Time series data

[Petr Skoda]

Dear Doug and WG members

I have just read the proposal and at the first look it looks consistent 
with the practical understanding of Light curve in (optical) stellar (and 
ground-based) astronomy - But there are two common problems coming 
probably from the orientation of the LCDM to transits (say exoplanet 
ones). It also answers some of the questions in Chap. 2 .

The periodogram is a method-dependent representation of simple 
(well-defined) mathematic transformation (often the FT - e.g. CLEAN 
algorithms, sometimes advanced statistics is applied - the people working 
in variable stars know about 20 different algorithms and new have been 
still devoloped - e.g. for GAIA).
The primary goal of a periodogram is the identification of periods 
conformant with the proposed physical model or expected behaviour (e.g. 
the 1-day aliases of Eart rotation must be excluded, but sometimes it is 
difficult to do it).

So in principle the periodogram is a derived product which is not 
necessary for data discovery or primary access (it may be easily 
constructed IF THE PERIOD and EPOCH is given.

The typical announcement in a journal or in the amateur alert is like
  For HMB07 a period P = 0.53284 +/- 0.00017 [d] with the epoch E0 = HJD 
2454155.683 and an amplitude of the light variation of dm = 0.45 +/- 0.05 
mag ....

I should emphasize that the EPOCH may be choosen arbitrarily and is 
different from what is referred as a time of observation of any point - So 
it requires the definition of another entity (say when the period is 
stated the EPOCH must be stated as well as well as an error in period)

BUT :
The period may be changing either in jumps (physical nature of the object 
is changing) or the gradually (e.g. systematic magnetic breaking of a 
rotation).

>From this it follows a lot of interesting physical theories -  people are 
investigating the variations of periods etc ... (the term O-C diagram is 
telling the difference between the computed value of certain effect (e.g. 
primary minumum of light curve computed from published period (valid for 
some Epoch) and its observed value.

Another problem is with multiperiodicity. Some stars have the real 
multiple periods - in asteroseismology it is the core of the succes of the 
analysis of stellar interior).
The astronomer may want to fold the lightcurve with one period to get some 
physical information or another to get different.

We may as well have stars with rotation period similar to the orbital one 
....


So as you see the light curve analysis is a crucial point of the 
astronomical research and if we want to devise a real standards for the 
community and to allow the precise transormation of most astronomical data
into  the VO format (with particular services) these issues have to be 
reflected somehow.

So the particular light curve (in addition to basic time/intensity 
information) may be accompanied by the list of (period,Epoch) information
which may be an array of many values.
In the provenance the reference to the article estimating the particular 
periods should be provided as well.

The information about  periods end epochs  is important for the analsis 
tools which can directly draw the particular curves folded (even if the 
primary data are not folded).


The another problem is common to both LCDM and PhotDM - it is the filter 
zero point - some  important discoveries were done using only RELATIVE 
calibrated data (i.e. the magnitudes were only instrumental not converted 
to standard filters - this means the influence of the atmosphere was not 
corrected.   Sometimes (on CCD photometry) the arbitrary magnitudes are 
just computed from photon cound acquired in aperture photometry.

I can have arbitrary offset in data numbers or even different offsets in 
different colours (including the atmosphere influence) but I can still 
find important periods (typically to show the object is a Blazar or 
magnetic cataclysmic star or polar etc ...
In fact the proper zero point calibration and atmospheric corrections are 
quite often not stated or applied.

I thing this should be discussed as well in the Photometric model session.
and allow some freedom in all parameters -of course it should be said in 
characterization part ....

Moroever the definition of the variability amplitude is still very vaguely 
done:

as it is said in 3.1.6 "The variability amplitude is a standard value but 
especially important for time series data."

So it should be explained properly but all we know from the STSP note is 
the meaning of zero (no variability) but what does mean 1 ? In SDM it is 
defined as a "fraction of mean"  - again very transit-centric statement - 
why I cannot have VARAMPL>1 or even >>1 ? It should be described more 
precisely - best giving examples and some formula - but IMHO it is WRONG 
idea to limit anything (what about the peak outburst - how the mean is 
defined ?????? We may have flat curve for a long time and suddently rapid 
oscillations occurs - or what about cataclysmic variables - by definition 
they are in quiescence state and suddenly increase their brightnes of 
several orders !



I think most of the problems described above should be solved before any 
serious work on introduction of GDS and their protocols will be presented 
to wide astronomical community.

*************************************************************************
*  Petr Skoda                         Phone : +420-323-649201, ext. 361 *
*  Stellar Department                         +420-323-620361           *
*  Astronomical Institute AS CR       Fax   : +420-323-620250           *
*  251 65 Ondrejov                    e-mail: skoda at sunstel.asu.cas.cz  *
*  Czech Republic                                                       *
*************************************************************************