Many geological applications involve the flow of multiple fluids through porous geological materials, e.g. environmental remediation of polluted ground water resources, CO2 storage in geological reservoirs and petroleum recovery. Commonly, to model these applications, the geological materials in question are treated as continuous porous media with effective material properties. Since these properties are a manifestation of what goes on in the pores of the material, we have to study the transport processes at the pore scale to understand why and how they vary over space and time in different rocks and under different conditions. As the high cost of acquiring and testing samples in many of these applications is often a limiting factor, numerical modelling at the pore scale is becoming a key technology to gain new insights in this field. This could be crucial in reducing uncertainties in field scale projects. The work presented in this thesis focuses on the investigation of two-phase flow in sedimentary rocks, and is an integrated numerical and experimental study. It deals primarily with two outstanding issues. First, image-based pore scale simulation methods have difficulties with representing the multiple pore scales in rocks with wide pore size distributions, due to a trade-off in the size and resolution of both modeling and imaging methods. Therefore, performing two-phase flow simulations in a number of important rock types, such as many carbonates and tight, clay-baring sandstones has remained an outstanding challenge. To tackle this problem, a new numerical model was developed to calculate capillary pressure, relative permeability and resistivity index curves during drainage and imbibition processes in such materials. The multi-scale model was based on information obtained from 3D micro-computed tomography images of the internal pore structure, complemented with information on the pores that are unresolved with this technique. In this method, pore network models were first extracted from resolved pores in the images, by using a maximal ball network extraction algorithm. Then, pores which touched regions with unresolved porosity were connected with a special type of network element called micro-links. In the quasi-static simulations that were performed on these network models, the micro-links carried average properties of the unresolved porosity. In contrast to most previous models, the new approach to taking into account unresolved porosity therefore allowed efficient simulations on images of complex rocks, with sizes comparable to single-scale pore network models. It was able to reproduce most of the behaviour of a fully resolved pore network model, for both drainage and imbibition processes, and for different pore scale wettability distributions (water-wet, oil-wet and different mixed-wet distributions). Furthermore, simulations on images of carbonate rocks showed good agreement to experiments. A sensitivity study on carbonate rocks and tight, clay-bearing sandstones produced results that were in qualitative agreement with experiments, and allowed to analyse how the two-phase flow behaviour of these rocks is influenced by their pore scale properties. The second issue which is treated in this thesis is related to the validation of pore scale models. Comparing predicted effective properties to experimentally measured values is useful and necessary, but is complicated by the typical difference in size between the model and the experiment. Furthermore, it does not always give a clear indication of the reasons for an observed mismatch between models and experiments. Comparing two-phase flow models to pore scale experiments in which the evolution of the fluid distributions is visualized is thus extremely useful. However, this requires to image the two-phase flow process while it is taking place in a rock, and it is necessary to do this with time resolutions on the order of tens of seconds and spatial resolutions on the order of micrometers. Previous experimental approaches used synchrotron beam lines to achieve this. In this thesis, we show that such experiments are also possible using laboratory-based micro-computed tomography scanners, which are orders of magnitude cheaper and therefore more accessible than synchrotrons. An experiment in which kerosene was pumped into a water-saturated sandstone is presented, showing that individual Haines jumps (pore filling events) could be visualized during this drainage process. Because the image quality is lower than at synchrotrons, care had to be taken to adapt the image analysis work flow to deal with high image noise levels. The work flow was designed to allow to track the fluid filling state of individual pores. The results indicate that the dynamic effects due to viscous and inertial forces during Haines jumps do not significantly impact the evolution of the fluid distributions during drainage, which may thus be adequately described by quasi-static models.