Airglow/Coadd Noise Study

I've made a brief study of the cnoise(4) parameter following today's telecon and offer the following quick and dirty analysis of the Hcn4 parameter and under what conditions it appears to trace airglow. The bottom line is that I propose that a reasonable formulation to detect OH airglow residual structure indicator appears to be:

Following recent discussions it appears we will relax the Jcn4 < 5.0 requirement and inspect such cases by hand since they are so few in number.

First, consider a plot of Hcn4 plotted against Jcn4:

Fiducial lines are drawn  at Hcn4 = 4.5, Jcn4 = 5.0. There is a significant population of points with high J & H cnoise scores. We believe that residual airglow structure should be dominant at H and not J, so this population must be attibutable to something else.

Next consider plots of Jcn4 vs. Kden and Hcn4 vs. Kden:


Clearly at high K source densities the reliability of the cnoise4 indicator becomes very poor. I've included a conservative fiducial cut at log density = 4.2. It's pretty clear that there is a significant correlation between J & H cnoise4 indicators above this line. In the Hcn4 vs. Kden plot (lower) it is also clear that there is a separate population of high noise scans at H with no J correspondants. I think these are our best airglow candidates.

Finally I present the first plot of Hcn4 vs. Jcn4, this time including only points with Kden < 4.2:

Now it looks like we've gotten rid of the bulk of those interlopers with both Jcn4 > 4.5 and Hcn4 > 5.0. The few remaining points in this regeme are most likely due to vertical diffraction spikes and similar phenomena which will show up at all bands.

Subject: analysis of upper left quadrant in cnoise4 plot
Date:    Wed, 22 Dec 1999 19:28:28 -0800
From:    Davy Kirkpatrick 


    I've now looked at objects which fall in the upper left quadrant
of the bottom plot on Robert's web page --

Specifically, I had Gene pull out of the SIT file all scans having
Hcn4<5.0, Jcn4>5.0, and Kden<4.2 (note, however, that Robert's plot
used Kden<4.5).  In all there were only 51 scans matching these criteria.  

    As can be seen from the analysis appended below, scans which live
in this region encompass both low-quality garbage (clouds, J-band
quadrant jumps, stray light from the moon, etc.) and high-quality data
(that happen to contain very bright stars).  The low-quality garbage 
was already filtered out through other means during QA review.

    My recommendation based on this analysis is that we should retain 
points in this region because we'd otherwise be tossing a few babies
out with the bathwater.

    I'm now working on the list of scans falling in the upper right
quadrant and hope to have that list ready on Monday.



Night    Scan(s)  Qo  Notes
-------  -------  --  -----
970918n  078       0  All of these scans were "badly non-photometric"
         083       0  and there was also a report of many of these
         099       0  scans having J-band quadrant jumps (but since these
         135-136   0  were quality=0 already, I'm not sure which of these
                      had jumps and which didn't).

980106n  071      10  (unknown -- after painstakingly bringing up all of the
                      coadds on the visualizer, I can see no reason why the
                      Jcn4 counts are so high)

970723n  023-037   0  The pre-amp box was changed out on this night and
         078-093   0  had wrong setting so all of the J-band photometry
                      was incorrect, forcing all of these to quality=0.
                      It was also noted that many of these suffered from
                      J-band quadrant jumps, too, but again the specific
                      scans affected were not spelled out.

980702n  012       0  Scan had "severe problem in coadds 150-173, looking 
                      as if something bright (and very blue) flew through or 
                      nearby the field".

980404n  011       0  Scan had severe J-band background variations due to
                      stray light from the moon.  (This scan was only about
                      25 degrees away from the moon.)

970920n  099      10  (unknown -- but source density must be very close to
                      the Kden=4.2 threshold)

981008n  041       0  Severely affected by clouds.

990508n  103      10  There is a monster bright star in this scan --  
                      g Herculis, M6 III, V=5.01.

990501n  027       0  Severely affected by clouds.

991020s  091      10  Stray light in coadds 139-245 from the very bright star
                      HD 39775, K0 III, V= 5.98.

990804s  087       9  (unknown -- source density must again be near the 
                      Kden=4.2 limit)

981107s  099      10  There is a very bright star in this scan --
                      S Pictoris, M7e, V=9.12.

990430s  036       0  Heavily impacted by clouds.

980607s  024       0  Severely affected by clouds.

990417s  092      10  This scan has Antares, M1.5 Ib, V=1.09.  It doesn't get 
                      much brighter than this.

990530s  024       0  Impacted by clouds.

981028s  122      10  This scan contains R Doradus, M8 IIIe, V=5.73.  Possibly 
                      even worse than Antares.

Subject: analysis of upper right quadrant in cnoise4 plot
Date:    Mon, 27 Dec 1999 16:05:18 -0800
From:    Davy Kirkpatrick 


    I've now looked at objects which fall in the upper right quadrant
of the bottom plot on Robert's web page --

Specifically, I had Gene pull out of the SIT file all scans having
Hcn4>5.0, Jcn4>5.0, and Kden<4.2 (note, however, that Robert's plot
used Kden<4.5).  In all there were only 48 scans matching these criteria. 

    As you suggested, I only inspected those scans having quality>4, and
that list is appended below.  In every case, Jcn4 and Hcn4 seem to be
high either because of a monster bright star causing severe background
variations (as Robert had pointed out earlier) or because of a combination
of moderately high source density combined with moderate or bad airglow.

    Of the 21 scans in the appended table, 13 do not contain monster
bright stars.  It may be worth rescanning these since airglow does appear
to have been a factor in their receiving high cn4 values.  



Night    Scan(s)  Qo  Notes
-------  -------  --  -----
990318n  131      10  This scan has the very bright star alpha Herculis,
                      M4 Ib-II, V=3.5.

980515n  063      10  This scan has Arcturus, K1.5 III, V=-0.04.

991020s  092      10  This scan has the very bright star HD 39775, K0 III,
                      V=5.98.  Although this one may not seem to be a monster
                      based on its type and V magnitude, it was noted in the
                      QA review that there was a lot of stray light from this
                      one over coadds 139-245.

990720s  073      10  This scan has V2392 Ophiuchi, M4 III, V=6.2.

990807s  089      10  Despite the moderately high density of this scan
                      (Kdens = 3.75), I see no reason from the visualizer
                      coadds why the cnoise4 numbers should be high.
                      However, a combination of bad airglow plus high density
                      may be enough to cause high values in cases like this,
                      and this scan does show airglow variations even above
                      the stellar background for a couple of coadds.

990721s  075,085, 10  Again, there's no obvious reason from the QA review 
         089          why the cnoise4 counters are high, but these scans are
                      of moderately high density (3.74 < Kdens < 3.92).  Some
                      bad airglow was also noted for these scans.

990723s  096      10  Again, just a high density field (Kdens = 4.04).

990724s  054-056, 10  Again, just high density fields (3.93 < Kdens < 4.00).
         064-066,     However, these scans (and ones around them) contain some
         096          particularly nasty episodes of airglow variability.

980602s  067      10  This scan has W Hydrae, M8e, V=7.0 (var).  Noted in the
                      QA review as "extremely bright" and causing background
                      jumps as far as three coadds east/west of the star 

980714s  036      10  This scan contains the southern diffraction spike from
                      Antares.  It is noted that this vertical spike is seen
                      more than 6 degrees away from Antares itself.

990617s  027       9  Most likely, this scan is triggered because of high
                      density (Kdens = 4.19) and bad airglow (as noted in 

991026s  113      10  Stray light over coadds 9-127 was noted in this scan
                      during QA review.  Most likely the culprit is Canopus
                      even though Canopus is not in the scan itself.

981024s  123       7  This scan contains vertical diffraction spikes from
                      the monster bright star R Doradus.

R. Hurt, last updated 12/29/99