websky extragalactic models

[xzackli]

This commit contains a basic interface for accessing extragalactic models from websky, like in so_pysm_models.

• I've removed the mention of E/C coordinates and the special case of nside \leq 512 for now
• radio galaxies are exactly like CIB
• I haven't finished validating the SZ yet, will push it soon
• testing these with a www folder on the CITA cluster for now, should be simple to paste them into the models directory at NERSC when merging

There's an important detail for the SZ: the sky map is not band-limited due to the cuspy nature of the halos, and this has been shown to cause issues at the 10% level. It would be better if I could supply a map generated in real space with a fiducial 1 arcmin beam, and provide an interface for reconvolving a user-supplied beam. I don't think this will make it into DC-1.

[marcelo-alvarez]

Thanks a lot for this @xzackli!

There is some mitigation of the SZ maps not being band-limited through normalization of the overall profile to preserve the total y-distortion of each halo and be consistent with the pixel window function when halos are poorly resolved, but there are some residual effects as you describe that will hopefully be addressed in subsequent updates.

It's likely there will be some re-scaling of CIB in order to reduce the small-scale power slightly and optimize these maps for cluster finding, by @sriniraghunathan. @zonca do you suggest we do this here or merge without any scaling applied and document that step in a separate PR?

[zonca]

Thanks can you also specify which files I should copy over and in what target folder?

[zonca]

Can you also fix the codestyle errors so the unit tests can run?

[zonca]

It's likely there will be some re-scaling of CIB in order to reduce the small-scale power slightly and optimize these maps for cluster finding, by @sriniraghunathan. @zonca do you suggest we do this here or merge without any scaling applied and document that step in a separate PR?

either way is fine

[xzackli]

To get the maps, please copy the contents of for_nersc_www directory from my scratch to pysm-data/websky/0.4.

mkdir /global/project/projectdirs/cmb/www/pysm-data/websky
cp -r /global/cscratch1/sd/xzackli/for_nersc_www /global/project/projectdirs/cmb/www/pysm-data/websky/0.4

[xzackli]

I added the CMB map from so_pysm_models but let me know if pysm3 has advanced past that for the purpose of lensing the map with the potential.

[zonca]

I added the CMB map from so_pysm_models but let me know if pysm3 has advanced past that for the purpose of lensing the map with the potential.

I always assumed that we want to use the same CMB parameters of Websky for consistency, so we are stuck with that CMB. However, should we reconsider? @xzackli @marcelo-alvarez, do you have a clear answer to this? or know who else we can ask?

[xzackli]

To clarify, I was wondering if pysm3 now has the ability to lens a Gaussian unlensed CMB realisation with the websky potential. Anyway, for the purposes of this PR I think the existing CMB map is totally fine.

[marcelo-alvarez]

I added the CMB map from so_pysm_models but let me know if pysm3 has advanced past that for the purpose of lensing the map with the potential.

I always assumed that we want to use the same CMB parameters of Websky for consistency, so we are stuck with that CMB. However, should we reconsider? @xzackli @marcelo-alvarez, do you have a clear answer to this? or know who else we can ask?

I think it's fair to say we are stuck with some of the websky cosmo parameters, for consistency of lensing with the primary, but not necessarily stuck with the same realization of the primary or parameters whose variation doesn't significantly affect the lensing, such as the scalar to tensor ratio r.

[zonca]

To clarify, I was wondering if pysm3 now has the ability to lens a Gaussian unlensed CMB realisation with the websky potential. Anyway, for the purposes of this PR I think the existing CMB map is totally fine.

no, we don't have any lensing software inside PySM

[zonca]

I copied the input maps to the pysm-data area at NERSC:

https://portal.nersc.gov/project/cmb/pysm-data/websky/0.4/

[zonca]

here a Notebook on Gist (or Google Colab, see link at the top of the notebook) to test this branch: https://gist.github.com/zonca/3e1df26fa69d279ae4a9e05658d408a1#file-test-websky-cib-pysm-3-ipynb

[xzackli]

This PR lacks tests, but I believe it is otherwise ready for merging. Adding tests in the form of a set of downgraded (nside 512) wouldn't be too difficult though, is that desirable?

[zonca]

thanks @xzackli I am working on importing the old tests, I can take care of that.
Could you please import the docs from so-pysm-models and make sure they are updated for 0.4?

[zonca]

also we need to create PySM names for the components, I think we can do:

• cib1
• rg1
• ksz1
• tsz1

For CMB:

• c2: unlensed
• c3: lensed
• REMOVED c4: lensed with solar dip

[giuspugl]

thanks a lot for this PR!
i am a bit against c4, coadding multiple signals can be misleading for the user. perhaps it might be more useful to have a separated component dipole that can eventually be coadded to the other signals..
the other names are fine to me!

[xzackli]

The names look good to me!

[marcelo-alvarez]

What is c1? I would have thought it was unlensed primary, i.e. the numbers start counting from one for each of the different flavors of each component.

[zonca]

What is c1? I would have thought it was unlensed primary, i.e. the numbers start counting from one for each of the different flavors of each component.

c1 is the PySM 2 CMB https://pysm3.readthedocs.io/en/latest/models.html#cmb

[marcelo-alvarez]

I see, thanks. I like this naming convention and it should be easy to document.

[zonca]

@marcelo-alvarez @xzackli @giuspugl @sriniraghunathan please review and improve the documentation:

• for websky: https://github.com/galsci/pysm/pull/129/files#diff-3646d27a0c10f8bbdc1d821ef3c6cd590ccad3eaae80827513534ea92dfc102d
• for the models: https://github.com/galsci/pysm/pull/129/files#diff-d606c1bcd72689c8bb949dcdafe4bcd0420da5cbaa7daf9614f0757734cffaf7

No hurry, I don't need this for the simulations.

[zonca]

@xzackli the instrument model for S4 that I am using requires a map at 17.5 Ghz or below, could you please generate it for CIB and Radio Galaxies?

[xzackli]

Yes, I'm going to add some of my favorite radio bands (S, C, X, etc)

"1.4", "2.7", "5", "10.7", "14.5", "17.5"

[xzackli]

Quick update: CIB at these low freq bands is done, but I still need to do some radio abundance matching. I'll give a directory to copy once I'm set up.

[zonca]

thanks @xzackli
implementation is done so I merged the PR, please notify me when you have the data available.