: observations, 3) the loss of their overall control of the program
129: (the weather lottery turns into the queue lottery), and 4) the loss of
130: their ability to be creative and innovative at the telescope."
131: \end{quote}
132:
133: A number of ground-based observatories have experimented with and/or
134: implemented queue observing over the past 10 years; examples of these
135: efforts can be found in Saha et al.~(2000), Martin et al.~(2002),
136: Robson~(2002), Kackley et al.~(2004), McArthur et al.~(2004),
137: Comeron et al.~(2006), and Puxley \& Jorgensen~(2006).
138: In this paper we describe the processes and the lessons learned while
139: operating the Hobby-Eberly Telescope (HET) for a decade in a queue mode.
140:
141: The HET is a 9-m class Arecibo-type optical
142: telescope located at McDonald Observatory near Fort Davis, Texas
143: (Ramsey et al.~2007). The HET was designed for narrow-field spectroscopy
144: of faint objects; to minimize the cost, the structure has a fixed
145: zenith angle of~$35^{\circ}$. The telescope can rotate in azimuth, but
146: during an observation the telescope is stationary; the sidereal motion
147: is tracked by a prime-focus instrument package that moves along
148: the focal surface.
149: The HET primary consists of 91 identical one-meter hexagonal
150: Zerodur segments with a spherical figure; these segments combine to form an
151: 11x10~m hexagonal primary mirror. The spherical aberration
152: corrector, carried in the prime focus instrument package, has a pupil
153: with a diameter of 9.2~m.
154:
155: The HET can access declinations (decl.) between
156: $-11^{\circ}$ and~$+71^{\circ}$
157: ($\approx$~70\% of the sky accessible from McDonald), but because of the fixed
158: altitude of the telescope, the observable area of the sky is a ring centered
159: at the zenith. On a given night an object is observable at most twice for
160: a period of an hour or two (the precise availability depends on the
161: decl. of the object) as the motion of the celestial sphere causes
162: the object to enter and leave the ``observing ring" (see Ramsey et al.~2007).
163: Figure 1 shows a projection of the ``observing
164: ring" in hour angle and declination. Targets at a decl.$ = 30^{\circ}$
165: have two tracks each about 1.2 hr long separated
166: by 4 hr. Targets in the north ($decl. > 65^{\circ}$)
167: and south ($decl. < -4^{\circ}$) have a single track.
168:
169: The HET has three facility instruments that are in principle available
170: at all times: the Marcario Low Resolution Spectrograph (LRS; Hill et al.~1998), mounted
171: in the prime-focus tracker; the Medium Resolution Spectrograph
172: (MRS; Ramsey et al.~2003); and the High Resolution Spectrograph
173: (HRS; Tull~1998).
174: The latter two instruments are fiber-fed spectrographs located in an area beneath
175: the telescope structure.
176:
177: The design of the HET naturally lends itself to the queue mode
178: of observing; indeed, it is quite difficult to efficiently operate the
179: telescope in the standard dedicated night mode.
180: Because each target has only one or
181: two relatively narrow windows of opportunity to be observed on a given night,
182: every observation can be viewed
183: as time-critical. For example, consider an observing
184: project that required one specific object to be observed for a total of
185: 7 hr; this
186: program could be completed in a single night at a telescope with
187: a classical design. However, the same program would require
188: more than three nights at the HET. To effectively execute a single program
189: on a night at the HET, the program must consist of a
190: list of targets whose possible HET observing times nicely
191: mesh with each other (e.g., regularly spaced in right ascension at a given
192: decl.), but even such a carefully constructed plan would suffer
193: practical limitations: standards, high-priority objects that
194: conflict between east and west tracks,
195: and the inevitable glitches that occur during a night would
196: wreak havoc with a tightly packed, time-critical schedule.
197:
198:
199: Given this situation, queue scheduling of the HET was an integral part of the
200: HET design, allowing
201: the HET to execute temporal projects, targets of opportunity, and surveys
202: as part of normal operations. One additional complication to
203: the scheduling arises from the governance of the HET; the telescope is
204: a joint project of five institutions:
205: the University of Texas at Austin,
206: the Pennsylvania State University,
207: Stanford University, the Ludwig Maximilians-Universit\"at M\"unchen,
208: and the Georg-August-Universit\"at at G\"ottingen. Each institution has
209: its own unique
210: share of the time. The queue must also keep the partner shares (including
211: dark and light time) in balance.
212:
213: The next three sections of this paper
214: describe the evolution of operations as experience
215: revealed shortcomings in the scheduling algorithm. Section 5
216: discusses the queue observing in the larger perspective and what works
217: well and what does not. The final section presents future plans for
218: the HET queue observing.
219:
220: \section{Development of HET Queue Scheduling}
221:
222: The initial concept for HET operations expected that 85\% of the nights
223: would be conducted in queue mode and 15\% would be assigned in the classical
224: dedicated observer mode.
225: The staffing proposed for this model was
226: two full-time Resident Astronomers (RAs) and three full-time Telescope
227: Operators (TOs). The RAs were responsible for target selection, configuration of the instruments,
228: execution of the observations, and monitoring of the data quality, while
229: the TOs were responsible for moving the telescope
230: to the science targets, guiding once the target is acquired, and monitoring
231: the weather conditions. The RAs were expected to be active researchers, with
232: 25\% of their time allocated for personal research, access to research and travel
233: support, and access to the HET and McDonald Telescopes through the University
234: of Texas allocation. The TOs were not expected to be active researchers, but
235: have been encouraged to participate (up to 7\% of their time)
236: in professional development in approved projects that
237: interest them. To provide adequate staffing for
238: continuous operation of the telescope, the HET facility manager
239: was expected to spend half of the time as a RA.
240:
241: First light of the HET occurred in 1996~December, but the first formally
242: charged
243: queue observations were not taken until 1999~October. During this time
244: frame the development of the queue received another level of
245: complexity with the addition of the National Optical Astronomical Observatory
246: (NOAO) to the HET time allocation process. (The National Science Foundation
247: partially funded the HRS and MRS; one condition of this support
248: was that the astronomical community be given partial access to
249: the HET. The public time on HET is administered by NOAO.)
250:
251: Each institution, including NOAO, was responsible for creating its own
252: HET proposal system, which was denoted "Phase~I"; i.e., each participant
253: empaneled an independent Telescope Allocation Committee (TAC), and there
254: was no interaction or coordination between the TACs. The TACs allocate
255: time in hours, not nights; this assignment includes the acquisition,
256: exposure, and detector readout times, as well as
257: any nonstandard calibration requirements. Typical
258: calibrations, such as flat fields and wavelength/photometric calibrations
259: that can be used by many observations for a given night are
260: considered part of the observatory overhead and are not charged to individual
261: programs. The initial plan adopted by the HET was for the TACs to
262: distribute their share of the telescope time in three priorities,
263: 1, 2, and~3, with 1 being the most important programs.
264: Priority~1 was reserved for targets of opportunity or especially time
265: critical observations.
266: The expectation was that each TAC would submit a list of programs whose
267: targets were widely distributed
268: across the sky and had an appropriate distribution of required observing
269: conditions (e.g., transparency, seeing, sky brightness).
270:
271: The five HET partners adopted the McDonald Observatory trimester schedule
272: (proposals every 4 months). At the start of a new
273: period, all proposals from the previous observing cycle were removed from the
274: queue. The TACs do grant long-term status to some programs but the
275: Principle Investigator (PI) must
276: resubmit a new list of targets at the beginning of each trimester.
277: The NOAO HET TAC operates on a semester schedule; unobserved
278: priority~1 and~2 targets remain in the queue until completed, while
279: priority~3 targets are removed from the queue at the end of each semester.
280: The NOAO methodology of keeping targets in the queue beyond the end of an
281: observing period has the advantage of completing high-priority programs
282: with minimal effort from the PIs and TACs.
283: The five HET partners' methodology of having
284: the PIs submit new proposals each trimester, with the exception of a few
285: programs granted long-term status, offers the advantage of keeping the PIs
286: engaged with the program; reducing their data to the point of being able
287: to offer progress reports to their TACs.
288:
289: The individual PIs are informed of their allocations
290: by the TACs. To activate the programs, the PIs submit to the HET RAs
291: the ``Phase~II" information required to execute the observations:
292: (1)~the proposal, which includes the science goals of the program
293: and a clear description of the data acquisition process, (2)~finding charts for
294: each target, and (3)~an electronic file that contains information
295: for each requested observation. This electronic target file consists of a
296: program number
297: (assigned by the TAC), name, coordinates,
298: $V$-band magnitude of the target, the instrument configuration,
299: the number of visits for the target, the total CCD shutter-open time on
300: each visit, the number of exposures into which that total CCD
301: shutter-open time is to be split, and constraints on image quality, sky brightness,
302: and transparency.
303:
304: Since the TACs are independent, it is inevitable that conflicts would
305: arise in the accepted programs. Examples include identical observations
306: of the same objects, requests for targets of opportunity (in particular,
307: supernovae and gamma-ray bursts), and a few issues that are produced by the
308: design of the HET (e.g., if one institution gives very high priority to a
309: program consisting of a large number of visits to a specific field, then
310: other targets that are accessible during that time period will rarely be
311: observed). There is no ``HET super-TAC" that creates a unified observing
312: program, but we have developed mechanisms for resolving these conflicts.
313:
314: The RAs examine the Phase~II information and identify problems in the
315: overall program. In addition to the conflicts mentioned above, occasionally
316: the combined programs leave a dearth of observations for a specific
317: sidereal time (a "hole in the queue"). Almost all conflicts have been resolved
318: by contacting the relevant TACs; a few cases had to be decided by the
319: HET Scientist, who is appointed by the HET Board of Directors and whose
320: charge is to maximize the scientific productivity of the telescope. The
321: HET Scientist operates independently of the individual partners and the HET
322: staff. The HET Scientist chairs the HET User Committee, which draws its members
323: from all HET partner institutions and both serves an advisory role
324: to the HET Scientist and acts as the conduit of information to the observers
325: at all partner institutions.
326:
327: The electronic files submitted by the PIs are collected into a single
328: database at HET as a single input file for the queue-scheduling
329: software.
330: The initial software design for queue observations was inspired by Spike,
331: which was developed for the {\it Hubble Space Telescope} queue
332: (Johnston 1991).
333: The early HET queue
334: software created a Tcl/Tk graphical user interface (GUI) that
335: could present a list of
336: targets available at any given time, sorted by various single-selection
337: criteria. In addition, the software creates
338: a list of the highest priority bright targets and the highest priority dark
339: targets for every 15 minute interval for each night. This basic software
340: package, denoted "htopx", was used through 2006.
341:
342: While observing, the RA would frequently run htopx throughout the night
343: to plan and update the night's observing schedule in real time.
344: The calibrations needed for
345: the night's suite of observations (frequently obtained with more than one
346: instrument, with occasional multiple configurations per instrument) were
347: also performed, usually at morning twilight.
348:
349: At the end of each night, all of the data acquired,
350: including the calibration files, are transferred to an ftp
351: site for individual PIs to retrieve. Each approved program has a separate
352: directory to receive those program's observations (only the PI has
353: access to the data for a given program).
354: The calibrations are placed in a directory that all PIs can access.
355: When the data are in place, the PI
356: is notified via e-mail that new data are available; this message also
357: contains any concerns that the RA has regarding the quality of the
358: data or presents any difficulties that were encountered in the observation.
359: This system
360: allows PIs to provide almost immediate feedback about the quality
361: of the observations or to suggest adjustments to the observing
362: strategy for future observations.
363:
364:
365: \section{Early Operations for HET Queue and Observing Support}
366:
367: One of the first advantages of queue
368: scheduling became clear in the earliest days
369: of HET operations: engineering time and instrument
370: commissioning were much more effectively combined with science observations
371: than with traditionally scheduled
372: telescopes. When an instrument commissioning run had to be rescheduled
373: (often on short notice), it was trivial to shift the telescope back to
374: science operations.
375: Similarly, when a problem arose that required an
376: engineering effort, it was easy to reschedule the science program. Thus,
377: an individual astronomer would not bear the entire cost of the loss of
378: telescope availability.
379: In addition, the local and visiting
380: engineering staff could halt their work for a brief time,
381: when feasible, to allow an observation to be obtained of a high-priority target.
382: Queue scheduling had the added benefit of allowing the engineers to release the
383: telescope to the RA when their work was completed ahead of schedule or
384: had reached an impasse.
385:
386: The initial plan of having two full-time RAs and the duties of a half-time
387: RA filled by the facility manager proved unfeasible for two reasons:
388: (1)~the facility manager was required
389: to spend all of his time with the daytime support staff to keep the facility
390: running smoothly,
391: and (2)~there were never any requests
392: from PIs for dedicated time. The HET immediately became
393: a 100\% queue-scheduled telescope by default, and hence would be
394: understaffed. To address this
395: staffing shortfall, a third full-time RA position was created.
396:
397: Target conflicts between the various TACs or PIs have been very rare;
398: since 1999, there have been requesting conflicts with two
399: gamma-ray bursts, four supernovae and two extra-solar planets.
400: The policy that we
401: developed calls for sharing data among the Target of Opportunity (ToO)
402: requests and for coordinating the
403: instrument setups. For non-ToO targets our policy has been to observe
404: each request separately according to the queue scheduling algorithm
405: and to distribute the data separately.
406:
407: During early operations the overhead times charged to each
408: PI's program
409: were found to be highly
410: unpredictable, with some visits taking a factor of 2 longer than the average length
411: due to technical problems with the telescope, including variable image
412: quality and mirror alignment issues. To give the PIs a sense of
413: predictability during this period of early operations we set a standardized
414: overhead of 10 minutes per requested visit until the actual overheads
415: could be more predictable, which occurred once full science operations began.
416:
417: Despite the e-mails that the PIs received each night, the PIs found that they
418: needed a Web-based interface to
419: monitor the progress of their programs.
420: To create this page it was necessary to access the night
421: reports (where the log of observed spectra were kept), the electronic
422: Phase II (where the log of the completed observations were kept), and a
423: file containing the TAC allocations for each program. Had this demand been
424: anticipated these various data sets would have been stored in a single database.
425: The resulting Web page shows the
426: observations that have been attempted, the targets in the queue, along with
427: their status
428: (completed or active), and the amount of TAC telescope allocation used for each program.
429:
430: Target selection by the RAs at the telescope was not found to adequately reflect the
431: TACs' wishes. Most of the time, the RA would be forced to choose from a very wide
432: range of targets, all with the same TAC priority.
433: The three-priority system did not offer enough dynamic range to represent
434: the TACs' scientific ranking.
435: In addition, there were times when significantly poor seeing or transparency
436: limited the available targets, but some types of science could still have been conducted.
437: The TACs did not want their institutions to be charged for the long setup times
438: (due to the
439: difficult conditions) and extended exposures (again due to bad
440: conditions) for these normally easy targets. To address these problems we
441: developed a new priority scheme, 0--4:
442:
443: Priority 0. -- Time is allocated for
444: targets of opportunity or very time-critical observations. Up to 25\%
445: of the priority 1 time can be assigned to priority 0.
446:
447: Priority 1. --
448: Targets would constitute one-third of the expected partner's observing time during
449: the period including average weather losses and time lost due to engineering
450: and technical problems.
451:
452: Priority 2. -- Targets would make up the second third of the weather-corrected
453: partner share.
454:
455: Priority 3. -- Time would make up the final third of the
456: weather-corrected partner share.
457:
458: Priority 4. -- Time is unlimited, but the
459: expectation is that this time would only be used when normal operations
460: could not be conducted (e.g. moderate cloud coverage or very poor seeing).
461: In addition, priority
462: 4 time would not have any charged overhead and would only be charged at
463: half the nominal conditions exposure time.
464:
465: The reason for only charging
466: half the nominal conditions exposure time for priority 4 (P4) targets
467: is to give the PI some advantage
468: and incentive to work with data that are acquired under sub-par conditions,
469: and to compensate the PI for the extra effort that can be required to reduce
470: such data.
471:
472: Despite some of the problems mentioned above, some aspects of HET's queue-scheduled
473: observing program were immediately successful (see Ramsey et al. 1998).
474: The RAs were directly involved in informing the PIs of data delivery
475: (and providing comments if useful) and they rapidly responded
476: to comments from the PIs. This highly personalized attention given to the
477: PIs required the RAs to make temporary notes in the PI's
478: Phase II materials and to communicate with the other resident astronomers about
479: these changes. This effectively added a customer service aspect to the
480: job description of the resident astronomer. The TACs and PIs overwhelmingly
481: approved of this aspect of the HET's performance.
482:
483: To better serve the PIs, several new fields were added to the Phase II files.
484: The first was an option for the PI to request radial velocity, spectrophotometric,
485: telluric or other types of standards. The second addition allowed the PI to
486: request a standard set of calibrations or to make a request for specific
487: calibrations, such as a wavelength comparison spectrum immediately after
488: their visit. We also created a flag that would allow the PI to group two
489: or more observations together, such as a specific standard to be observed
490: directly after their main target.
491: The final change to the electronic Phase II specifications
492: was the addition of two synoptic-specific
493: fields: one to specify the frequency of the visits, and the second
494: to specify when the visits should occur, using a date string. The latter allows for
495: specifying specific visit dates and/or ranges of dates in which to
496: make visits. A complete list of the entire Phase II format can be found
497: on the HET Observing Support Web pages.\footnote{See http://het.as.utexas.edu/HET/hetweb/}
498:
499: \section{Full Operations with the HET Queue and Observing Support}
500:
501: Under full science operations, the typical queue contains 1500 entries, with
502: each entry describing from one to 20 requested visits. The average
503: requested number of visits is 1.9. Typical requested visit times are
504: 1300 s for LRS and 600 s for HRS (the HRS visits are
505: dominated by planet-search proposals, which involve a large number of short
506: visits spread out over the trimester). We typically have 45 programs at
507: the beginning of the trimester. Figure 2 exhibits the cumulative distribution
508: of TAC allocations for each priority. In 2006 the total shutter-open time made
509: up 50.3\% of the clear science time. The remaining 49.7\% includes target overheads,
510: mirror alignment, science calibrations and time lost due to problems. In
511: 2006, we lost 31.9\% of the nights to bad weather and spent 3.2\% of the
512: nights conducting scheduled engineering tests.
513:
514: Our TACs have found that allocating
515: a blend of priority classes to a single program allows the more critical
516: targets to be observed and gives some flexibility to the PIs. The projects
517: that have successfully made use of the lowest priority time, P4, have been
518: programs that submit a large number of targets distributed widely over
519: the sky, with exposure times that are short ($<600$ s).
520: When the number of submitted
521: programs is compared with partner share, we find a rough average of one program
522: for each 2\% share of telescope allocation. The smallest HET partners have
523: roughly 6\% share in the observing time. During a trimester, we typically also
524: receive two to five new
525: proposals from the TACs. These are allocated from time that the TACs hold back
526: for unforeseen exciting science, such as ToO programs.
527:
528: The expectation for
529: the partners was that they would, on average, share equally in all observing
530: conditions. For some of the partners, this has not been the case. For
531: example, in 2006 two of the smaller HET partners had less than 1\% of their
532: total submitted targets in bright time (sky brightness V$< 19.5$). The HET
533: Board of Directors and HET User Committee do not have any formal mechanism or
534: policy to deal with this inequity.
535: By the end of the trimester, bright time is
536: largely dominated by P4 targets submitted by the two larger HET partners.
537:
538: When an observation is completed, the RA attaches a flag to the night report
539: entry to indicate if the observation is charged or uncharged. The charged
540: category is further broken down into acceptable and borderline.
541: In 2006, the borderline observations made up 5.3\% of the charged
542: observations. The PIs have the option of requesting an observation be repeated;
543: usually these are borderline observations that the PI believes are of too poor a
544: quality to use for their analysis. In 2006, PIs rejected a total of
545: 3.1\% of the science-collecting operating hours.
546: From 1999 to 2006, there has been
547: only a single case of a disagreement between an RA and a PI as to whether
548: an observation should be reobserved at no cost to the PI.
549:
550: Once the HET entered full science operations it became clear that the
551: HET facility manager could not effectively monitor the night-time science
552: operations without assistance. A position was created to manage the
553: science operations and supervise the science operations team. The science
554: operations supervisor reports to the facility manager, but also interacts
555: with the TAC chairs and the HET scientist. The science operations
556: supervisor maintains metrics on science operations, approves procedures for
557: night time operation, monitors the nightly operations, and makes night-to-night
558: decisions about the operation of the queue, including balancing
559: engineering time with science operations. The science operations
560: supervisor also creates a monthly report for the TACs. This report describes
561: the status of the observing programs and that of the facility, which
562: allows the TACs to make changes to the priorities to ensure that
563: their most important programs are completed.
564:
565: Under the three-TO/three-RA staffing plan,
566: normal attrition left the HET partially crippled. The re-hire and
567: training process takes approximately 6 months, and operations with just
568: two RAs or TOs led to observing inefficiencies,
569: decreased morale, and an increased attrition rate. Just as spares for
570: critical components are required for the telescope, we required a plan for
571: replacement of critical staff. The science operations supervisor would
572: be required to act as a part-time RA during normal
573: operations and to fill in as a full-time RA during any
574: hiring and training periods. A fourth TO was hired, leaving night operations somewhat over-staffed under normal conditions, so extra duties were added to
575: the TO position. These new duties included working in the
576: afternoon with the day staff to assist in telescope maintenance and
577: engineering operations, and working before sunset to prepare the telescope
578: for operations. A summary of the HET science operations staff is
579: given in Table~1. The HET scientist is also included in the table, even
580: though that position is independent of HET operations. The HET has a
581: total of eight full time equivalents (FTEs) dedicated to science operations.
582: This is 40\% of the
583: total FTE compliment for the HET, and is 27\% of the total HET budget.
584:
585: One of the obstacles to program completion was found to be PIs who
586: submit programs with unrealistic observing requests (e.g., extreme image
587: quality constraints, expectation of a large number of visits).
588: A Phase I tool was developed to allow the PI to get a
589: sense of which targets and observing programs have a reasonable expectation of
590: being completed.
591: Some TACs require that each PI explain the feasibility
592: of completing the observing program in the allocation period, based on results
593: from this Phase I tool. Although this tool
594: allows a PI to determine if the program is feasible, it does not show
595: conflicts that might arise with other programs in the queue. The most
596: common form of conflict is overly subscribed portions of the sky;
597: the north Galactic pole and the Coma cluster of galaxies are typical examples.
598: A first attempt
599: to understand these conflicts is made by breaking down the queue into
600: histograms based on the sidereal time (ST) for each requested visit
601: and comparing the expected number of visits that may be completed at each
602: ST bin. Figure 1 shows that for targets with declinations between
603: $-3^{\circ}$ and $+64^{\circ}$,
604: there are two available windows
605: in which they may be observed (east and west), and that these tracks are
606: often separated by several hours. However, during any given observing period
607: (trimester), some targets may only have a single track available, since the east or
608: west tracks may not be observable because of temporal conflicts with morning or
609: evening twilight.
610:
611: Figure 3 exhibits the histogram for the 2006-1 period for priority~0--3
612: dark time (sky brightness constraint of V $>=$ 20.6) targets and includes all visits (east
613: and west).
614: The unfilled histograms are the requested dark time
615: visits in the queue, and the filled histogram are the
616: completed visits. A target at $decl. = 30^{\circ}$ with a single
617: requested visit will appear
618: twice in the figure, once for the east track and once for the west track. Similarly,
619: completed visits in Figure 3 are double counted for targets with both east and west
620: tracks. The solid line represents our estimate for target
621: completion at each ST, based on the Phase I program completion expectation tool
622: that is mentioned above and found on the HET Observing Support Web pages.
623: The completed
624: visits exceed the estimated visits between ST 21 and 0 because these are targets
625: with double-counted visits; e.g., the east track falls during the day, but the target
626: was observed in the west track.
627: At the beginning of the trimester, the RAs report to the TACs
628: any programs that are in jeopardy of not being completed because of target
629: density on the sky, e.g., the large number of requested visits over
630: the number expected to be completed at ST $=$ 10 was largely due to one
631: program. The TACs can report this information to the PIs, or they can
632: take any appropriate steps to modify the queue. The RAs can also attempt
633: to exceed the estimated completion rate for an overly dense region by systematically
634: observing targets with two available tracks
635: at STs that are less densely populated, such as STs between 6 and 8. This was
636: done in ST bins $ 10$ and $ 14$.
637:
638: While the setup times during full science operations were more predictable than
639: during early operations, the setup times charged to PIs were found to
640: be larger, on occasion, than the assumed setup times adopted in the Phase I proposal, which can
641: make program completion in the allocated TAC time difficult. In order to ensure
642: that the program could be completed, and to give the PIs
643: an element of predictability, a cap is placed on the amount of overhead time
644: charged to any requested visit. The overhead cap is instrument dependent.
645:
646: The choice of targets to observe during night operations is made by
647: the RA, who bases the selection on a balance of object
648: availability (how many more visits can be made to the target during the current trimester),
649: TAC priority, synoptic constraints,
650: and current observing conditions. To assist the RA
651: with these choices
652: and to add a level of standardization, a modified-priority scoring system
653: was created.
654: A set of modifiers are added to the TAC's priority, and the results are sorted
655: to aid in target selection.
656: The modifiers are listed
657: in Table 2, along with the maximum magnitude of the modifier. For
658: targets that have synoptic constraints,
659: the PI can give either a flexible or a fixed range;
660: the synoptic modifier given in Table 2 is for the flexible range.
661: The modified-priority system is tuned such that:
662:
663: \begin{enumerate}
664: \item The total modifiers on a priority~4 target never allow it to outrank
665: a priority~3 target.
666: \item The synoptic modifiers on a priority~2 synoptic target can
667: allow it to outrank a priority~1 target.
668: \item The object availability modifier on a priority~2 target can
669: allow it to outrank a priority~1 target.
670: \item The total modifiers on a priority~1 target can combine constructively to
671: allow it to outrank a priority~0 target.
672: \item The total modifiers on a priority~3 target can combine constructively to
673: allow it to outrank a priority~1 target.
674: \end{enumerate}
675:
676: The modified priority system was created and implemented by the RAs
677: but it is continuously evaluated by the HET User Committee. The
678: system allows for the creation of an initial observing plan for each
679: night by selecting the targets with the highest modified
680: priorities and then filling out the schedule with lower modified-priority
681: targets. The modified-priority algorithm creates a suggested plan
682: for the RA, and it provides some predictability
683: as to the order in which the targets will be observed, which is useful
684: for planning instrument changes. However,
685: as the observing conditions change, the RA is still required
686: to make critical decisions about which targets to observe.
687:
688: Program completion is one of our critical metrics.
689: For over-subscribed or perfectly subscribed programs (programs that have
690: enough targets to be completed in the TAC-allocated time), the program completion
691: percentage is calculated from the amount of TAC-allocated time completed at
692: each priority. For under-subscribed
693: programs (not enough targets to fill their TAC-allocated time), the program
694: completion percentage is calculated from the completed versus
695: requested shutter-open time at each priority.
696: Figure 4 exhibits the median completion rates for the individual
697: programs from 2004 through
698: 2006. The modified-priority system was implemented at the beginning
699: of 2005, and further modifications were made in 2006 to fine-tune the
700: algorithm. From this plot, we have concluded that our completion
701: rates for the highest priority targets (those with the lowest priority number)
702: are not driven by the algorithm,
703: but by the weather and the feasibility of the program. The lower
704: priority targets have a better completion rate under the modified-priority
705: methodology than under the subjective, RA-specific nightly
706: decisions.
707:
708:
709: \section{Review and Status: Summary of what does and does not works well}
710:
711: A subjective analysis of the HET science return reveals that we have been most
712: successful by concentrating on target-of-opportunity surveys (e.g.,
713: Frieman et al. 2007), synoptic programs (e.g., McArthur \etal 2004; Kaspi
714: \etal 2007),
715: and surveys with a
716: wide distribution on the sky (e.g., Sowards-Emmerd et al. 2005). For
717: more information about completed science programs see Ramsey (2005)
718: and Ramsey \etal (2007).
719: Most of the incomplete programs
720: have been pencil-beam surveys (e.g. many
721: visits to a single target, or many targets in a single, small region of the sky).
722: Most of the failed programs have failed due to technical problems at the
723: telescope and are not
724: due to problems with the HET queue-scheduling methods.
725: While the efficiency benefits of queue scheduling under full science
726: operations are documented in the literature (see references quoted in the
727: introduction), we found
728: that in early operations the benefits were even more substantial. The ability
729: to begin or end an engineering effort to attack a subtle or intermittent
730: problem without disrupting a specific PI's allocated time, or the ability to
731: pause engineering for a very high priority science target made the first
732: years of science operations more productive. Even today,
733: planning an engineering run is far easier on a queue-scheduled telescope
734: than on the traditionally scheduled telescopes at McDonald Observatory.
735:
736: The basic lesson learned from the Hobby-Eberly Telescope queue-scheduling
737: effort is that a customer-oriented observing effort can be highly successful.
738: While the HET would have benefited from greater software development early on,
739: the evolutionary manner in which the software and observing styles developed
740: has allowed us to understand the needs and improve service to the institutional partners and PIs.
741: Since it is the PIs who ultimately will be analyzing the data, making an
742: effort to involve them as active participants of the process is critical.
743: The HET does this in several ways:
744:
745: \begin{enumerate}
746: \item The PIs are allowed and encouraged to e-mail the RAs
747: of the Hobby-Eberly telescope during their Phase I and
748: Phase II planning.
749: \item The PIs have access to their data on a nightly basis.
750: \item The PIs can make some changes to their Phase II information after submission.
751: \item The PI can request that observations that may have been compromised
752: by the weather or improperly observed be repeated.
753: \end{enumerate}
754:
755: These steps to involve the PI address many (but not all) of
756: the concerns that most PIs have about queue scheduling as
757: summarized by Boroson~(1996) and listed in the introduction, but only if the
758: PIs are encouraged to actually examine the data and provide timely feedback to
759: the RAs. In addition, we have allowed our software tools
760: to evolve as the PIs and TACs request new features or
761: interfaces. This necessary evolution arises from a customer-oriented model
762: for the science operations. To further our efforts to improve
763: customer service, we created a survey to be filled out by the PIs
764: after Phase II and again half way through the trimester. Our response rate
765: for the Phase II survey was 22\%, while that for the
766: mid-trimester survey was 28\%, out of 32 PIs. The results of the survey
767: can be summarized as follows: 56\% of the responses had positive comments
768: and no constructive criticism, 19\% had constructive comments on our Web
769: documentation, 19\% had constructive comments on our Phase II
770: process, and 6\% had constructive comments on the instrumentation. While the
771: feedback was moderately useful, we believe the low response rate did not warrant
772: a survey be repeated every trimester, since the vast majority of our
773: PIs are the same from trimester
774: to trimester.
775:
776: One unique feature adopted by the HET is our poor-conditions priority 4 policy.
777: This has been tremendously useful and successful. Not only have P4
778: targets been observed under poor conditions, but the short-exposure targets
779: are also useful for filling in between high-priority targets on nights with good
780: weather. Recall that the HET must wait for specific targets to enter
781: the observing annulus on the sky to be observed, so having an abundance of
782: short-exposure targets in the queue to fill in the "dead-time" leads to
783: higher scientific productivity. In 2006, 35.8\% of accepted CCD
784: shutter-open time was for P4 targets. Even for a conventional telescope,
785: there will still be reasons to have a large pool of short-exposure
786: targets to be observed in poor weather conditions or to fill
787: in between high-priority time-critical targets. Without some incentive,
788: such as our policy of half-charge for the nominal
789: shutter-open time and no overhead charges, the PIs and TACs would have reduced
790: incentive to submit targets for poor observing conditions, and the
791: queue would be less efficient and the telescope less productive.
792:
793: Another important lesson that we have learned from running fully queue-scheduled
794: operations at the HET is the need to have a plan for
795: replacement of critical staff. Specifically, we must be able to cope
796: with the normal attrition of TOs and RAs.
797: This plan can be as simple as identifying who will take up
798: the work load until new staff are hired (as is the case for our RAs
799: and the night operations supervisor) or the more extreme
800: measure of having an extra person in the rotation so that the observatory
801: is never under-staffed (as is the case for our telescope operators).
802: With the increasing costs of telescopes, the expense of having an additional
803: telescope operator and resident astronomer is small but crucial.
804:
805: While the staff require considerable training to properly operate a queue,
806: we have found
807: that the TACs and PIs also need training to properly
808: populate the queue. The PIs must have a good working knowledge of how the
809: telescope and instrumentation operate and what their capabilities are.
810: One of the
811: areas in which the HET operations staff have been less successful is in maintaining
812: a living repository of knowledge of how the HET and
813: its instruments are performing. Without this information, the PIs have
814: relied on communication with each other to determine if projects are feasible.
815: This problem could be solved with one more FTE. This additional person could
816: take on the documentation of the facility's capabilities as a principle
817: duty, or else the person could be designated as another RA and then have the
818: documentation duties split among the RAs.
819:
820: Just as educating the PIs is critical, the TACs must be staffed with people
821: knowledgeable about the telescope's capabilities in order to
822: determine if the
823: science proposals have merit. In the case of the partner institutions this
824: does not seem to be a problem, since most of the TAC members are or have been
825: PIs. There have been some instances in which programs approved by non-HET partners
826: have not been well suited for the HET; while technically feasible, such
827: programs have a low expectation of completion.
828: For future queue-scheduled telescopes, we recommend that the observatory staff
829: participate in the time-allocation
830: process during early operations until a significant
831: fraction of the TAC members are familiar with the facility. This participation could
832: take the form of nonvoting members or as an external review of project
833: feasibility.
834: We have found that continued interaction with the TACs or a TAC
835: representative is the best approach to maximize each institution's science
836: productivity. Most of our interaction has been
837: in the form of monthly reports from which the TACs can follow the progress
838: of program completion, the institutional shares and the instrument and
839: facility status. In addition, the TACs' policy of retaining a small percentage
840: of their telescope allocation to be
841: distributed later during the trimester to cover unforeseen opportunities has
842: frequently yielded significant science results.
843:
844: One of the major ongoing problems we have had in completing programs has
845: come from an unforeseen impact of adding new features and improvements to our
846: instrumentation. One of the reasons that HET works well is that the
847: instrument complement and design allows HET to rapidly change between
848: instruments to take advantage of changes in the observing conditions and
849: to complete the most difficult science that the observing conditions allow.
850: As more features have been added to the HET instrument complement,
851: one of the greatest challenges to queue scheduling has been balancing
852: requests for mutually exclusive
853: instrument configurations. The HET has one instrument, the LRS, that
854: cannot have all modes available all the time. Four grisms are used for
855: LRS observations, but the instrument can hold only two on a given night
856: (grism changes are limited to daytime operations). One of the slots
857: is devoted to the most requested grism (g1). The demand for the remaining
858: dispersers (g2, g3, and e2) is quite unequal, with g2 being the most popular
859: and e2 rarely being requested. Figure 5 displays the completion rates for
860: priority~0--3 LRS programs in 2005 and 2006, divided into three groups:
861: g1, g2, and g3$+$e2. The median completion rates are roughly
862: equal for the high-priority programs, largely due to the
863: roughly one to four instrument changes per dark run. For the lower priority
864: time (larger priority numbers), the completion rates for the less subscribed
865: setups are considerably lower. This problem was particularly complex
866: when one of the partners with only a small share requested an instrument
867: configuration that cannot be mounted at the same time as one of the more
868: popular configurations. The continuing desire to upgrade to specialized
869: instrumentation must be balanced against the impact that that upgrade
870: would have on the flexibility of the queue to respond to different
871: observing conditions and configurations.
872:
873: Any new collaboration on a queue scheduled telescope should have a well-designed
874: plan for dealing with conflicts between the competing target requests
875: and for dealing with partners who wish to specialize in one specific part
876: of the sky or one specific observing style. This last issue can manifest
877: in requests for all dark time or in an institution requesting only one
878: specialty instrument. This situation can lead to the awkward problem of
879: a partner not receiving their full allocation in an observing period which
880: begs the question: what techniques, if any,
881: can be employed to rectify the situation?
882: At the HET we have encouraged any partner who has fallen significantly behind
883: their partner share to submit more targets and have given those targets extra
884: emphasis through the modified-priority algorithm.
885:
886: \section{Next Generation of Queue Scheduling}
887:
888: Our work with HET and its queue methodology is an ongoing program;
889: most of our operations efforts are now going into improving the PI, TAC,
890: and RA software interfaces. We wish to automate many of the
891: features that are currently performed manually. For example,
892: checking the status of a program with respect to its TAC allocation
893: and determining whether the program has used its allocation requires one
894: to manually place on hold any
895: remaining targets for that program. Automation will not only reduce
896: the RA work-load, but will also increase reliability by removing the human
897: element from the loop.
898: Some examples of these changes are:
899:
900: \begin{enumerate}
901: \item Automate the checking of program completion.
902: \item Allow a PI to add or remove targets directly to or from the queue.
903: \item Allow the TACs to create programs and allocate time directly to the queue.
904: \item Create an automated list of calibrations and standards to be taken for each target.
905: \end{enumerate}
906:
907: One of the strengths of the HET has been the fast response
908: to PI requests, and we wish to build on this strength. Current requests
909: by PIs for changes to their programs
910: are not common outside of the synoptic programs.
911: Medium-term planning has been more sensitive to the site's variable
912: weather than to changes to PIs' programs, and we anticipate that allowing
913: the PIs greater control over their programs will not change this.
914:
915: These changes will require a transition from the current HTML-based night reports and
916: Tcl/Tk$+$tab-delimited text file to an integrated database. Our choice for
917: this upgrade is a
918: MySQL database with a variety of interfaces, including HTML, PHP, JavaScript, and
919: AJAX technologies.
920:
921:
922: The current working prototype uses PHP sessions to authenticate users, then allows
923: for flexible, individual and graduated access to program and operational data.
924: Permissions can be granted according to the status of active programs or membership
925: on allocation committees. Access to Web applications to plan, administer, and
926: evaluate programs, and to interact with the queue can also be customized. It is
927: envisioned that a well-designed suite of tools that provides timely and accurate
928: information on operational conditions, including queue activity, for PIs, TACs and
929: observatory staff has the potential not merely to improve efficiency and
930: productivity through automation, but to give all users the power to leverage the
931: advantages of the service model to refine their own programs in a dynamic
932: queued environment.
933:
934: \acknowledgments
935:
936: We are deeply indebted to the
937: night staff of the HET over the last 10 years: Grant Hill,
938: Ben Rhoads, Teddy George, Gabrelle Saurage, Frank Deglman,
939: Mike Soukup, Michelle Graver, Martin Villarreal, John Caldwell,
940: Vicki Riley, Chevo Terrazas, and Heinz Edelmann. We would
941: also like to thank the day staff in West Texas and the
942: Austin engineering
943: staff for all of their efforts on the HET.
944: We would like to thank Larry Ramsey, Rob Robinson,
945: Roger Romani, and the HET User Committee for their
946: continuing guidance. Support for
947: A. W. was under REU program grant NSF AST-0243745.
948: The Hobby-Eberly Telescope is a joint project of
949: the University of Texas at Austin,
950: the Pennsylvania State University,
951: Stanford University, the Ludwig Maximilians-Universit\"at M\"unchen,
952: and the Georg-August-Universit\"at at G\"ottingen.
953: The HET is named in honor of its principal benefactors,
954: William P. Hobby and Robert E. Eberly.
955:
956: \clearpage
957:
958: \begin{thebibliography}{}
959:
960: % Queue Scheduling - One Approach Toward Evaluating Gains and Drawbacks.
961: \bibitem[Boroson 1996]{tab96}
962: Boroson, T. 1996, in ASP Conf Ser 87, New Observing Modes for the Next
963: Century, ed. T. Boroson, J. Davies \& I. Robson (San Francisco: ASP), 13
964:
965: % VLT Scheduling paper - VLT service mode operations at seven years
966: \bibitem[Comeron et al. 2006]{c06}
967: Comeron, F., Mathys, G., Kaufer, A., Hainaut, O., Hanuschik, R.,
968: Romaniello, M., Silva, D., \& Quinn, P. 2006, Proc. SPIE, 6270, 30
969:
970: % SDSS-II Supernova Survey: Technical Summary
971: \bibitem[Frieman et al. 2007]{jaf07}
972: Frieman, J., et al. 2007, AJ submitted
973:
974: % LRS Paper
975: \bibitem[Hill et al. 1998]{gjh98}
976: Hill, G.J., Nicklas, H.E., MacQueen, P.J., Tejada, C., Corbos~Duenas,
977: F.J., \& Mitsch, W. 1998, Proc. SPIE, 3355, 375
978:
979: % SPIKE: AI Scheduling Techniques for Hubble Space Telescope
980: \bibitem[Johnston 1991]{mdj91}
981: Johnston, M. D. 1991, in NASA Goddard Space Flight Center Space
982: Network Control Conference on Resource Allocation Concepts and
983: Approaches, (REPT-91B00130; Greenbelt: GSFC), 183
984:
985: % The JCMT Observing Queue and Recipe Sequencer
986: \bibitem[Kackley et al. 2004]{rdk04}
987: Kackley, R.D., Rees, N.P., Walther, C., \& Jenness, T. 2004,
988: Proc. SPIE, 5496, 718
989:
990: \bibitem[Kaspi et al. 2007]{kaspi07}
991: Kaspi, S., Brandt, W.N., Maoz, D., Netzer, H., Schneider, D.P., \&
992: Shemmer, O., 2007, ApJ, 659, 997
993:
994: % The Queued Service Obseerving Project at CFHT
995: \bibitem[Martin et al. 2002]{pm02}
996: Martin, P., Savalle, R., Vermeulen, T., \& Shapiro, J.N. 2002,
997: Proc. SPIE, 4844, 74
998:
999: \bibitem[McArthur et al. 2004]{mca04}
1000: McArthur, B., et al. 2004, ApJ, 614, L81.
1001:
1002: % Five years of queue observing at the Gemini Telescopes.
1003: \bibitem[Puxley \& Jorgensen]{pj06}
1004: Puxley, P., \& Jorgensen, I. 2006, Proc. SPIE, 6270, 28
1005:
1006: % Early HET Science paper
1007: \bibitem[Ramsey 2005]{lwr05}
1008: Ramsey, L.W. 2005, AIP Conference Proc, 752, 3
1009:
1010: % MRS Paper
1011: \bibitem[Ramsey 2003]{lwr03}
1012: Ramsey, L.W., Engel, L.G., Sessions, N., DeFilppo, C., Graver, M., \&
1013: Mader, J. 2003, Proc. SPIE, 4841, 1036
1014:
1015: % Early HET paper
1016: \bibitem[Ramsey et al. 1998]{lwr98}
1017: Ramsey, L.W., \etal 1998, Proc. SPIE, 3352, 34
1018:
1019: % HET Paper
1020: \bibitem[Ramsey et al. 2007]{lwr07}
1021: Ramsey, L.W., et al. 2007, in preparation
1022:
1023: % Lessons learned from four years of queue-based flexible.. JCMT
1024: \bibitem[Robson 2002]{ib02}
1025: Robson, I. 2002, Proc. SPIE, 4844, 86
1026:
1027: % NOAO Queue-oberving Experiment on the WIYN Telescope.
1028: \bibitem[Saha et al. 2000]{as00}
1029: Saha, A., Harmer, D.L., Smith, P.S., \& Willmarth, D.W. 2000,
1030: Proc. SPIE, 4010, 25
1031:
1032: % A Northern Survey of Gamma-Ray Blazar Candidates.
1033: \bibitem[Sowards-Emmerd et al. 2005]{dse05}
1034: Sowards-Emmerd, D., Romani, R.W., Michelson, P.F., Healey, S.E.,
1035: \& Nolan, P.L. 2005, ApJ, 626, 95
1036:
1037: % HRS Paper
1038: \bibitem[Tull 1998]{rgt98}
1039: Tull, R.G., 1998, Proc. SPIE, 3355, 387
1040:
1041: \end{thebibliography}
1042:
1043: \clearpage
1044:
1045: \begin{deluxetable}{lr}
1046: \tablecaption{Science Operations FTEs}
1047: \tablehead{
1048: \colhead{Position}&
1049: \colhead{FTE}}
1050: \startdata
1051:
1052: Science Operations Supervisor & 0.5 \\
1053: Resident Astronomers & 3.5 \\
1054: Telescope Operators & 4.0 \\
1055: HET Scientist & 0.1 \\
1056:
1057: \enddata
1058: \tablecomments{The HET Scientist is not formally a part of the
1059: night operations team.}
1060: \end{deluxetable}
1061:
1062: \clearpage
1063:
1064: \begin{deluxetable}{lrc}
1065: \tablecaption{Priority Modifiers}
1066: \tablehead{
1067: \colhead{ Modifier }&
1068: \colhead{ Algorithm }&
1069: \colhead{Maximum } \\
1070: \colhead{ }&
1071: \colhead{ }&
1072: \colhead{Magnitude } }
1073: \startdata
1074:
1075: Priority 0 & If P0 then -1 &$-1.0$ \\
1076: Priority 4 & If P4 then +2 &$+2.0$ \\
1077: Object availability & Log[($N_{\rm visits}-N_{\rm request})/(f_{\rm min}*N_{\rm setups})]$ &$+2.4$ \\
1078: Object completeness & $ 0.6*(N_{\rm request} - N_{\rm done})/N_{\rm request}$ &$+0.6$ \\
1079: Partner share & $ 4*(F_{\rm current} - F_{\rm HET})/F_{\rm HET}$ & $-1.5$ \\
1080: Synoptic modifier & If Date $> $Date$_{\rm max}$ then & \\
1081: & $-0.6-0.15*$log[(Date$-$Date$_{\rm max})/(f_{\rm max}-f_{\rm min})]$ &$-1.0$ \\
1082: & else $-0.6*($Date$-$Date$_{\rm max})/(f_{\rm max}-f_{\rm min})$ & \\
1083:
1084: \enddata
1085: \tablecomments{\\
1086: $N_{\rm visits}$ is the number of visits left in the observing
1087: period, including any restrictions imposed by firm synoptic deadlines. \\
1088: $N_{\rm request}$ is the number of visits requested by the PI. \\
1089: $f_{\rm min}$ is the minimum length the PI prefers between visits. \\
1090: $f_{\rm max}$ is the maximum length the PI prefers between visits. \\
1091: $N_{\rm setups}$ is the number of exclusive setups that compete with the requested setup. \\
1092: $N_{\rm done}$ is the number of visits completed. \\
1093: $F_{\rm current}$ is the actual cumulative fractional share the partner has. \\
1094: $F_{\rm HET}$ is the cumulative fractional share the partner should have. \\
1095: Date is the current decimal date. \\
1096: Date$_{\rm max}$ is the last day in the preferred observing window.}
1097: \end{deluxetable}
1098:
1099: \clearpage
1100:
1101: \begin{figure}
1102: \epsscale{0.8}
1103: \plotone{f1.eps}
1104: \caption{Annulus observable by the HET,
1105: translated
1106: to hour angle and declination. The shaded region is the
1107: area observable by the HET. Targets with $decl < -10$ or
1108: $decl > +70$ are not observable by the HET. The longest tracks
1109: for the HET are at $dec = +63$ and are 170 minutes long.
1110: \label{fig1}}
1111: \end{figure}
1112:
1113: \clearpage
1114:
1115: \begin{figure}
1116: \epsscale{0.8}
1117: \plotone{f2.eps}
1118: \caption{Cumulative distribution of TAC allocations in 2006. Typically
1119: high-priority times are allocated in small increments to a number of programs.
1120: The lowest priority time, P4, is generally allocated in large increments to a
1121: few programs (e.g., 50\% of the P0 programs were allocated 2 hr or less).
1122: \label{fig2}}
1123: \end{figure}
1124:
1125: \clearpage
1126:
1127: \begin{figure}
1128: \epsscale{0.8}
1129: \plotone{f3.ps}
1130: \caption{Distribution of the completed and requested dark time visits
1131: for priority $1--3$ in the 2006-1 period. The open histogram represents
1132: the requests, while the filled histogram represents the completed visits.
1133: The connected triangles gives a model of expected completion rates for this period,
1134: based on our Web tools. The completed visits fall above the expected completion
1135: rates because many of the targets can be accessed in two ST windows (east
1136: and west).
1137: \label{fig3}}
1138: \end{figure}
1139:
1140: \clearpage
1141:
1142: \begin{figure}
1143: \epsscale{0.8}
1144: \plotone{f4.ps}
1145: \caption{
1146: Median individual program completion rates for all programs in
1147: 2004, 2005, and 2006.
1148: The completion rate for each program is based on the percentage
1149: of targets or TAC-allocated time successfully completed.
1150: In 2005, the HET staff adopted the modified-priority algorithm.
1151: In 2006, the User Committee made significant modifications to
1152: this algorithm.
1153: \label{fig4}}
1154: \end{figure}
1155:
1156: \clearpage
1157:
1158: \begin{figure}
1159: \epsscale{0.8}
1160: \plotone{f5.eps}
1161: \caption{
1162: Median individual program completion rates for all programs
1163: in 2005 and 2006 that used either
1164: the permanently mounted grism (LRS\_g1), the most requested
1165: interchangeable grism (LRS\_g2), or the less often requested
1166: interchangeable grisms (LRS\_e2 and LRS\_g3).
1167: The completion rate for each program is based on the percentage
1168: of targets or TAC-allocated time successfully completed.
1169: \label{fig5}}
1170: \end{figure}
1171:
1172: \end{document}
1173:
1174:
1175: