The Gemini Time Allocation Process --- A New Game

Todd Boroson

For the first time, the recent series of TAC meetings included proposals for Gemini North. The manner in which time is assigned and observations carried out on the Gemini telescopes is somewhat different from that of the NOAO-operated telescopes, and it was clear that some of the implications of that were not appreciated by proposers. First, let's acknowledge that many things were atypical about this semester. We had only 17.5 nights to give out. The 78 proposals that we received oversubscribed this time by a factor of 6.4. Instruments available were limited to two--University of Hawaii's Hokupa'a/QUIRC and University of Florida's OSCIR. All observations were to be obtained through queue scheduling by Gemini staff. Predictions about sensitivities and efficiencies were only guesses.

However, even with all these special conditions, there are several aspects, particularly of the Gemini queue, that determine who gets data and who does not. Here's how the process works.

1. Proposers write and submit proposals using the forms and process set up by each country's national Gemini office or national TAC. For the US, this is the standard NOAO LaTeX proposal form that is available on the Web and submitted electronically to NOAO. Information about the capabilities Gemini is offering and expected performance comes from the Gemini Web site (mirrored by NOAO for US astronomers). In addition to the instrument desired, proposers must indicate what quality of observing conditions they need. For this first semester, the conditions were limited to image quality and sky transparency, and the choices were limited to one or two for each.

2. Proposals undergo a technical review by US Gemini Program (USGP) scientific staff at NOAO. They are evaluated scientifically by the NOAO TAC panels (membership of which is listed on our Web site) and merged into a ranked list based on scientific assessment.

3. A US Gemini merging TAC (with representatives from the various discipline panels) then goes through the list in detail. Our 17.5 nights are divided into two, half for each instrument, and are further subdivided into the different bins defined by observing conditions. Once the merging TAC is satisfied that the proposals are in the proper ranked order, these bins are filled by going down the ranked list. When a bin is filled, a proposal that needs those conditions cannot go to the telescope unless it can be put into a bin with better conditions. To give us some latitude, the bins were initially overfilled by a factor of two.

4. The resulting ranked list of proposals (about twice as many as were needed to subscribe the US time) is sent to Gemini. The Gemini North operations team takes these lists from the seven partners plus host (Hawaii) and Gemini scientific staff, and merges them into a single ordered list of programs. This list is filled top to bottom using a scheme that allows approximate balance of the partner shares to be maintained.

5. The International Time Assignment Committee (ITAC), including representation from each country, meets to discuss the merged queue of programs. The main charge to this committee is to deal with conflicts, such as proposals that went to more than one country (typically, the "cost" is split among the countries involved) or identical proposals from two or more countries (typically, an attempt is made to form a collaboration). The ITAC also decides how to deal with proposals for which a Gemini technical review has identified a problem. Finally, the ITAC decides how to divide the list into "bands." The bands are meant to be ranges of programs that can be considered of equal scientific priority, so that the staff executing the programs have a simple way to pick the best observation to make at any given time from a pool of reasonable size. This time, the ITAC divided the queue into three, roughly equal bands. The final list is forwarded to the Gemini Director for approval.

6. For each approved program, a contact scientist at Gemini is designated. The contact scientist works with the PI to ensure a complete understanding of the observations desired. As the semester proceeds, the staff execute the observations, attempting to complete all the Band 1 observations before the Band 2 observations are started. At any given decision point, weight will be given to the best match between program and conditions, completing programs that have been started, and maintaining the balance of partner shares. Partner shares can only be expected to balance over two to three semesters.

In working through this process from beginning to end, it became clear that the constraints on conditions play a major role in determining which programs get into the queue. Proposers should understand that the tighter the constraints they put on the quality of the conditions for their program, the less time is available for that program. In the most recent round, several proposals requested more than 100% of the time that would be available to US programs with the conditions specified! In the merging TAC, we had to skip over a number of excellent programs because the conditions they requested were already used up by higher ranked programs.

Why don't we select programs purely on the basis of scientific merit and use however much good quality time as there is? We have agreed with the other Gemini partners that we will share the time in an equitable way. We won't try to load the queue with programs that will use up all the best time, but will limit our request of the best time to the same proportion as we get of the total time. Alternatively, we could put a lot of good quality time proposals at the bottom of the queue, with the idea that these will get executed if there is an excess of good quality time in a given semester. Our experience with the WIYN queue convinced us that this is a bad idea. Leaving programs in the queue all semester and never executing them results in (justifiably) upset proposers. If we run out of programs to execute midway through the semester, we can always go back to our list and contact proposers to see if they are still interested in getting data--the usual response is, "Are you kidding?"

And so the advice that comes from the results of this first semester is:

• Do the math. Divide the number of nights you are asking for by the frequency of the conditions you require to calculate an "equivalent nights requested." See if this is a rational request.

• Think carefully about the data quality you need. Make sure that you specify conditions that will allow you to get that data quality, but not better. Read the instrument and telescope information carefully to ascertain what that is.

• Don't be discouraged. Recognize that this semester was a special case that made things a lot harder than they will be in the future.