This thesis describes a two-part study aimed at understanding the physical processes leading to the ignition of forest canopy fuels above a spreading surface fire. The study comprises a theoretical component focusing on the development of a crown fuel ignition model based on heat transfer principles (Chapter 1) and an experimental component aimed at comprehending how fuels, weather and fire behavior variables determine upward convective and radiative heat fluxes (Chapter 2). The crown fuel ignition model integrates the properties of the heat source as defined by the surface fire flame front and crown fuel characteristics, which determine the heat requirements for crown ignition. Fuel particle temperature increase was determined through a simplified energy balance equation relating heat absorption to fuel particle temperature. The final model output is the temperature of the crown fuel particles, which upon reaching ignition temperature are assumed to ignite. Model results indicate that the primary factors influencing crown fuel ignition are those determining the depth of the surface fire burning zone and the vertical distance between the ground and surface fuel strata in the fuel complex and the lower base of the crown fuel layer. The coupling of the crown fuel ignition model with models determining the spread of crown fires allows for the prediction of the potential of sustained crowning. A number of laboratory and outdoor experimental fires were instrumented to measure upward radiative and convective heat fluxes. No evidence was found to a preponderance of one heat flux process over the other. The fire behavior characteristics that were most related with the upward heat fluxes were reaction time and predicted flame height. No significant relationships were found between fire intensity measures, such as fireline and reaction intensity and various measures quantifying upward heat flux, namely peak and cumulative heat fluxes. The use of models to predict upward radiative heat flux and buoyant plume behavior showed no evidence of bias, although predictions showed some degree of variability. Analysis of the observed heat fluxes and model predictions indicates that the heat flux partitioning into convective and radiative processes is highly dynamic in time and space, and determined by fuel complex characteristics and burning conditions.