The growing demand for energy in the future will necessitate the production of natural gas from fields which are located farther offshore, in deep water and in very cold environments. This will confront us with difficulties in ensuring continuous production of the fluids (natural gas, condensate and water) emerging from a natural gas well. Often, all the three phases are transported together through a multiphase flow pipeline to processing facilities onshore. The natural gas production pipelines that carry the wellstream fluids from the subsea wells to the processing facilities are designed using engineering multiphase flow models. It is known that the currently available flow models cannot predict the most important flow parameters such as pressure drop and liquid holdup with sufficient accuracy when the gas production is low. At these conditions, the liquid accumulates in V-sections (lower elbows, low spots) of a pipeline, which are present because the pipeline profile follows the undulations in the seafloor topography. When the flow is low, the existing engineering models show large uncertainty in predicting the shear stresses, particularly at the gas-liquid interface. This uncertainty leads to the inaccurate prediction of, for example, the liquid holdup, which can cause production problems connected with liquid slugs, corrosion and the formation of gas hydrates. In this Thesis, the problem of liquid accumulation in an undulating pipeline is studied both at a more general, macroscopic level in a flowloop that contains a V-shaped section, as well as on a detailed, fundamental level in a straight horizontal pipe. Both configurations, however, have the same flow pattern: stratified gas-liquid flow. The main issues addressed in the V-section setup are the conditions of liquid accumulation and removal in the case of zero net liquid flow with gas flowing over a stagnant liquid pool, as well as the appearance of multiple steady-state solutions in the two-phase models at these conditions. The detailed measurements in the horizontal setup aimed at simultaneously capturing the velocities in both phases in the entire streamwise cross-section of the pipe and the position of the gas-liquid interface. This provided detailed information on waves and turbulence, which are the main phenomena in stratified flow. The occurrence of multiple solutions in stratified flow models was studied by applying a steady state and a transient model to various conditions for which lab experiments exist and by verifying the structural stability of the obtained solutions. It was found that the applied transient model (supplied with the criterion for structural stability) can qualitatively predict the measurements in zero net liquid flow, at conditions where hysteresis occurs and in experiments with a holdup discontinuity. Based on the comparison with available experimental data, it was concluded that hysteresis can only occur in fully laminar flow of both phases, and it is not expected to occur at typical field conditions. However, to achieve a better quantitative agreement with the measurements, the closure relations used in the model need to be improved. Measurements of the gas velocity, liquid holdup and pressure drop in zero net liquid flow were performed in a setup containing a V-shaped section. It was shown that the critical gas velocity (i.e. the minimum gas velocity at which the liquid is removed from the low spot) and pressure gradient increase with increasing inclination angle and with increasing liquid density and viscosity, while the liquid holdup stays approximately the same. The results were compared to the predictions of a mechanistic flow model, which was modified to account for the recirculation in the stagnant liquid layer at critical conditions by employing a theoretical solution for the wall shear stress at laminar flow conditions of the liquid. Good agreement was found between the measurements and the simulations. Stratified two-phase flow of air and water was measured with Particle Image Velocimetry (PIV) in a horizontal transparent pipe, with a laminar, transitional or turbulent liquid, and a smooth or a wavy interface. An advanced experiment was designed and built, which uses two lasers and three cameras to simultaneously measure the liquid velocity, interface shape and gas velocity. The data were time- and phase-averaged to obtain detailed and accurate insight into the turbulent and wavy structures. The cases with a smooth interface were shown to approximately follow the velocity laws valid in single-phase flows. The wavy region of the flowmap (constructed with superficial gas and liquid velocities at the axes) had waves which are asymmetric, with gravitational and capillary forces of similar magnitude. The linear wave theory provided a good approximation of the wave-induced velocity profiles, although the wave non-linearity caused a deviation close to the interface. The separation of wavy and turbulent motion was, however, only partly successful due to the wide range of wavelengths and wave heights in all the wavy cases. The laminar-to-turbulent transition of the liquid phase in stratified gas-liquid flow was also studied. The boundaries of transition were determined in both the smooth and the wavy region of the flowmap. In both regions, the Reynolds number at the start and at the end of transition decreased with increasing gas flowrate. The two occurring wave patterns (labelled `2D small amplitude' and `3D small amplitude' waves) corresponded to the capillary-gravity and the gravity-capillary solutions of linear wave theory. This led us to recast the flowmap of the wavy region into Weber number - Froude number coordinates, which in turn provided a physical interpretation of the interaction between the developing turbulence and the changing wave patterns. Finally, the interfacial characteristics and the velocities were investigated in both phases of stratified flow in two wave patterns: `3D small amplitude' and `2D large amplitude' waves. The 2D LA waves (corresponding to gravity waves) had higher and longer waves, that changed the liquid velocities in almost the entire liquid layer. The 3D SA wave pattern (corresponding to gravity-capillary waves) had smaller and shorter waves whose influence was limited to only a part of the liquid height. The effect of the two wave regimes on gas phase velocities, however, was rather similar. In all cases, waves produced an increase in the Reynolds stresses in the air close to the interface, which was linked to the occurrence of boundary layer separation at the interface. The occurrence of separation correlated well with the wave properties. The results of the current study provide a step forward in understanding stratified gas-liquid flow through straight pipes. The physical explanations and insights of the measured phenomena can be used to develop new correlations for the engineering flow models, which is the type of models that are currently widely applied in the industry. Furthermore, a high-quality experimental database for an elementary two-phase pipe flow configuration has now been established. This database can be used for the improvement and validation of more advanced models for such two-phase flows. In particular, these are Computational Fluid Dynamics (CFD) models of increasing complexity, such as Reynolds-Averaged Navier Stokes (RANS), Large Eddy Simulation (LES) and Direct Numerical Simulation (DNS).