filter transmission and atmosphere in VISTA filters

[silvanfischbacher]

When comparing the VISTA filters in sedpy with the ones from SVO (link), it looks like the sedpy filters match the no atmosphere versions from SVO.

But the sedpy filter notes say:

"Includes filter transmission, atmosphere, detector QE, and mirror reflectivity."

So either the atmospheric transmission is missing, or the notes in the filter files are incorrect.

[bd-j]

Thanks. Please see #35 and #39, I believe these came directly from the VISTA/ESO website, but the documentation there is not clear about the contributions included in the referenced file. Tagging @stevet40.

[stevet40]

Hmmmm, good spot! I agree that it is a little ambiguous whether the ESO files include all of the components. What I added in #39 is taken directly from the ESO website, but I added the comments based on the best of my knowledge. The current curves are consistent with the previous VISTA curves that were in sedpy (and with EAZY, as the investigations in #35 showed), up to the presence of the red/blue leaks which are included in the updates made in #39. I will investigate this more when I get a chance, and will get back to you both. I have the atmosphere, reflectivity, and QE curves from the ESO page, so I'll take a look......

[silvanfischbacher]

Great, thanks a lot!

[stevet40]

I've tried combining the different curves in the ESO tarball, and have ended up with something that looks by eye similar to what's in SVO, and to what's in Fig. 3 here: https://doi.org/10.1093/mnras/stx3073.

Fig. A: VISTA transmission functions with 3mm PWV atmosphere model.

There are two things that were a bit ambiguous for me...

1. The ESO tarball provides two different atmospheres, with different PWVs and airmass. I made the above plot (Fig. A) using the model atmosphere with 3mm PWV, and airmass=2.0 (trans_30_20.dat in the tarball). There is an alternative option (trans_10_10.dat) with 1mm PWV and airmass=1.0. I've posted the plot of that below (Fig. B). It seems like SVO used the 1mm PWV model. But the janky plot on the ESO webpage used the 3mm model (except longward of the $K_S$-filter.
2. The tabulated curves for QE, reflectivity, etc., don't necessarily cover the full wavelength range spanned by the filters (or the same wavelength grid). I've interpolated the different bits onto the same wavelength grid as the filter curves. But, a question mark is what to do beyond the edges. The way we handle this will affect e.g. if the red/blue leaks get preserved, or if they get wiped out by the QE. I really don't know what the correct answer is there, but also it may not matter too much as these things are small-fry compared to the other changes we're potentially making here!

Fig. B: VISTA transmission functions with 1mm PWV atmosphere model.

In any case, I'm happy to open a PR to add something new (and/or to improve the documentation of the current filters, which I guess is wrong). But this feels like a fairly dramatic change (compared to adding some tiny red leaks). I wonder if the right thing to do is to preserve the current curves in some form, and to add some "with corrections" versions with all of this extra stuff included?

[silvanfischbacher]

Thanks for your checks @stevet40! Regarding the PR, I agree that the changes are quite significant - this is also why I noticed them when comparing a catalog generated with the sedpy filters and one with the SVO filters. Perhaps you could update the documentation for the existing filters and introduce new ones that include the atmosphere model? Either way, I guess it’s important that users can still reproduce their previous results, so I agree that the current filter curves should be preserved in some form.

[stevet40]

Hi @silvanfischbacher! I've proposed something to this effect in #42, but a little bit uneasy about such a change for the reasons you mention. I'm also not 100% convinced that the new curves I made by combining the files from ESO are exactly the same as what's in SVO (perhaps they used their own Paranal atmosphere model or something like that). But they seem pretty close.

[silvanfischbacher]

I like your proposition. Regarding the exact choices of PWV and airmass, I'm not an expert in this area, so I don't have strong opinions on what the most reasonable values would be. That said, your choices seem reasonable to me.

[bd-j]

I think having two sets of filters is a reasonable solution, especially if the documentation is clear about what is done.

Also just to mention, in 53bd36b I added a new capability to include multiple transmission columns, e.g. for different detector chips with Roman, that could in principle be used here to have different atmospheric transmissions (including a filter-only column)

[bd-j]

Ok, I'm going to merge #42 Thanks @stevet40 and @silvanfischbacher for your attention to this!