The SCOC is finalizing the Phase 2 recommendations, which are a substantial narrowing of the parameter space of the survey strategy. We are presenting four simulations that attempt to include the currently formulated recommendations for the initial survey strategy based on the SCOC Phase 1 and Phase 2 recommendations, with few small variations on implementations
(see the Phase 1 recommendations in PSTN-053; the draft Phase 2 recommendations are temporarily summarized in this slidedeck).
The four simulations here (called “v2.99” to emphasize that we are so, so close to v3.0 but these are still drafts) are:
“draft_connected_v2.99_10yrs.db”
This is a 2 band 0.9 strength rolling cadence (but starts after a full 1.5 years, so not “early”).
There are long gap triplets every 6 nights, and revisits within 0-2 hours are suppressed (at the lowest strength).
The twilight NEO microsurvey is added starting year 1, using riz filters and 4 repeats per night, and active every 4th night – the micro survey has been added in only morning twilight, and only taking part of the twilight time, which is less than the SCOC recommended. (this will be an area of further testing and evaluation).
The galactic plane footprint is based on the TVS/SLWMV priority map, modified to reduce isolated areas. (the filter balance and details of this footprint will be an area of further testing and evaluation).
The galactic plane filter balance is very red (the filter balance is likely to be tweaked slightly back toward more g band).
The NES weight has been lowered slightly from baseline.
The DDF visits make up about 7.5% of the total visits, with COSMOS acquiring ~double depth (10yr depth by year 3, then continuing at the standard cadence for the remainder of the survey). The DDF strategy is a fairly simple pre-scheduled season. (the DDF sequences are likely to be modified from the simple flat cadence used here).
MAF outputs:
Preview - the footprint looks familiar but slightly different.
“dd6_v2.99_10yrs.db”
This is similar to the above simulation, but reduces the fraction of time spent on the DDFs to 5% total. The reduction is implemented uniformly across all of the DDFs. This does not follow a particular SCOC recommendation, but was intended to help illustrate some of the areas of uncertainty in the recommendations.
MAF outputs:
“low_gp_v2.99_10yrs.db”
Again, similar to the draft_connected simulation, but the weight of the footprint within the galactic plane WFD area is strongly reduced, which reduces the number of visits in the galactic plane ‘high cadence’ (WFD-level) area. This does not follow a particular SCOC recommendation, but was intended to be a placeholder to indicate uncertainty in the galactic plane footprint.
MAF outputs:
“roll_early_v2.99_10yrs.db”
This is similar to the draft_connected simulation, except in some areas related to cadence.
The 2-band rolling cadence starts “early”. In addition the repeat weight is stronger. These changes do follow SCOC recommendations, but are also intended to illustrate areas of uncertainty.
MAF outputs:
This notebook contains a high level summary comparison of metrics between these runs (and comparing against the baselines from v2.0, v2.1, and v2.2). (and nbviewer is an alternate view option, if the original at GitHub keeps failing to render).
Is there any paritcular reason by no_twilight_neo_v2.2_10yrs is missing from summary_2022_10_25.csv? The v2.2 baseline has no Vatira/ 'Ayló’chaxnim detections, so I was trying to find a non-v2.0 simulation to compare to.
The v2.2 baseline has some 'Ayló’chaxnim detections (at least in the 1 quad category). I didn’t add no_twilight_neo_v2.2_10yrs to the big csv file because it was intended to be a stand-in for baseline_v2.2 before we had baseline_v2.2 available (since we made the twilight neo runs in v2.2 before we had sorted out the baseline_v2.2 parameters fully). I think the major difference between baseline_v2.2 and no_twilight_neo_v2.2 would be that no_twilight_neo probably doesn’t have pre-scheduled DDFS. I can have a look at what metrics we ran with that (I’m sure the SSO ones were the same, but the remainder of the metrics may be difference) and add it in.
Thanks. That makes sense. Maybe it should be in there for comparison to the neo twilight simulations for posterity, but again as it is not a baseline others won’t confuse it. Given SSP is only the moving object search tool currnetly that will be guaranteed to run on the LSST data, I am looking at the SSP metrics (3 pairs in 14 days), and the current v2.20 baseline has zero. Good to be reminded of the differences with no_twilight_neo_v2.2_10yrs . The v2.99 do better than no_twilight_neo_v2.2_10yrs, much better than the 2.20 baseline in this context.
It seems to me that these cadences are overcompensating on the 2-14 hour time gaps. The metric went up by a factor of 2. It seems to me that some of these visits could be redirected to increase the number the number of visits in the wide-fast-deep that dropped from v2.2 to v2.99
Is this something about the combined suppress visits and long gaps combining, and one of these could be relaxed (like how often log gaps run)?
The Tgaps metric went up by a factor of 2 because the initial value was incredibly small. The long gap visits mostly happen in the WFD, so changing how often it runs shouldn’t change the total number of WFD observations.
Right - we picked a relatively small “nights off” value (nights off 6) which showed in the previous long gaps simulations only a very small impact on the overall number of visits. I don’t think this is reducing number of visits in the WFD by a significant amount (<1%).
The drop in visits in (the low-dust) WFD is being primarily driven by the increased time in DDFs (~2.5%), the addition of the twilight NEO micro survey (~2.5%), and to some extent, a larger area in the galactic plane WFD (maybe ~1/1.5%?). None of these visits are entirely redirectable to “anywhere on the sky”, although they can be within a limited RA range. Overall we’re seeing a decrease of 5-6% in the median number of visits in the WFD (depending on which run you compare to), which is consistent with the impacts we saw with each of those individual changes in previous runs.
For these kinds of comparisons, I usually find the median number of visits per pointing (evaluating the ‘best’ 18,000 sq deg) tends to scale much more linearly as visits are redirected to ‘other than WFD’ purposes, because the “typical” number of visits scales relatively slowly; the same is mostly true of the minimum number of visits per pointing. However, the area which receives at least X visits tends to show more dramatic changes, because we do cover the WFD area fairly uniformly (not entirely, but mostly) and so when one point on the sky falls below the threshold X, large chunks of sky usually follows fairly quickly.
Another question I was trying to look into the Vatiras (Inner Venus Objects/ 'Ayló’chaxnims) - the increase in the discovery metrics for them jumped significantly in the new v2.2 twilight simulations. It’s much more than the v2.99 simulations. I am not sure why it’s different. I am missing something, but I am not sure what’s different between v2.99 and the new v2.0 twilight simulations.
I realized that the twilight_neo micro survey in v2.99 is only using morning twilight, and not morning + evening (as it was in v2.2). This was one of the suggestions from the SCOC as a possibility, but wasn’t what I had originally thought was implemented, so hadn’t brought it up earlier. This definitely has an impact on the discovery rate. None of the v2.2 runs used morning twilight only.
I’ll communicate this to the SSSC. Thanks for looking into it. Is this change because evening and morning twilight would be ~5% of the time and there’s not 5% to give to that?
To be honest, I am not sure.
I think this is something (along with swapping the u/z/y filters with lunar cycle instead of just u/z and changing the DDF sequences to something like either quads or the accordion cadence, as well as tweaks to the filter balance in the galactic plane and south celestial pole regions) that we will have to add some extra work into checking before we can finalize a v3 simulation.
There was mention last week that there was a bonus v2.99 simulation with a twilight near-Sun survey that was both morning and evening twilight. I am not sure what the name of that simulation is. Are the metrics from that simulation available? Many Thanks.
The bonus simulations that were run at the time of the workshop still had the near-sun twilight survey only operating in morning, rather than both morning and evening.
I’ve just run (over the weekend) one which does take place in morning and evening, at the same level as twi_neo_brightest_repeat4_riz_np4.
I can send that along to you, or rather maybe what I should do is send along an example of the metric output – but I will say, I think we still have work to do as the time needs some redistribution from where it’s ended up when running near-sun twilight at both evening and morning. Mostly, I would like to verify that we get the expected 'Aylò’chaxnim performance when running like this, while also swapping z and y band, and taking the rest of the visits for other parts of the survey (DDF running at 7.5%,etc).
