Fog has long been identified as one of the most difficult weather phenomena to forecast. With the increasing need of air, marine, and land transportation, the significance of the societal impact caused by fog has been growing in the recent decades. Numerical models are an essential tool to gain more insights of fog formation and dissipation processes and subsequently to improve fog forecasts. Numerous factors and physical processes influ- ence fog at a wide variety of spatial and temporal scales. The conditions favourable to fog formation and development are highly dependent on the area studied. Identifying the site-specific meteorological controls on fog is crucial to improve the understanding of local fog events and fog forecasts. While numerical models have been improved and up- dated constantly in recent years, they are still unable to represent all the scale-dependent physical processes relevant to fog formation accurately. Accurate fog forecasting remains difficult, especially in regions with complex and heterogeneous land surfaces.

This thesis focuses on the meteorological controls on radiation fog in Christchurch, New Zealand. Both observational data and numerical simulations were used in this thesis to understand the meteorological controls across various scales. Mesoscale (104 to 2 x 105 m) and orographically induced flows were found to modify the local atmospheric dynamics significantly. Furthermore, at microscale (10−2 to 103 m), several processes have impacts on fog formation, dissipation, and duration. This thesis considers soil moisture heterogeneity as an important part of land surface characteristics at microscale, which can lead to spatial heterogeneity in the surface energy balance and subsequently fog.

Developing a fog climatology is a crucial first step to understand the processes involved in local fog events. To achieve this, this study used 12 years of observational data obtained from the automatic weather station (AWS) operated at Christchurch international airport (CHA). Fog events over the 12-year period were identified. A novel fog type classification method based on the Modified Richardson number (MRi) was developed. The fog events identified were classified into five types of fog. A climatology was developed for each fog type to identify and understand their characteristics, including fog intensity, duration, and diurnal and seasonal variability of fog occurrence frequencies. Radiation fog is the predominant fog type in Christchurch and occurs most frequently during local winters (June, July, and August). In addition, the wind analysis showed a strong signal of oro- graphically induced drainage flow in association with radiation fog events. These results form the basis for understanding the dominant processes involved in radiation fog events in Christchurch.

To include mesoscale flows and local microscale land surface characteristics, suitable modelling tools were developed for conducting numerical fog simulation in Christchurch. This thesis used two numerical models: the Parallelised Large-Eddy Simulation Model (PALM), and the Weather Research and Forecasting (WRF) model. PALM can read data from mesoscale models and geospatial data sets as inputs, while no suitable modelling tools were available for Christchurch. Therefore, firstly, a Python-based tool, WRF4PALM, was developed to pass data from WRF into PALM. Two case studies were presented to assess the performance of WRF4PALM. Secondly, to utilise the complex and heterogeneous urban environment of Christchurch in PALM simulations, a static input generator was developed for geospatial data sets in Christchurch and New Zealand. Finally, a Python- based tool has been developed to process high-resolution (30 m) LANDSAT 8 observations into soil moisture data that can be used to initialise PALM simulations.

With the identified radiation fog characteristics and the developed modelling tools, a case study was carried out. A set of eight simulations was conducted at microscale (grid spacing of 81 m) and designed to understand the roles of land surface characteristics and soil moisture heterogeneity in radiation fog. The simulations included heterogeneous soil moisture derived from LANDSAT 8 imagery, topography, and land use. A readjustment method was applied to amplify the dry and wet signal from the heterogeneously distributed soil moisture. Several meteorological controls were identified. The important role of the local drainage flow was recognised in agreement with the climatological analysis. Within such a complex environment, the near-surface fog occurrence locations were not largely altered by soil moisture heterogeneity. While the soil moisture heterogeneity led to changes in fog duration, no pixel-to-pixel correlation was found between them. Spatial analysis was carried out using the watershed image segmentation (IS) method. The variations in the percentage of vegetation cover influence the evolution of surface fluxes at fog formation and dissipation over each watershed clustered area. This, consequently, was found to cause variability of the effect size of soil moisture on fog duration.

Overall, this thesis identified and investigated the meteorological controls on radiation fog from mesoscale to microscale. This thesis is the first to investigate fog climatology and typology in detail for Christchurch. The fog simulation case study is the first fog study carried out using soil moisture derived from high-resolution satellite imagery in addition to heterogeneous topography and land use. As such, this work established a basis and provided guidance for future fog research for Christchurch. Furthermore, both the WRF4PALM tool and the static input generator are open-source. While this work focused specifically on Christchurch, the tools and findings presented in this thesis have the potential to be applied worldwide.