Uploaded image for project: 'Data Management'
  1. Data Management
  2. DM-12793

implement color terms for jointcal photometry

    XMLWordPrintable

    Details

    • Epic Name:
      jointcal color terms
    • Team:
      Alert Production

      Description

      Analysis work by Pierre Astier suggests that the common failures to fit the ConstrainedPhotometryModel in HSC are due to excessive outlier rejection of stars with beyond median color. To fix this we likely need to implement a focal plane color term as a fittable object, to correct for the color variation. For now it can be a constant across the focal plane or constant per-visit, but it should allow for position dependence, and will have to provide parameter derivatives for the fitter.

      Pierre is also investigating algebraic methods to identify outlier stars whose removal may cause the problem to become underconstrained. Progress on that may result in this becoming less critical.

        Attachments

          Issue Links

            Activity

            Hide
            jbosch Jim Bosch added a comment -

            Are you saying that you need different color terms for different visits, or just that you need color terms at all?

            If you can just take a constant color term as given, we can provide that (I don't know how the machinery works, but Paul Price certainly does).

            If you need something spatially varying, for HSC that'll be dominated by the radial dependence of the HSC-R and HSC-I filters. My new TransmissionCurve stuff might help with that, and I could try to push the parts you'd need in sooner rather than later.

            Show
            jbosch Jim Bosch added a comment - Are you saying that you need different color terms for different visits, or just that you need color terms at all? If you can just take a constant color term as given, we can provide that (I don't know how the machinery works, but Paul Price certainly does). If you need something spatially varying, for HSC that'll be dominated by the radial dependence of the HSC-R and HSC-I filters. My new TransmissionCurve stuff might help with that, and I could try to push the parts you'd need in sooner rather than later.
            Hide
            astier Pierre Astier added a comment -

            I think the idea is that we should be able to fit color terms together with other parameters, without relying on some existing solution. It seems that a spatially variable color term (in some sort of focal plane coordinates) is desirable given what we now know of the filters fabrication. There is also the possibility that chip coatings cause differential color terms. I think that the idea here is to devise an interface to the photometry fit that allows one to eventually address these kind of issues, but for now, limit the implementation to a single color term extra parameter across the whole instrument and input dataset, and allow for e.g. a radially varying color term once the photometry fit no longer fails presumably for lack of references.
            It seems that being able to fit color terms is a desirable functionality. Of course, we can take the temporary short cut of cooking up a color-matched reference catalog, but it is not clear that this requires much less programming.

            Show
            astier Pierre Astier added a comment - I think the idea is that we should be able to fit color terms together with other parameters, without relying on some existing solution. It seems that a spatially variable color term (in some sort of focal plane coordinates) is desirable given what we now know of the filters fabrication. There is also the possibility that chip coatings cause differential color terms. I think that the idea here is to devise an interface to the photometry fit that allows one to eventually address these kind of issues, but for now, limit the implementation to a single color term extra parameter across the whole instrument and input dataset, and allow for e.g. a radially varying color term once the photometry fit no longer fails presumably for lack of references. It seems that being able to fit color terms is a desirable functionality. Of course, we can take the temporary short cut of cooking up a color-matched reference catalog, but it is not clear that this requires much less programming.
            Hide
            jbosch Jim Bosch added a comment -

            Robert Lupton and/or Merlin Fisher-Levine should comment here, but the only reason I hesitate is because I think we have mostly planned for JointCal to use wavelength-dependent components that are better constrained elsewhere, at least for LSST. That includes of auxiliary telescope data, the collimated beam projector, and either a specially processed Gaia catalog or Eli Rykoff's FGCM.

            That said, we won't have all of that extra information for precursor data, and if we end up using FGCM I don't think we've really defined a clear boundary between it and JointCal. But I think I also recall Eli Rykoff saying that there actually isn't enough information to constrain some of these wavelength-dependent terms in an FGCM- or JointCal-style fit.

            I suppose the big question is how hard this is to add, and what the timescale is - if it's easy, I could imagine it being generally useful to have around. But if we need it for the JointCal vs. meas_mosaic comparison, it may be better to go with the pre-derived HSC color terms we're currently using with meas_mosaic.

            Show
            jbosch Jim Bosch added a comment - Robert Lupton and/or Merlin Fisher-Levine should comment here, but the only reason I hesitate is because I think we have mostly planned for JointCal to use wavelength-dependent components that are better constrained elsewhere, at least for LSST. That includes of auxiliary telescope data, the collimated beam projector, and either a specially processed Gaia catalog or Eli Rykoff 's FGCM. That said, we won't have all of that extra information for precursor data, and if we end up using FGCM I don't think we've really defined a clear boundary between it and JointCal. But I think I also recall Eli Rykoff saying that there actually isn't enough information to constrain some of these wavelength-dependent terms in an FGCM- or JointCal-style fit. I suppose the big question is how hard this is to add, and what the timescale is - if it's easy, I could imagine it being generally useful to have around. But if we need it for the JointCal vs. meas_mosaic comparison, it may be better to go with the pre-derived HSC color terms we're currently using with meas_mosaic.
            Hide
            rhl Robert Lupton added a comment -

            The question of spatio-temporal variation of filters is more complicated than a spatial model in jointcal.

            John's initial issue says, "common failures ... are due to excessive outlier rejection of stars with beyond median color." I don't see how spatially varying colour terms help here – isn't the problem that the colour terms (i.e. corrections for different effective filter transmission curves) are not correct for extreme objects, possibly due to complex spectra for which a simple "colour term" isn't sufficient? In this case I see no alternative to trimming extreme objects out of the reference catalogues, which is what HSC does. Of course, bad (constant) colour terms exacerbate the problem of clipping, and that's probably what John meant.

            If the problem is the spatial variation of the colour terms (which is what I think Pierre's concerned about) then I agree that that's something that we could estimate in jointcal, but I don't think we want to do that in the long term because (as Jim notes) there's lots of external information. This isn't really a pure spatial variation but changes in time, either due to dropping or upgrading a filter, or due to changing atmospheric absorption. If we neglect the temporal dependency (which we cannot for HSC, as we've upgraded the i and r filters – and I'd argue that we should plan for the LSST filters evolving and LSST certainly needs to track the atmosphere) then jointcal is one place that we could track this.

            In the short term a fixed spatial model of colour terms coefficients is reasonable, and whether jointcal estimates the terms or we get them from filterscans + models is a detail (we certainly wouldn't estimate them per jointcal run). Jim thought that Eli Rykoff had some problem with jointcal's (actually FGCM's) ability to estimate these constant-in-time spatial variations, so it'd be good to get that clarified. I think that for jointcal to measure spatial variation of colour terms would require that jointcal has colour information about each object even though I think it runs per-band (is that right, John?); maybe it does – they don't have to be all that good colours as errors only enter at higher order.

            In the longer term we need to track effective filter curves for each measurement of an object: "I was measured on an ITL chip at this position in the focal plane on Thatcher Day 2027 under these atmospheric conditions" right through to the science database, and jointcal is probably the place to do this (it'll need per-object colour information to do this). It does mean that some of this information that jointcal is gleaning/assembling needs to be persisted as per-source "measurements", and that implies something about who has to be involved in designing the data products.

            Show
            rhl Robert Lupton added a comment - The question of spatio-temporal variation of filters is more complicated than a spatial model in jointcal. John's initial issue says, "common failures ... are due to excessive outlier rejection of stars with beyond median color." I don't see how spatially varying colour terms help here – isn't the problem that the colour terms (i.e. corrections for different effective filter transmission curves) are not correct for extreme objects, possibly due to complex spectra for which a simple "colour term" isn't sufficient? In this case I see no alternative to trimming extreme objects out of the reference catalogues, which is what HSC does. Of course, bad (constant) colour terms exacerbate the problem of clipping, and that's probably what John meant. If the problem is the spatial variation of the colour terms (which is what I think Pierre's concerned about) then I agree that that's something that we could estimate in jointcal, but I don't think we want to do that in the long term because (as Jim notes) there's lots of external information. This isn't really a pure spatial variation but changes in time, either due to dropping or upgrading a filter, or due to changing atmospheric absorption. If we neglect the temporal dependency (which we cannot for HSC, as we've upgraded the i and r filters – and I'd argue that we should plan for the LSST filters evolving and LSST certainly needs to track the atmosphere) then jointcal is one place that we could track this. In the short term a fixed spatial model of colour terms coefficients is reasonable, and whether jointcal estimates the terms or we get them from filterscans + models is a detail (we certainly wouldn't estimate them per jointcal run). Jim thought that Eli Rykoff had some problem with jointcal's (actually FGCM's) ability to estimate these constant-in-time spatial variations, so it'd be good to get that clarified. I think that for jointcal to measure spatial variation of colour terms would require that jointcal has colour information about each object even though I think it runs per-band (is that right, John?); maybe it does – they don't have to be all that good colours as errors only enter at higher order. In the longer term we need to track effective filter curves for each measurement of an object: "I was measured on an ITL chip at this position in the focal plane on Thatcher Day 2027 under these atmospheric conditions" right through to the science database, and jointcal is probably the place to do this (it'll need per-object colour information to do this). It does mean that some of this information that jointcal is gleaning/assembling needs to be persisted as per-source "measurements", and that implies something about who has to be involved in designing the data products.
            Hide
            erykoff Eli Rykoff added a comment -

            A lot to chew on in the comments here, and I won't be able to comment on everything, but let me specify what I mean about FGCM and the focal-plane variant color terms. First, there are definitely significant variations in the color term radially in HSC i and r filters (which the filter scans predict well) and the ccd-by-ccd variations in qe in g and r which we don't have any scans for. However, while these are "significant", and definitely necessary to take into account if you need to get below 1% photometry, I don't see how these can be causing fitting failures, since you can do 1-2% perfectly fine without them. Now, if the problem is that the selection of the input reference photometry is overly strict in terms of colors and this is removing all reference sources, then I would say that's a somewhat orthogonal problem in reference selection, but I could very well be misreading the comments above.

            Second, what I mean about "not being enough information". There is certainly enough information in the stars to compute color terms that are appropriate for main-sequence stars. The problem becomes when you want to apply these to the galaxies/quasars/whatever. In this case you need to know the passband in order to properly multiply it by the source SED. And just given variations in star colors, it is not possibly to uniquely map these back to a true wavelength-dependent passband.

            Now, if jointcal is only trying to figure out a gray correction to the zeropoint, then maybe using the stars to give an empirical color term in the fitting would be fine (and would certainly help the fit over a broader range of star colors). But how this interacts with the true passband variation is something that we need to figure out, and the best option would be using external information (but there's the question of timescales of course). Of course, you're better off having the color terms constrained on the proper spatial/temporal scales (the instrument part doesn't vary over the sky and basically only varies when the filter is switched; the atmosphere part varies per exposure, but is correlated from exposure to exposure, etc.)

            Show
            erykoff Eli Rykoff added a comment - A lot to chew on in the comments here, and I won't be able to comment on everything, but let me specify what I mean about FGCM and the focal-plane variant color terms. First, there are definitely significant variations in the color term radially in HSC i and r filters (which the filter scans predict well) and the ccd-by-ccd variations in qe in g and r which we don't have any scans for. However, while these are "significant", and definitely necessary to take into account if you need to get below 1% photometry, I don't see how these can be causing fitting failures, since you can do 1-2% perfectly fine without them. Now, if the problem is that the selection of the input reference photometry is overly strict in terms of colors and this is removing all reference sources, then I would say that's a somewhat orthogonal problem in reference selection, but I could very well be misreading the comments above. Second, what I mean about "not being enough information". There is certainly enough information in the stars to compute color terms that are appropriate for main-sequence stars. The problem becomes when you want to apply these to the galaxies/quasars/whatever. In this case you need to know the passband in order to properly multiply it by the source SED. And just given variations in star colors, it is not possibly to uniquely map these back to a true wavelength-dependent passband. Now, if jointcal is only trying to figure out a gray correction to the zeropoint, then maybe using the stars to give an empirical color term in the fitting would be fine (and would certainly help the fit over a broader range of star colors). But how this interacts with the true passband variation is something that we need to figure out, and the best option would be using external information (but there's the question of timescales of course). Of course, you're better off having the color terms constrained on the proper spatial/temporal scales (the instrument part doesn't vary over the sky and basically only varies when the filter is switched; the atmosphere part varies per exposure, but is correlated from exposure to exposure, etc.)

              People

              Assignee:
              Parejkoj John Parejko
              Reporter:
              Parejkoj John Parejko
              Watchers:
              Dominique Boutigny, Eli Rykoff, Jim Bosch, John Parejko, John Swinbank, Pierre Astier, Robert Lupton, Russell Owen
              Votes:
              0 Vote for this issue
              Watchers:
              8 Start watching this issue

                Dates

                Created:
                Updated:

                  Summary Panel