Morphology parameters in DP0 truth catalogue

This is related to the topic discussed here:

but seems distinct enough to merit its own thread.

I was wondering if there was any way to access the ground-truth Sersic parameters for galaxies in DP0, not just best-fit exponential and de Vaucouleurs values from the measurement catalogue. The TruthSummary table available via the RSP only seems to have positions, fluxes, and a handful of other parameters like redshift and Galactic extinction-related parameters.

I was considering whether I could use the DP0 object catalogue to make some simulated images for my own purposes, for which I need noise-free images, model galaxies drawn out to very low surface brightness (>30), and stars convolved with an extended (~arcminute-scale) PSF model I have on handā€”things that arenā€™t really available in the DP0 images, so it made sense to me to redraw the galaxies locally for small test-regions on the sky.

1 Like

Hi @aewatkins, thanks for this question and for participating in Data Preview 0.

As far as I know, there is no way to access the ground-truth Sersic parameters for galaxies in DP0, which is based on the DESCā€™s DC2 simulation. These features are not included in the DECSā€™s DC2 data release (Appendix B of arXiv:2101.04855), and so theyā€™re not available in the DP0.2 data release either (and are not in the additional truth data available as described in this contributed DP0.2 notebook).

However, DESC members might have access to the underlying DC2 simulation data for galaxies, if thatā€™s an option for you?

I asked some DESC members and they clarified that in the cosmoDC2 catalog, all the disks and bulges are assumed to have n=1 and n=4 Sersic profiles, respectively and half-light radii are specified for the sizes. These were then rendered in the image simulations, but for some of the galaxies, some of the flux was converted to knots as described in the DC2 paper.

He Melissa. Thanks for the quick reply. I think what would be useful for me would be a table showing which objects in the TruthSummary were built with which components (n=1, n=4, or both), and magnitudes and half-light radii associated with those components for each object. The presence of knots isnā€™t needed for what Iā€™m planning.

Is it possible to request such tables for specific parts of the sky (say, of order the size of a tract or a patch)? I wouldnā€™t personally need this for the whole simulation (although maybe someone would want it in the future, e.g. to derive uncertainties on Sersic parameters or some such thing).

Hello,

As I understand it from my conversation with DESC members, all simulated galaxies have true n=1 for their disk and n=4 for their bulge, but the true magnitudes and true half-light radii for each component are not part of the DESCā€™s DC2 public release, and so I think are unavailable.

But, Iā€™m double checking with DESC members, and have passed along your description of specifically what youā€™re looking for, in case Iā€™m wrong about how much truth information has been made public. Iā€™ll report back!

Thanks again for participating in DP0,
Melissa

Hi @aewatkins! Thereā€™s a way to obtain these quantities, but the process is a little involved. The TruthSummary table does not contain all the columns from the original truth catalog that went into the DC2 images (cosmoDC2_v1.1.4_image), but the full catalog is actually available following the instructions here. If you donā€™t need the full area, the catalog files are organized in healpixels, so you would get the healpixel ID of interest from the TruthSummary table (I believe the column is called cosmodc2_hp) and just download the piece of catalog that corresponds to the healpixel of interest. Then, you could join the TruthSummary table using the cosmodc2_id and objectID columns from the TruthSummary and cosmoDC2_v1.1.4_image catalogs, respectively. Once you join the tables, you can use the bulge_to_total_ratio_i column to get the bulge fraction (if I remember correctly, we fixed the morphology across bands), and I believe that the disk_to_total_ratio is just 1-bulge_to_total_ratio_i. As a side note, we added some ā€œknotsā€ to the brightest sources, but I donā€™t remember if that was included in the total input magnitude or added after the fact. I hope this helps!

2 Likes

Just to add a few more details to what @fjaviersanchez said:

To transfer the files, youā€™ll want to log in to Globus from the posted link. The cosmoDC2 truth tables are in the ā€œLSSTDESC Publicā€ collection, under the path ā€œ/~/lsstdesc-public/dc2/cosmoDC2_v1.1.4/ā€. There you will find files like ā€œz_0_1.step_all.healpix_9684.hdf5ā€ (that particular tract covers much of the two tracts used regularly by DM for testing, 3828 and 3829). Note that these files are very large - the z_0_1 files are ~6GB each, z_1_2 ~16GB, and z_2_3 ~21GB, partly because DC2 is very deep (every source of r>30 mag was rendered, AFAIK, and it occurs to me that these truth files might have more galaxies than that?), and because there are hundreds of columns. Anyway, you can transfer these files to any Globus endpoint you have access to.

Once the files are transferred, you can use h5py to load the file, and the relevant morphological properties are under ā€œgalaxyProperties/morphologyā€, the IDs under ā€˜galaxyProperties/galaxyIDā€™, and luminosities/magnitudes under ā€œgalaxyProperties/LSST_filtersā€. There are also stellar masses and a variety of other properties.

ā€¦ and come to think of it, @fjaviersanchez, I donā€™t see those cosmoDC2_v1.1.4_image catalogs youā€™re referring to - are they part of LSSTDESC Public or another collection?

1 Like

Thank you for your answer @dtaranu, and apologies. It is simply cosmoDC2 v1.1.4 in the desc data portal.

Thanks, all! I think I can work with this; it sounds like everything I need is in that collection. Iā€™ve set up a Globus account and can see the data, so Iā€™ll just need to transfer what I think I want and go from there.

Ah, I see what you mean. https://data.lsstdesc.org/transfer is a handy shortcut to transferring DC2 files through Globus (as I described above; it basically skips the step of choosing the input collection).

Just a quick note to wrap up this topic ā€“ Iā€™ve marked @fjaviersanchezā€™s response as the ā€œsolutionā€ but am noting here that @dtaranuā€™s advice on Globus was, according to @aewatkinsā€™s confirmation replay, also essential. Thanks all!