Cadence studies N. Regnault, P. Gris et al (many thanks to D. Rothchild and P. Yoachim, E. Rykoff, C. Stubbs and many others for past and future helpful discussions)
Metrics SN cosmology Survey uniformity Cosmology metrics Light pipeline & ubercal toy model, ● DETF figure of merit to evaluate ○ Using SNe to probe LSS ○ ⇓ If we can fit a ubercal solution ● After 1, 2... 3+ year(s) ○ How many well sampled SNe ? ● The quality of the ubercal ● Sampling quality ○ solution requirements from Photo-id Fisher matrix studies ○ & distance measurements Multiple fits ○ Redshift limit of SN survey ● How cadence can be tweaked ● z above which ○ To improve ubercal errors ○ measurement error > SN GAIA may help ! intrinsic dispersion (PCWG pop-up session on 2 Thursday)
Three cadence families White paper call Jan 2018 simulations (Tests of the feature scheduler) Baseline 2018a ● Kraken 2026 (new baseline) Minion ● ● Colossus 2665 Feature baseline ● ● Pontus 2002 (very wide WFD) Feature rolling ½ ● ● Released files come Colossus 2664 Feature rolling ⅔ ● ● with ditherings Colossus 2667 (1 visit / night) ● Pontus 2489 ● AltSched Kraken 2035 ● AltSched ● Mothra 2045 ● AltSched rolling ● Pontus 2502 ● Kraken 2036 ● No ditherings in released files AltSched’s own (added after the fact dithering scheme with MAF) http://astro-lsst-01.astro.washington.edu:8080/?runId=1 http://www.rothchild.me:8080 3
Three cadence families All give very similar metrics ● Number of visits per filter per healpix superpixel ○ Total survey depth ○ Average image quality ○ … ○ … but cadences differ very significantly w.r.t ● Median interval between visits for a given field ○ This is what Filter allocation strategy ○ matters for SN Integrated depth in a ~ 45 days time window ○ science ... ○ 4
Example 5
Example 6
Different filter allocation strategies AltSched rolling Minion 1016 Guarantees that each field observed in Keep observing in a given filter over long At least 2 bands during a given night durations 7
SN metrics 8
SN and cadence Cosmological impact of SNe is primarily a function of cadence (and ● calibration) Indeed, cadence determines the number and redshift ○ distribution of well measured SNe Using SNe Ia to constrain the Provided enough well DETF FoM measured SNe growth of structure (several 10 5 ) 9 DESC SRD (Howlett, et al, 2017)
Method 1. Superpixellize the sphere (nside=64) 2. Build a focal plane model 3. Play the cadence : for each exposure a. Project the focal plane on the sky b. Determine the healpixs observed (= center in the focal plane) c. Build an observing log for each pixel Global metrics SN specific metrics (depth, seeing, effective Use SN light curves as a cadence, per pixel) “probe” to determine cadence quality 10
Choice of a fiducial SN Sample of SNe with ● Median SN a. “Good LC sampling” to allow photo-ID ■ good distance measurement ■ b. Redshift-limited (i.e. all SNe Ia up to the redshift limit pass criterion (a) ⇒ size of SN redshift-limited cosmology sample is a good metric for LSST cadence Region of interest in (X1, Color) parameter space 11
Method (II) Use the LC of a fiducial SN as a probe of the cadence ● Method ● For each MJD, for each healpix pixel, in a given redshift bin ○ Generate a fiducial SN that peaks at MJD ■ Demand ■ At least 1 visit every 4 days, in either g,r,i or z ● > 1 visit in [-20, -10] days (restframe, g,r,i or z) ● > 1 visit in [+35, +45] days (restframe, g,r,i or z) ● σ(color) < 0.04 ● If pass, mark the pixel, ■ otherwise, go to the next lower z-bin ■ 12
Can be summarized as a movie SN peak date Integrated Instantaenous 13
Examples Minion Baseline 2018 AltSched AltSched rolling 14 http://supernovae.in2p3.fr/~nrl/lsst_sn_cadence/
Examples New baseline candidate Kraken 2026 Very wide WFD Altsched experiment By P. Gris, with guidance and help from D. Rothchild Very wide WFD AltSched very wide Pontus 2002 AltSched rolling 15 http://supernovae.in2p3.fr/~nrl/lsst_sn_cadence/
Size of WFD SN sample Fiducial SN = median SN Bug in the weather files (now corrected) 16
Size of WFD SN sample Fiducial SN = median SN 17
Size of WFD SN sample Fiducial SN = median SN 18
Size and redshift lever arm (WFD) Fiducial SN = median SN Number of well sampled SNe Good ! Fiducial SN = median SN “Under-optimal” Sample redshift limit 19
Size and redshift lever arm (WFD) Fiducial SN = median SN 20
Size and redshift lever arm (WFD) Fiducial SN = median SN 21
Ongoing experiments with SLAIR Tune the feature-based scheduler (a.k.a. SLAIR), to get an ● “Altsched-like” behavior Experiments carried out by P. Yoachim (thanks a lot !) ● Key points are: ● ○ Observe near the meridian ○ Observe fields in blocks that are repeated twice a night ○ Second visit in a different filter (use the filter changer!) ○ Make sure that any revisit occurs at least 2 days later 22
Size and redshift lever arm (WFD) Fiducial SN = median SN 23
Conclusions for SN metrics LSST can yield ~ 4 10 5 well sampled SN in the WFD ● New SN Ia science (?) beyond the HD ○ All types of fast transients ! ○ Ongoing effort to tune OpSim/SLAIR to obtain a AltSched-like ● Work for Tucson a& behavior New York workshop Ongoing effort to explore how we can improve on best cadences ● identified so far Wider survey ? ○ 80%-20% rolling ? ○ shallow survey to increase size of very low-z sample ? ○ More g & r-band, in order to go deeper ? ○ 24
Size and redshift lever arm (WFD) Fiducial SN = median SN Lots of help from D. Rothchild Help from P. Yoachim and Project very much appreciated 25
Survey uniformity 26
Survey uniformity Why do we (DESC) care about survey uniformity ? ● Flux calibration ● Primary flux reference(s) in specific locations on the sky ○ Flux scale must be transported on the full survey footprint ○ Essential for SN cosmology, target accuracy ~ 1 mmag ○ Specific calibration error modes on the sky ? ● may affect PZ determinations ○ at specific scales that are relevant for cosmology ? ○ 27
Questions Main question is ● How well can we transport the flux scale carried by a handful of ○ flux reference on the entire sky ? Technical questions are ● For a given cadence, can we solve the ubercal problem ? ○ Does interlacing DDF obs help improving the calibration ? ○ Are some dithering patterns significantly better than others ? ○ What is the impact of non-photometric sequences on solution ? ○ Are there specific error modes, at specific scales that have an ○ impact on the analyses ? Will adding GAIA help ? ○ 28
Ubercal toy model Measurement Uniformity map Calibrated mag Exposure ZP ~ 1 / month ? ~ 1 / week ? ~ / day ? Fitting simultaneously: ● Calibrated magnitudes ○ Calibration parameters (ZP + uniformity maps) ○ With constraints from ● PCWG GAIA pop-up session Primary references ○ Padmanabhan 2008 Thursday@8 am Schlafly et al, 2015 Future uniform star catalog (GAIA ?) ○ 29 Burke & Rykoff, 2017
Fast ubercal simulator Cadence Uncertainty budget (Fisher matrix) Observing Ubercal simulator conditions Ubercal fit Ubercal model External constraints (Primary standards, GAIA) Power spectrum Two simulators available so far: - Using healpixs (NR) - Using stars (F. Feinstein et al) 30
Example (with 2 years of survey) Many cadences New cadences not connected (no dithering applied yet) ℓ ~ 150 Error power spectrum Jan 2018 cadences (dithering applied) ℓ 31
Conclusion & work ahead Transform fast ubercal simulator into a MAF metrics ● Systematics checks of all cadences available ○ Add ditherings to the white paper call metrics ● No white paper cadence allows us to constrain the ubercal ○ model without the dithers Open questions ● How can we make sure that the cadence allow to constrain an ○ ubercal model with 1 yr of data only ? Understand how DDF observations help rigidifying solution ○ Location of fundamental flux standards ? In the DDFs ? ○ Calibration specific minisurvey ? ○ Additional dithered observations ? ■ Specific observations to transport CALSPEC -> DDF ? ■ 32
Backup slides 33
Different mean observing conditions 34
Different mean observing conditions 35
Global filter allocation 36
Size of WFD SN sample Fiducial SN = median SN AltSched very wide (P. Gris) Pontus 2489 37
Recommend
More recommend
