In Chapter 2 of this thesis I present my database of molecular absorption cross sections. These were developed using public molecular transition line-lists (from the ExoMol group). I use them to find limitations in the modelling of exoplanet atmospheres due to pressure broadening. Pressure broadening, where collisions between molecules in atmospheres cause a Lorentzian broadening of molecular transitional lines, is little understood in the field. In this chapter I consider its effects on real exoplanet atmosphere observations, both with current and future instruments. I show that pressure broadening may affect future observations of exoplanets in the JWST era. Pressure broadening primarily affects cooler, small exoplanets such as Earth analogues.

In Chapter 3 I present the pipeline I have developed to reduce HST WFC3 spectra of exoplanet hosts during transits to create transmission spectra. This code corrects several instrumental systematics, from varying dark signal in the detector to subpixel shifts in the target position over time. By creating a pipeline to process all targets, regardless of observing strategy, systematics are dealt with uniformly and different planets’ spectra can be meaningfully compared. I show that the height of the water feature in 30 unique exoplanets’ transmission spectra is strongly correlated with the most simplistic absorption model. I use this to predict a list of the best future targets for observations with HST WFC3 to find water.

In Chapter 4 I discuss my work with the stellar spectra from WFC3, which utilise the sub-pixel shifts in target position to oversample the spectra and increase the resolution. I have compared these exoplanet host stellar spectra with stellar models to investigate how well stellar atmosphere models describe the near IR. I find a small discrepancy in temperature when WFC3 alone is used to assess the stellar temperature, particularly with cooler stars. I attribute this firstly to an error in the WFC3 sensitivity curve and secondly to an inaccuracy in models of cool, small stars due to molecular absorption.

In Chapter 5 I present my work on K2 light curve data using machine learning to find young stellar objects that display unusual, transit-like behaviour. These objects are known as dipper stars due to their distinctive occultations with depths of 10-50% in flux and very fast orbital periods of a few hours to a few days. Such large occultations are difficult to explain and are currently attributed to material at the inner edge of the protoplanetary disk. This behaviour is often variable and aperiodic, suggesting that the occulting material is changing in morphology on the time scale of a single orbit. Using python’s scikit-learn I have developed a code that utilises a Random Forest algorithm to classify stars in K2 Campaign Field 2 and distinguish these objects from other types of variables, such as eclipsing binaries and pulsating stars. This method has proved very successful and has allowed me to nearly quadruple the number of known dipper candidates in the Upper Scorpius and Rho Ophiuchus regions.