Christchurch, New Zealand, has a serious air pollution problem during the winter season, due to frequently occurring stagnant air masses and strong inversions, combined with increased emissions caused by burning wood and coal for domestic heating. During this research project an investigation was made of the role of meteorological processes affecting the dispersion of air pollution in the Christchurch environment. At the smaller end of the scale, analysis was conducted of the relationships between surface layer fluxes and land use in the Christchurch area, while at the larger scale drainage flows and interactions with advection were studied. A data collection system was established during the winter of 1995 to obtain a three dimensional picture of the atmosphere, being most detailed near the surface and in the Christchurch area. Two 10.5 m, two 21 m towers and two mobile weather stations were deployed in the area, combined with a network of existing weather stations. Tethered balloon flights with a sonde provided detailed profiles of temperature, humidity and wind up to 500 m, while pilot balloon soundings gave information of wind profiles up to 6000 m. Air pollution data were provided by the Canterbury Regional Council. Analysis of the heat fluxes in the Christchurch area indicated that during fine days in winter, the Bowen ratios over the urban areas were much higher than over the surrounding rural environment, which could have implications for dispersion of air pollution. Drainage flows from the Port Hills and the Canterbury Plains were observed together with their impact on the dispersion of air pollution in Christchurch. During clear nights very stable thermal conditions developed up to an elevation of 100 to 200 m. Above these stable layers strong low level jets were observed. Atmospheric waves occurred frequently and at times they had a dramatic impact on dispersion of air pollution. Simulations using the three dimensional prognostic model RAMS, version 3b, have shown that drainage flows can still develop under clear sky conditions with advection at a synoptic scale, due to decoupling. Verification of two case study days showed that simulated airflow patterns agreed reasonably well with observations, although some significant deviations occurred also. Dispersion of CO was simulated by using a Eulerian mode, and the results agreed reasonably well at the general level, although the results were less accurate at the detailed level. It is suggested that the overall performance of the simulations will be improved by adjusting and refining model settings and boundary conditions, and by the applications of inhomogeneous initialization procedures, cloud physics, an improved surface flux routine and dispersion studies using a Lagrangian or hybrid mode.