Bushfires are a natural hazard and necessary disturbance of Australian ecosystems as well as many others around the world, where they may also be known as wildfires. However, a small portion of these fires can lead to widespread destruction of built and natural environments resulting in human and ecological disasters. Research in the pursuit of better understanding of the physical dynamics of these events has frequently encountered a dearth of observational data, especially from events that cause most impact. This is particularly true of fire behaviour that involves interaction with the atmosphere, and is most relevant when this interaction takes place on the scale of a thunderstorm, known as storm scale (meso-gamma).The process of pyroconvection occurs when fire-released heat, moisture and/or aerosols induce or augment convection in the atmosphere. This fire-atmosphere coupling may lead to extreme rates of fire spread, as well as rates and distances of spotting - ignitions from ignited debris ahead of fire. Under favourable conditions, pyroconvection may create pyrocumulus which may develop further into a pyrogenic thunderstorm described by the cloud type of pyrocumulonimbus. The lifecycle and thermodynamic properties of these bushfire initiated and augmented thunderstorms are poorly understood. Despite this, they present substantial risk to firefighters and the public both on, and in proximity to, the fire ground through enhanced potential for ember transport and associated spotfire ignitions, pyrogenic lightning and near surface wind modifications that feeds back to fire behaviour.Meteorological radar has been demonstrated in some studies as a potentially useful tool for profiling the microphysics and thermodynamics of bushfire plumes and the associated atmosphere-fire coupling. This thesis aimed to quantify fire behaviour and bushfire thunderstorm lifecycle with radar. A synthesis of knowledge on the theory of radar for the analysis of wildfire is presented and the new term ‘pyrometeor’ is introduced to describe the range of pyrogenic scatterers observed by radar. The methodology and results of a three-year field campaign (Bushfire Convective Plume Experiment) is detailed. This experiment tested the hypothesis of a dedicated ground based deployable radar platform to describe fire behaviour and fire-atmosphere interaction. Based on the results of the experiment, a machine learning method for the detection and tracking of pyrometeors is presented. Storm-scale fire-atmosphere interaction is analysed from direct observations of a fire-initiated and augmented thunderstorm that occurred during the experiment. The findings of the Bushfire Convective Plume Experiment and a review of observational studies of fire with radar are synthesised to combine the theory of fire and radar with observations. The review of observational case studies is compared across plume microphysics, plume thermodynamics and deep pyroconvection, and operational usage of radar to monitor wildfire. Recommendations are provided for methods and datasets to advance the application of radar for observing and understanding bushfire – atmosphere interaction, including for plume microphysics and bushfire thunderstorms.