Use this URL to cite or link to this record in EThOS:
Title: Optimal extraction of planetary signal out of instrumental and astrophysical noise
Author: Danielski, C.
ISNI:       0000 0004 5362 1142
Awarding Body: University College London (University of London)
Current Institution: University College London (University of London)
Date of Award: 2014
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Astonishingly for a discipline that did not even exist twenty years ago, the science of Exoplanets has arguably evolved exponentially, delivering transformational science. Planetary formation theories had to be completely revisited. The critical step yet to take is the determination of the chemical composition of the atmospheres of these exotic worlds. Detecting atmospheric features, which have a contrast of about 10^(-4) compared to the host stars radiation, is a challenge, especially since none of the instruments available can provide an absolute calibration at that level. To measure even the smallest flux variations, the data reduction techniques adopted are therefore critical. In particular, instrument systematics and stochastic errors need meticulous corrections. The menagerie of parametric correction models used in the field, has often led to fierce debates due to the high level of degeneracy between the correction model and the scientific result. However, some of these concerns can be addressed by adopting more robust and objective statistical techniques to remove instrumental systematics. In this thesis, I present two robust methods to extract the planetary signal out of instrumental and astrophysical noise with no required knowledge of the data or instrument itself. The first method uses Fourier analyses to enhance the astrophysical signal in a condition of low signal-to-noise. It benefits from the sparsity of the signal itself in the Fourier domain. I tested the procedure on ground-based data of the transit of HD-189733b, presenting for the first time the planetary spectrum at 0.94-1.4 micron. The second one makes use of a supervised machine-learning algorithm to de-trend the long- term stellar activity. The algorithm constrains a probability distribution over the function space by using a Gaussian Process prior and conditioning it with the data. This allows the extrapolation of the long-term information even when the star is not continuously monitored. I proved the effectiveness of the technique by applying it to the high precision photometric lightcurves of the NASA/Kepler space observatory. Both these two methods can be applied to other datasets recorded by different instruments. In particular I plan to analyse stellar light curves observed by CoRoT and Spitzer or ground-based facilities. Given the experience I have developed with the exoplanet spectra observed by IRTF I will be in a strong position to study the advantages and pitfalls of future facilities such as James Webb Space Telescope and the ELT.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available