October 7, 2016 at 9:00 pm #1415
Here is the description of Part 2 of my GMOS data reduction pipeline, introduced in my previous post.
This part of the pipeline is focussed on high-precision timeseries observations of exoplanet transits. It may not be suitable as-is for other uses but could be adapted for other timeseries spectrophotometry if needed.
Here is a summary of steps in the second part of the pipeline:
– Extract variable parameters from FITS headers, such as airmass (can be used later for de-trending transit lightcurves)
– Interpolate over gaps between CCDs in the detector using linear interpolation. A list of interpolated pixels is given at the end of the pipeline so these wavelengths can be ignored in further analysis
– Fit a linear function to pairs of pixel location and corresponding wavelength to obtain a wavelength solution for each star. This step requires the user to identify features manually using IRAF first
– Compute a model for differential atmospheric refraction for each exposure in an observation and each wavelength. Use this to correct wavelength-dependent dispersion-direction shifts over time in the spectral timeseries
– Cross-correlate spectra in time to take account of remaining changes in the wavelength solution as a function of time and wavelength. This step is taken because there is an additional “stretching” of the wavelength solution over the course of an observation that is not accounted for by differential atmospheric refraction. Our procedure uses spectral features identified by the user to measure lags as a function of time for multiple wavelengths
– Use cross-correlation to correct for the constant offset in wavelength between the target and reference star (due to the instrument PA not being exactly the same as the PA between the stars). Linearly interpolate the reference stellar spectra onto the target star’s wavelength grid
– Test for the effects of spectral shifts and cross-correlation uncertainties on measured planet transit depths as a function of wavelength. This is done by first producing simulated transits and injecting known spectral shifts. We then measure the transit after the shifts have been introduced and compare the measured depth with the known input depth.
– Compute optimal spectral bin size to use for transmission spectra. This is done by measuring the effects of spectral shifts on the planet transit depth as above, but in a variety of bin sizes.
The output files are a timeseries of corrected 1D spectra for both the target and reference star. These are saved into files that can then be read into IDL or Python and used directly to create transit lightcurves using the optimal binning identified in the pipeline.
The pipeline makes use of the following procedures:
– Differential atmospheric refraction code written by Enrico Marchetti at ESO, available at https://www.eso.org/gen-fac/pubs/astclim/lasilla/diff\_atm\_refr.pro.
– the fitting package mpfit (Markwardt, 2009, in ASPC Conf. Ser. Vol. 411)
– the Exofast package (Eastman \& Agol, 2013, PASP, 125, 83)
– the analytical transit models of Mandel & Agol, 2002, ApJL, 580, L171
– the NASA IDL Astronomy Library, available at http://idlastro.gsfc.nasa.gov
Let me know if you have questions or comments about the procedures (or suggestions for improvements). I will post a link to the public pipeline and documentation when it is available.00
You must be logged in to reply to this topic.