Timber could revolutionize the construction industry by allowing rapid, cost-efficient, and sustainable construction of high-rise buildings. However, insufficient knowledge of the fire performance of timber hinders its uptake. The structural performance of timber in a fire is evaluated by the char depth after standardised heating inside a furnace. This standardised heating only represents fires in small compartments (floor area<100 m2). In larger compartments, char depths under uniform fires have been studied sparsely (area < 500 m2) and under non-uniform fires not at all (area>500 m2). This thesis aims to understand the charring behaviour of wood under uniform and non-uniform fires. To this end, a novel charring model of timber was developed across scales. At the microscale, a novel kinetic model reproduces over 80 experiments from the literature accurately. The model was developed using novel methodologies to arrive at an appropriate reaction scheme and identify experiments free of heat and mass transfer effects. At the mesoscale, this kinetic model is coupled with a heat and mass transfer model. The combined model reproduces blindly experiments under a wide range of conditions. With this model, I unified the three existing theories on the role of chemical kinetics. Furthermore, I demonstrate that chemical kinetics is key to reproduce transient charring behaviour, but that the variability in charring rates of different softwoods stems from their difference in material properties. At the macroscale, I predict that in non-uniform fires char depths are up to 34 % smaller than in uniform fires, but the strength loss ahead of the char front can be up to 129 % larger than in uniform fires. When taking the latter into account, the strength decay under uniform and non-uniform is comparable. Overall, this thesis provides a fundamental understanding of charring and enables engineers to improve their current design tools.