Forecasting

From CMB-S4 wiki
Jump to: navigation, search

Forecasting non-r parameters (for the DSR)

To get started for large-area forecasts for the DSR, we will use the following. We expect the noise tiger team will provide updates in a month or so (~end of August).

We will use numbers from the CDT (Table 1 from https://www.nsf.gov/mps/ast/aaac/cmb_s4/report/CMBS4_final_report_NL.pdf).

That means (via Tom Crawford and Matt Hasselfield):

large-area survey map depths
frequencies 40 90 150 220 270
map depths (T/uK-arcmin) 5.6 1.35 1.81 9.1 17.1

area of sky: 40% of the full sky

For beams, baseline would be to assume a diffraction-limited 6m telescope. To be concrete, let's say a 1.4' beam at 150 GHz that then scales as (1/freq).

To include the effects of atmosphere and 1/f, the recommendation is to use the ell scaling at https://cmb-s4.org/CMB-S4workshops/index.php/Update_on_Neff_Forecasts. That means putting in a knee at ell=3400 in TT and at 340 in EE (and presumably BB): NlTT= N0TT(1+ (l /3400)-4.7) NlEE= N0EE(1+ (l /340)-4.7)

It isn't immediately clear how these should scale with frequency, though, so it is also fine to go with white noise for the initial go-around.

Forecasting non-r parameters (for the Science Book v1)

The expectation is that these forecasts will be done mainly using Fisher forecasts for the Science Book, with prudent use of frequencies and ell-range to avoid being over-optimistic about the impact of Galactic and extragalactic foregrounds and noise from the ground.


Parameters

Names of people responsible for generating forecasts in brackets. Sign yourself up and we can divide up tasks when we speak if there are multiple people.

  • LCDM 6-parameters (all)


Inflation section


Neutrino section


Dark matter and dark energy section


BSM Physics section

Here is place-holder for some suggested fiducial parameters and step-sizes, can be different though. Suggested new one: File:Params steps v2.txt (Original one here: File:Params steps.pdf)


Fisher codes

With info on what data they can handle and who is available to run them during next two months

  • Errard/Feeney code -unlensed TT, EE, TE and BB, plus deflection dd. Can do iterative EBEB, CIB or LSS delensing, and foreground removal on a statistical level. Current white noise, but simple to extend (Stephen,Josquin) -- web interface accessible at : http://portal.nersc.gov/project/mp107/index.html
  • Allison et al code - lensed TT/TE/EE plus kk plus BAO, can take in non-white N_ell but no explicit FG handling (Erminia Calabrese, Danielle Leonard, Jo Dunkley)
  • Extension of Allison et al code for axions (Renee Hlozek)
  • Stony Brook code - TT/TE/EE/kk with iterative delensing plus BAO, can take in non-white N_ell but no explicit FG handling, being expanded to include LSST shear and cluster counts with halo lensing (Mat Madhavacheril, Neelima Sehgal, Nam Nguyen)
  • de Bernardis code - includes kSZ likelihood (Francesco de Bernardis)
  • Manzotti code- similar to Allison et al and already partially checked against that code; used for http://arxiv.org/abs/1512.02654 (Alessandro Manzotti)
  • CITA code - includes delensing of spectra to all orders; forecasts of delensed covariances are in progress (Joel Meyers, Alex van Engelen, Dan Green)
  • who has a tSZ likelihood? (David Alonso can do N(M,z) and tSZ fluxes. Who has code to push through to parameters?)
  • other codes and people?



Settings

Nominal:

  • S4 TT/TE/EE/kk over 40% of sky, 30<ell < lmax
  • Planck TT/TE/EE from 30<l<2500 over additional 20% of sky. Use these 'Planck-pol' specs for noise:File:Planck pol.pdf
  • Planck TT at l<30 over 80% of sky
  • Tau prior 0.06+-0.01


  • lmax(TT)=3000 unless explicit foreground cleaning is done in code for kSZ etc
  • lmax(TE,EE)=5000 unless explicit foreground cleaning done in code
  • kk reconstructed from 30<l<lmax using MV estimate


  • quadratic estimator for lensing, ideally with iterative delensing
  • Gaussian likelihood neglecting T/E/k covariance is ok, but non-Gaussian better (do any codes have full lensed T/E/B/k covariance?)
  • non-linear power spectrum for kk, e.g. ok to use halofit in CAMB

Extensions:

  • cluster masses calibrated with LSST lensing or CMB halo lensing (need to refine what that means)

Specs

Nominal test case

  • Single channel (e.g. 150 GHz) at 1 uK/amin in T and 1.4uK/amin in P, 3 arcmin resolution.
  • White noise, no FG inflation
  • Useful if your code can spit out errors as function of noise level in 1-10 uK/arcmin range and resolution in range 1-10 arcmin.

Next steps

  • Define multiple frequencies of the survey
  • Decide if an N_ell that captures non-white-noise is necessary
  • Define residual FG level if any