Round Table 3:

​

Tackling the unknowns

Table B Scribe - Michael Wood-Vasey

Moderator - Chuck Claver

Emphasize systematics with moderate excursions + fiducial observations to provide improved leverage over models and thresholds for survey data.

Airmass; Stellar density (l, b); galaxy density (Abell clusters); ecliptic latitude; �Determine null tests of things that should be independent of � airmass, conditions, seeing: eg, PSF miscalibration, photometric calibration.�Degrade optics (figure, active control, louvers...); sensors (bias voltage, CTE, ...). Map out wavefront. Spin camera separate from optics.�Figure out how to calibrate y-band observations. Seasonal range? (hard)�Track a field for a night.�Exposure time as a function of wind speed. Open filter. Calibrate y ?�Auxiliary telescope + aux calibration + in-dome calibration; “Open” filter�Existing data: SDSS 82, HST, HST WD calibrators, Herschel(?), � DES, Abell clusters, VISTA.�What observations bootstrap doing science at survey start?�Stellar metallicity narrow-band survey (on DECam?) for dust calibration.

Table C Scribe - Rachel Mandelbaum

Moderator - Mike Jarvis

Question: Applying “null tests” in cosmic shear analyses is one way to identify residual systematic uncertainties. Which existing or new null tests will be most relevant for the full LSST data sample? Will these null tests have the required statistical sensitivity for cosmic shear measurements with LSST? (How do they scale?) What do we do when we see a 3-sigma effect, say, in a null test?

• last question should be rephrased: doesn’t matter if it’s 3-sigma if it’s < systematics requirements
• null test diagnostics:
• use on catalogs with multiple shear algorithms. This could help distinguish shear measurement method limitations from other stuff (like hardware-related systematics)
• look for subset of data where failures are even stronger, very informative
• simulations:
• should decide what quantities can be robustly tested by constructing null tests w/ sims (depends on what physics we know)
• MC simulations of data? Used in other contexts. Validation test: sim has to match the data
• identifying problematic exposures / images: don’t want to base on shear statistics but we need metrics to do this. Jackknife tests, since galaxies won’t be well detected on single exposures?
• change observing strategy in way that will stress specific null tests? (e.g., don’t change rotation angles, go out of focus, …) - try this during commissioning to test our understanding? Or use DDF
• Null tests that involve rerunning pipeline: e.g. compare shear estimates after estimating PSF with bright stars vs. with faint stars; change which exposures are used; etc. Increase sensitivity to certain systematics, selectively; but this is expensive.
• Do we need cross-validation (rather than a single training set) to reduce errors on null tests?
• Null tests with Euclid / WFIRST: s/g separation, shear calibration, … (modulo errors in other data)

​

Table D Scribe - Srikar Srinath

Moderator - Steve Ritz

Due to the planned short exposure times, atmospheric turbulence dominates the spatial variability of the LSST PSF, especially at small angular scales.

1. How do we determine an appropriate model for the atmosphere (how many layers?, what parameters?), and what information is or should be made available to constrain this model (MASS/DIMM, scintillation of guider images, wavefront sensors, ...)?
1. We need > 1 layer to avoid unrealistic wind speeds and probably fewer than 6 currently used. Wind layers sufficiently decorrelated, stratified, differ in power content
2. MASS/DIMM exists at Gemini-S. Other data: Sufficient telemetry probably exists.
3. Model-schmodel! What value does a model have?
2. What measurements are useful to obtain “statistically” for producing realistic simulations, and which are useful to obtain simultaneously with the survey to constrain PSF spatial variability in specific images?
• Ensembles of dense star field images - DES, Subaru will have them
3. Number of numbers needed to fully describe 15s atmospheric PSFs?
• Basically we want 2-D power spectrum at sub-arcminute scales and effect on PSF
• Do higher order moments of PSF variability matter? Yes! PSF is more than ellipticity.
• Best dataset we will have is LSST itself.

​

Table E Scribe - Will Dawson

Moderator - Jim Bosch

Q.: How can the variable seeing in the large number of LSST epochs be used to learn about the impacts of the atmosphere - e.g., for blended objects, inferring the PSF from galaxy images? Others?

What might we be able to constrain:

• (de)blending
• algorithmic / observing strategy
• Good seeing in u/g everywhere or small area (WFIRST/Euclid enough?)? How much does current cadence satisfy the needs?
• PSF
• Galaxies (in addition to stars) are the same in every image but PSF’s aren’t?
• For shapes do you gain by observing in variable seeing vs. always in the best seeing?
• To what degree can you carry over PSF information from exposure to exposure?
• Chromaticity/DCR/airmass
• Going to high air mass in good seeing conditions (needs to be entire survey).
• Photometry
• Same objects in variable seeing enable one to test how well PSF effects are being corrected.
• High resolution objects (e.g. strongly lensed arcs) of varying colors
• Star/Galaxy separation
• Calibrating brighter fatter
• same stars vs different stars (same stars take out the chromatic effects)

Table A Scribe - Satoshi Miyazaki

Moderator - Tony Tyson

Before, we start the discussions two things are mentioned

• We should try all the possible combination of observing parameters without prior bias
• Bad weather condition intentionally to see, for example, the external water vapor measurement is useful for calibrations

RT3A

Important Parameters ?

​

• Instrument Rotator Angle
• CCD characterization
• Mechanical flexure
• Exposure Time
• The output must scale to the exposure time , basic but need to check first. SN goes up with sqrt(Nphoton) ?
• Sky level
• Use moon light to change the sky level

RT3A

Field appropriate for commissioning…

• DES Supernova field
• COSMOS (HST/ACS imaging, a lot of spec-z) useful for HSC as well
• HDF-S

​

• Low galactic field / outskirt of globular clusters
• Denser stellar field to characterize higher spatial frequency error

​

RT3A

Some other parameters that might be useful: (may be mostly engineering)

- Azimuth/Elevation w.r.t. the moon … scattered light

- Focus position … test of optics modeling

- Tilt/de-center of focal plane … test of optics modeling

- Elevation …

(Lensing analysis is the most strict engineering diagnostic)