Leo P. Singer (2467bf0f) at 29 Mar 00:13
Add additional RAVEN citation; fixes #347
... and 1 more commit
The BNS sensitive volumes for Virgo (55 Mpc) and KAGRA (still struggling due to EQ) need to be updated. The use of Virgo for PE/Skymaps could also be mentioned
https://emfollow.docs.ligo.org/userguide/capabilities.html should be updated accordingly
Leo P. Singer (882a9b59) at 29 Mar 00:12
Resolve "Virgo and Kagra sensitivities"
Closes #354
Closes #354
Let me rephrase. I think that these are small effects compared to the gross question that most astronomers have, of whether there is the remotest possibility of any BNS mergers in O4. While these small effects would probably be important to investigate in a paper, I think it's too much detail for the User Guide.
I want to stress again that the User Guide is not a journal article. It's a scientific outreach product for users of our data.
This MR is a valuable contribution and addresses a timely question, but I think that by making it shorter and more to the point we can increase its impact.
@michael-coughlin, would you please take a look at this and see if there is anything that we can do to tighten up this section? My critique consists of the following suggestions:
I think it looks alright, except for at least one remaining grammatical error. @michael-coughlin, would you please finish reviewing this or delegate it to someone else? I will be fairly busy for the next 9 days due to APS and I might not be able to respond as promptly as @vinaya.valsan requires.
Thank you!
This is an incomplete sentence. Please proofread.
I've taken a pretty deep dive on this, and it's going to require development of some new tools for IPN. See #772 (comment 980745).
I contacted the IPN developers and took a deep dive into why their sky maps are so large. TL;DR: there are straightforward techniques to generate small, conservatively sampled localizations, but those techniques are not currently captured in any HEALPix Python library. Here's the full text of the message that I sent them:
I see why your sky maps are so large. There are two main problems:
- For each of your Gaussian distributions (the Gaussian annulus and the Gaussian circle), you are starting by creating a HEALPix mesh that accurately describes the boundary of the 1-sigma level. However, you don’t want to concentrate high-resolution pixels on that border; you want to concentrate them in the regions of highest probability (i.e., along the middle of the annulus, and at the center of the circle) and you want the resolution to fall off as the probability density drops to zero.
- The maximum nside is much higher than it needs to be. If you concentrate the highest resolution a the maximum of the distribution, then it should be enough to have about 4 pixels per standard deviation within the 1-sigma region, dropping to the base resolution of nside=1 outside of the 3-sigma region.
I’m attaching some example images and FITS files illustrating this sampling scheme. (I changed the sigma of the annulus from 1 to 10 degrees so it’s easier to see the grid, but this all works fine with the 1 degree annulus too).
Note that in the combined (multiplied) sky map, there are artifacts due to using nearest-neighbor sampling to interpolate the low-resolution pixels in the tails of the circular distribution. These artifacts would not be visible if you are combining IPN annuli with each other, but they may sometimes be visible when you are combining IPN annuli with error cones from individual satellites. I recommend using linear interpolation, which results in a combined sky map that is free of visible artifacts.
The final combined FITS file occupies about 62 KB. (With the sigma of the annulus set to 1 degree, it occupies about 360 KB.)
Neither mhealpy nor mocpy directly support the above sampling scheme. Also, neither supports linear interpolation on a multiresolution grid. I have written code for both, but I’m still working on cleaning it up. It is not clear to me whether it should live in healpy, astropy-healpix, mhealpy, ligo.skymap, or a new library.
Eric, in the coming days and weeks, I would like your help to facilitate a discussion with some of the authors and users of the main Python implementations of HEALPix to discuss these issues.
01-circle.multiorder.fits 02-annulus.multiorder.fits 03-combined.multiorder.fits 04-combined-linear-interpolation.multiorder.fits
Leo P. Singer (0bd327b4) at 27 Mar 20:57
WIP
The ligo.skymap.plot tests started failing on macOS arm64 recently. Check to see if this was due to the recent release of astropy 6.0.0.
The ligo.skymap.plot tests started failing on macOS arm64 recently. Check to see if this was due to the recent release of astropy 6.0.0.
Leo P. Singer (5e11c848) at 27 Mar 20:05
Leo P. Singer (5e11c848) at 27 Mar 20:05
Collect pytest-mpl summary artifact from all test jobs