Deep drilling whitepapers

opsim
special-surveys
Tags: #<Tag:0x00007f7f6f0a3ef8> #<Tag:0x00007f7f6f0a3d40>

(Lynne Jones) #1

I keep losing the links to the deep drilling white papers that were submitted by each of the science collaborations in 2011, and that played a significant role in shaping the current deep drilling strategy in Opsim.
I also find that other people can’t find these white papers.
The link is here!
https://docushare.lsst.org/docushare/dsweb/View/Collection-2279

There was a series of pages on the (long languishing but still online) science wiki:
http://www.lsstcorp.org/sciencewiki/index.php?title=Deep_Drilling_Fields
but this material should be moved somewhere publicly viewable.

A synthesis of the requests from these DD proposals comes down basically to:

  • 5 ‘extragalactic’ fields (most likely the 4 extragalactic fields ELAIS S1, XMM-LSS, Extended Chandra Deep Field-South, and COSMOS, along with 1 more undecided pointing) observed as follows:
    265 sequences (on different nights) of 20 g exposures, 40 r exposures, 40 i exposures, 52 z exposures and 40 y exposures (~68 minutes per sequence), spread over 10 years. This corresponds roughly to observations taken every 3-4 nights, if the ‘observing season’ is assumed to be 4 months long for each field.+
    185 sequences of 40 u band exposures, spread over 10 years. (no time constraints relative to grizy observations).
    This would achieve final coadded limiting magnitudes of approximately u=28.5, g=28.7, r=28.9, i=28.4, z=28.0, and y4=27.0 (with varying single-night depths depending on lunar cycle, ranging from approximately g=26.2-24.4, r=26.1-25.1, i=25.6-25.1, z=25-24.8, y=24.1-24.0).
    These fields require about 335 hours per field in real-time observing.
    Observations take place over all 10 years of the survey, and require 1675 hours total (all 5 fields).
    +[Note camera limitation of 5 filters at a time; during the time around new moon the u band filter will be in the camera and the grizy sequence will be somewhat interrupted.]

  • 5 (different) extragalactic fields (unchosen but with high ecliptic latitude, high galactic latitude, low extinction, and no foreground clusters) observed as follows:
    265 sequences (on different nights) of 20 g exposures, 40 r exposures, 40 i exposures, 52 z exposures and 40 y exposures (~68 minutes per sequence), spread over 10 years, as above.
    57 sequences of 40 u band exposures, spread over 10 years.
    This would achieve final coadded limiting magnitudes of ~ u=28.0, g=28.7, r=28.9, i=28.4, z=28.0, and y4=27.0 (similar single night depths to other 5 extragalactic fields).
    These fields require about 309.4 hours per field in real-time observing.
    Observations take place over all 10 years of the survey, and require 1547 hrs total (all 5 fields).

  • 3 Milky Way/Local Volume fields (South galactic pole & Blanco 1 (RA 00:53, Dec -26:33), Galactic anticenter (RA 04:02, Dec 11:01), and one of three proposed open clusters (11:05, -58:44; 08:03, -28:01; 17:25, -49:58), observed as follows:
    30 sequences (on consecutive nights), consisting of 3 subsequences of [2 y exposures, 2 i exposures, 2 z exposures], subsequences separated by ~30min-1hrs (this set of exposures takes place once per field for 30 days).
    133 sequences of 30 g exposures, 2 r exposures, 70 i exposures (within 1 week of new moon), spread over 2.5 years
    113 sequences of 70 i exposures, spread over 2.5 years.
    This would reach final coadded depths of ~ g=29, i=28.9, r=27, z=25.6, y4=24.9. (although in z and y, the individual image depths are the relevant parameters and are equivalent to the main survey).
    These fields require about 135.8 hrs per field in real-time observing.
    Observations take place over ~2.5 years per field, but fields could be spread over 10 years of survey if needed, and require 407.3 hrs total (all 3 fields).

  • 6 Transients/Variable stars fields (4 pointings toward LMC, 1 at SMC & 47 Tuc, 1 at IC 4651) observed as follows:
    1 hour of continuous g band exposures
    sequences of [2 g exposures, 2 r exposures] spread over next 3 nights at various time intervals, for 7 hours of total real-time exposures
    Within the next few months -
    1 hour of continuous r band exposures
    followed by sequences of [2 g exposures, 2 r exposures] spread over next 3 nights for 7 hours of total real-time
    The pattern should be repeated again at a later time for period validation. All observations should occur within 1 week of new moon.
    This would reach coadded depths of ~ g=28.3, r=28.1 after all 4 3-night cycles, although the individual exposure depths are similar to the main survey (and are the relevant quantity for most of the transient/variable phenomena).
    These fields require about 32 hours real-time per field.
    Observations take place over ~1 year for each field, but fields could be spread over 10 years of survey if needed, and require 192 hrs total (all 6 fields).

  • 9 solar system fields (9 pointings in a 3x3 grid centered on 19:35:04, -21:38:32 where Jupiter and Neptune Trojans will coincide, in year 2022), observed as follows:
    8 sequences of 283 exposures in r band (continuous) on different nights, epochs specified as 2 nights 1.5 months before opposition, 2 nights at opposition, 2 nights 1.5 months after opposition, and 2 nights 1 year later (approximate timing). Each of the ‘2 night’ sets should be consecutive or at most a few nights apart. (Total of 8 epochs of observations, with 283 exp in each epoch).
    This would reach a total coadded depth of ~r=28.1, with each individual night being coadded to ~ r=27.0 (which is the relevant quantity for the solar system detections).
    These fields require about 11.3 hours real-time per field.
    Observations take place over 1 year, and require 101.7 hrs total (all 9 fields).

The combination of the above produces an oversubscription rate of about 2 compared to then-current time invested in DD (in opsim 3.61). The total time required above is 3923 hours, and the time currently programmed as DDF’s in Opsim 3.61 is about 1972.5 hours. It also gives us about 28 various deep drilling pointings to satisfy all the science requirements, although not all of these have similar observational sequences.


LSST2017: Deep Drilling & Mini-surveys breakout planning
(Lynne Jones) #2

The current set of opsim runs only includes the first of these bullet points: 5 extragalactic DD fields, of which 4 are the pre-selected DD fields.

The pre-selected fields were announced in 2012:
http://www.lsst.org/News/enews/deep-drilling-201202.html


(Bethwillman) #3

Lynne - This is a great summary of otherwise hard-to-find (and often asked about) information.

One question - What % of the total 10-year survey observing time is taken by the 4 pre-selected DD fields? I’m very interested in, and we are discussing with the PST and the SAC, (i) how to settle on a total % time to allocate to DD fields, and (ii) how to build upon this past work, while also giving the full LSST data rights community the opportunity to submit proposals for DDFs before the final selection.


(Lynne Jones) #4

Okay - so considering the DD fields included in minion_1016 (our soon-to-be baseline run):

| fieldID  |   fieldRA    | fieldDec   |  Name
| -------- |  ----------- | --------   | ---------
| 290      |  349.4       | -63.3      |    (not pre-selected)
| 744      |   0.00       | -45.5      | Elias S1
| 1427     |   53.0       | -27.4      | Extended Chandra Deep Field-South
| 2412     |   34.4       |  -5.1      | XMM-LSS
| 2786     |   150.4      |   2.8      | COSMOS

All together, the pre-selected fields receive about 3.7% of the total number of visits in the run, in sequences as requested.

Timewise, the four fields used about 991 hours (about 99 nights if we assume an average of 10 hrs per night) in sequences as requested, if I count as separate ‘sequences’ any DD observation of a given field separated by less than 4 minutes.
The overall survey got about 27779 hours of ‘uptime’, over 3026 open nights, or an average of 9 hours per night (counting anything more than a 2 hour gap as ‘downtime’).

This means that the four DD fields actually only took about 3.5% of the available time.

@ivezic was right again … the number of visits was pretty close to the amount of time available, even though these sequences require a lot of filter changing. I did not expect that, due to the filter changes – but it’s worth noting that the DD fields here only used about 248 hours per field, compared to my estimate of 335 hours per field above.

I think the difference is that I assumed fewer u band sequences and more long grizy sequences, while in minion_1016, the sequences are often split into shorter blocks. Out of the 367 sequences in total, over half - 199 - of them are in blocks shorter than 15 minutes. These include sequences in u band (which we would expect to be in short blocks, because they occur during dark time and are thus just ‘u’ band observations) as well as 136 sequences in other filters. This means that either the sequences get started, then stopped due to the field setting or bad weather, or something else (maybe the filter change took longer than 4 minutes in some changes or the rotator had to slew or something else??). It does not seem that they restart again later in the night, because if I make the allowable ‘interval’ before going back to the DD field longer, the number of DD sequences does not change.
If the problem is that the DD fields are setting - and thus the sequences are aborted after 15 minutes - then when opsim adds better ‘lookahead’ (to avoid this problem), then the amount of time required by DD ought to increase (and the acceptable and desired number of sequences and number of visits should increase as well).

I made a notebook with this analysis (we’ll add it to the list to fold into MAF at some point too) and put it online at:


(Willclarkson) #5

Hi Lynne - do you know if the DDF whitepapers referenced in your original post (at the link below) ever did get moved somewhere publicly viewable?

https://lsstcorp.org/content/whitepapers32012

Thanks!

Will cc @jgizis


(Lynne Jones) #6

I apparently can’t edit my original post anymore (I’ll see if I can get @jsick to fix that), but the updated link is
https://docushare.lsstcorp.org/docushare/dsweb/View/Collection-2279
for all the DD white papers.


(Lynne Jones) #7

White papers are also available here: https://project.lsst.org/content/whitepapers32012


(Tim Jenness) #8

I wish that web page linked to the Docushare versions at https://docushare.lsst.org/docushare/dsweb/View/Collection-2279 – it’s nice for it to be pointing to a more permanent location that is easier to cite.


(Lynne Jones) #9

We will need to make some updates on the opsim related pages (and presumably also DD pages) on lsst.org, so I will keep that in mind when we do that.


(federica bianco, NYU) #10

Hi @ljones - the link at the beginning of this thread https://lsstcorp.org/content/whitepapers3201223 is dead. This is alive https://project.lsst.org/content/whitepapers32012 - can we substitute the link in the top thread or create a redirect from it? The link is in the arxiv version of the white cadence paper (note 3 page 28) and @drphilmarshall can we change it in the current version of the paper (i’ll post an issue on that)


(Lynne Jones) #11

Unfortunately, community isn’t built that way so much and I can’t edit the top post anymore.
I don’t think we can make a redirect from it, because of lsstcorp.org being decommissioned.
If people read the whole thread, there are updates below – but yeah, I agree it’s hard to find.


(Ranpal GIll) #12

https://project.lsst.org/content/whitepapers32012 has been updated to point to docushare and i updated the Lynne’s original post but the science wiki link is still dead, i’m trying to find where that content went to…