A couple of questions / project ideas (photometry)

[billyharris72]

Hi all:

Just had a couple of ideas as regards doing some variable star photometry and wanted to get some input from someone with a bit of experience.

Idea 1. A regression based photometric model that includes a term for altitude.

I've experimented with doing comparative photometry using regression (linear regression on log transformed ADU), doing the photometric measurement in AstroArt and importing the data into R for modelling. That seems to work well, but I was wondering if it's possible to extend the field of view restrictions. My idea was just to use the RA and DEC data from the AstroArt output with Julian date from the FITS header and my Latitude and Longitude and use that to calculate Alt for each object. Then feed the log transformed cosecant of alt into the regression along with ADU in order to account for atmospheric extinction and thus (potentially) widen the usable field. In principle this might work, but my concerns are:

a) This will do nothing to manage differential extinction at different wavelengths. How important is that relative to absolute extinction? Would this allow me to work with comparatively distant comparisons (say a degree or so) or is there no real gain.

b) How many comparison stars would this need to have a chance of being an improvement on a single variable model? With small numbers of comparison stars I expect it would be much worse due to correlations between the two independent variables - how many comparison stars would be needed before regular tests for multicollinearity would be valid? (I think this one might be the clincher and could scupper the whole idea).

Idea 2. Using regression to find variable stars (which might possibly allow enough check stars to make idea 1 work).

One issue with differential photometry is obviously finding "proper" photometric comparison stars in a narrow field (the vast majority of variables in the AAVSO database, for example, don't seem to meet this criterion). That set me to wondering what would happen if we allowed any star not clearly flagged as a variable to be a comparison star and managed the error statistically (i.e. by picking maybe 200 stars to build our regression model on). Even if many of the stars had some intrinsic variability, the sheer number of stars would presumably mean the effects on the model should be very minor (we'd have to use diagnostics to manage stars with high leverage).

Assuming that the outputs might not be good enough for accurate reporting of magnitudes, I then thought of the following. Say I grabbed a few hundred images of the same field over a year or so, ran photometry as in idea 1 above (with or without altitude as a variable) and then grabbed the residuals from each model and pasted them row-wise into a matrix, could the column in this matrix then be used to identify variable stars and provide reasonable approximate photometry for them?

If that worked then a nice compact target like an open cluster could presumably be searched for variables without a huge degree of difficulty.

Both just ideas, speculative at the moment, but would be interesting to know what people think.

Cheers,

Billy.

[JeremyS]

Interesting ideas Billy. However, they do seem a lot of work for questionable return.

For Idea 1, one normally selects comparison stars in the same field as the variable for CCD photometry. Over this range there is no need to correct for altitude. The minimum requirement is to have one comparison star and one reference. 

For idea 2, it's rare to find a star without suitable comps in the AAVSO catalogue. If you come across this case, then all one needs to do is to ask AAVSO. They have a dedicated chart team that will produce the sequence for you quite rapidly. They have a huge APASS photometry catalogue, which they built using pretty much  the same system as you propose. It's just they have important it into the how variable star chart catalogue. APASS has a high degree of precision and the photometry is fully accredited accepted by professionals.

I hope this does not put you off; there are many  variable stars you could usefully do photometry on. 

[billyharris72]

3 hours ago, JeremyS said:

I hope this does not put you off;

Not off putting at all - good to learn that AAVSO are able to produce sequences other than those produced by the web tool.

My understanding of comparative photometry is also that you don't need to correct for extinction, as you say. But how big a field can you get away with this over? I've seen conflicting accounts, and I suppose it depends on how much accuracy is needed.

I've always assumed that accuracy would suffer badly above maybe a one degree field (one degree at 45 degrees alt I think would increase air mass by a couple of percent or so)....

It's funny how discussing things with someone can clarify your thinking. I've just realised that I'd assumed this equated to a couple of percent error in the magnitude calculation, when in fact it's a couple of percent error in the extinction, which will be so much lower.

So my ideas above may be a solution to a "problem" that doesn't exist.

Right- where's that V filter?

Billy.

[JeremyS]

2 minutes ago, billyharris72 said:

Right- where's that V filter?

For some photometry, you can go unfiltered, Billy. An example is photometry of cataclysmic variables where we want to find the period of the variation precisely and the actual magnitude is not required. This is mostly the sort of thing I observe. I've therefore not used my V filter for ages s I can get ~1.5 mag fainter without it (or get a faster cadence  - shorter exposure times - which be important to get higher time precision)

[JeremyS]

BTW if you are interested in cataclysmic variables, I wrote a bit about them in the BAA Journal. You can loaded it via NASA ADS or get the PDF directly