Star-forming emission line galaxies (ELGs) are being used as tracers of the dark matter distribution at z~1. Model [OII] emitters have luminosity functions in reasonable agreement with observations. ~95% of model ELGs are centrals hosted by haloes with M>10^10.5Msunh^-1. The mean halo occupation distributions of central ELGs are far away from a canonical step function and we have proposed a split contribution from disks, asymetric Gaussian, and spheroid central ELGs, soft step function. Half of the ELGs live in filaments and a third in sheets. This distribution is similar to that of star-forming galaxies selected in different ways. ELGs, star formatio rate and [OII] luminosity selected samples with equal number density, have similar large scale bias but their clustering below separations of 1h⁻¹Mpc is different.

Galaxies are thought to form within haloes of dark matter, whose gravity allows the galaxies to exist. The formation and evolution of galaxies is affected by a multitude of other processes besides gravity and computational modelling is the only way we can attempt to understand all these processes. In this work we presenta a new development of the GALFORM semi-analytical model of galaxy formation and evolution, which exploits a Millennium Simulation-class N-body run performed with the Wilkinson Microwave Anisotropy Probe 7 cosmology. We use this new model to study the impact of the choice of stellar population synthesis (SPS) model on the predicted evolution of the galaxy luminosity function.

The study of high redshift galaxies, z>2.5, is fundamental in unveiling initial phases of galaxy evolution. However, the interpretation of the observational data is dramatically affected by dust attenuation. We explored the impact of dust on UV luminosities using a published model of galaxy formation. I found that UV colours are very sensitive to the shape of the extinction curve used as an input for the radiative transfer model that calculates dust attenuation, stressing the difficulty of using the UV continuum slope as a tracer of dust attenuation without an understanding of the dust characteristics and geometrical distribution in high redshift galaxies.

The observed abundance of massive, high redshift galaxies with very red colours posed a major challenge to hierarchical clustering cosmologies. I presented the first example of a model which could explain these observations, whereas previous work had failed dramatically, by an order of magnitude. The key to understanding these objects proved to be the inclusion of heating of gas following the accretion of matter onto super-massive black holes at the centres of galaxies.

• Tabulated compilation of data from the literature: EROs number counts.

Following the breakthrough above, I studied the clustering of Extremely Red Objects (EROs) as a further test of the model. The model predicts a very different distribution of galaxies across halos of different masses than is typically assumed, which has a major influence on the clustering predictions. These predictions agree with the currently available observations, once the small size of the samples is taken into account.

The internal properties of galaxies contain information of their recent evolution. Our analysis relates, on average, steep colour gradients to a higher presence of young stars within a galaxy. Close pairs of interacting galaxies were found to present bluer cores than the average population. We also found that nuclear activity is a marginal driver for creating steep colour gradients in massive galaxies.