2.3 HIDROCARBUROS HALOGENADOS
2.3.3 IMPACTO AMBIENTAL DE LOS HIDROCARBUROS HALOGENADOS
The pipeline in Warwick currently used to analyse NGTS data from Paranal is similar to the basic version I designed to analyze the data from Geneva, as described above. It has been extensively updated by Simon Walker, Richard West, and other members of the NGTS consortium. However, the same basic software tasks and workflow were followed for the production of early NGTS science data.
I have produced several quality control plots for the pipeline to track its performance.
Vector astrometry
In order to track the quality of the astrometric solution over time, the pipeline produces vector astrometry plots. The detected positions of the sources are marked with black points, and the vector to the catalog position of the source, multiplied by a factor of 106 to make it visible, is also plotted. The center of the distortion
terms are also indicated with a red mark. An example of this pipeline product is given in Figure 3.5.
Focus performance
My Point Spread Function (PSF) recovery code utilises the extremely precise astro- metric solution described in the previous subsection.
The image is divided into 9 sections, and in each Section the regions sur- rounding the brightest 100 stars aresuper sampled by a factor of ten. These regions are centered on the catalogue position for that source. The local sky background is removed, and the pixel counts are rescaled, so the peak of the star psf is at 1. These 100 super sampled regions are combined, which produces a PSF sampled with higher resolution. This works in a similar way to the Drizzle algorithm, which is often used to improve the resolution in HST data (Fruchter and Hook, 2002). A 2 dimensional gaussian is fit to the oversampled PSF in each of the 9 sectors, and the
Figure 3.9: A single partial transit of TrES-2b recorded with NGTS (top) is compa- rable in significance to multiple seasons of WASP data, which combine 39 transits (bottom).
semi-minor, semi-major, and rotation angle of these fits are stored for each image frame. The purpose is to check that the PSF remains stable over time, but these data could also be used to test for, e.g., the tilt in the focal plane. An example output plot is shown in Figure 3.10.
Photometric performance
If the noise in the lightcurves behaves as white noise, then binning together multiple points will decrease the RMS as √N, where N is the number of points. At some point, the red noise limit will be reached. Red noise does not have a frequency dependence, so will not be reduced by combining more frames. Quantifying the levels of white and red noise in the data is an important diagnostic, so a useful plot to generate for a set of data is the RMS level as a function of the number of binned points.
Due to the instrument response, it is possible that there are different red and white noise levels as a function of star brightness. An example of this would be stars bright enough to be saturated, which will have an additional source of red noise from non conservation of charge. Therefore, the diagnostic plot can be made more useful by separating the stars into brightness bins. An example of this diagnostic plot from recent NGTS data is shown in 3.11, both before and after detrending.
The apparently very high and consistent RMS level in the first plot is due to airmass trends in the data, as this task covered an entire nights worth of data. As a result, every star has an airmass curve, with an amplitude of several 10s of mmags that cannot be reduced by binning, so appears as a red noise source. The airmass curve is one of the first trends to be removed bysysrem, so in the second plot the RMS is allowed to decrease much further with binning, in this case most brightness bins reaching levels of a few millimags on timescales of∼100 minutes, except for the brightest bin, which flattens off earlier due to saturation effects.
3.5
Conclusions
In this Section I have summarized some of my contributions to the ongoing NGTS project.
I demonstrated with my analysis of the Geneva data gathered with an NGTS unit in June-August of 2013 that the instrument was performing well, capable of regularly reaching precisions better than 10 mmag in 10 second exposures on bright stars. The data are white noise dominated and stable over long timescales, as demonstrated by the lightcurves of variable stars, including a short period delta-
Figure 3.10: Top: Recovered PSFs for the 9 image sections. Bottom: evolution of the PSF fit parameters over a typical observing night.
Figure 3.11: An RMS binning plot divided by magnitude bins, from red (faintest), to yellow (brightest). Top: raw data Bottom: after detrending by sysrem. Data is from one night in Paranal on 06/09/15. Stars near saturation (yellow) do not improve with binning. Dashed lines indicate how data would improve if noise was purely white. The precision on this night was limited to a few mmag by red noise from weather conditions.
scuti with an amplitude of 1 mmag. I demonstrated that the detector could detect an exoplanet transit at very high significance, with a partial transit of TrES-2b being comparable to several seasons of WASP data. These are the highest precision lightcurves ever achieved in a wide-field ground-based survey. My validation of the CASU photometry tools led to them forming the core of the current development version of the NGTS production pipeline.