Forest ecosystems store more carbon than the atmosphere and harbour the majority of the world's biodiversity, yet their response to changing climate is uncertain. Forest simulation models make landscape-level predictions of forest dynamics by scaling from key tree-level processes, but models typically have no climate dependency. In this thesis I demonstrate how large-scale national inventories combined with improvements in computational methods mean that models that incorporate the climate dependency of demographic processes may be parameterised at regional scales. In Chapter One I outline historical approaches to modelling forest dynamics and present a discussion of competing methods of parameterisation and model selection. In Chapter Two I present a model of individual tree mortality in the eastern United States which incorporates species, climatic and competitive effects parameterised using Markov Chain Monte Carlo methods. The remainder of the thesis concentrates on modelling Spanish forest dynamics, so in Chapter Three I present a brief introduction to Spanish forest ecology. In Chapter Four I examine how aboveground allometry - the scaling of tree height and crown shape - varies with climate and competition in Spain for 26 species. Hierarchical modelling suggests that scaling theories based on wood properties do not explain differences between species, but climatic factors, and in particular hydraulic limitations, do. In Chapter Five I parameterise a model of recruitment in Spanish forests using Approximate Bayesian Computation, a novel computational method which allows parameterisation of individual-based models without individual-based data, and demonstrate that it produces ecologically reasonable results. Chapter Six presents a forest dynamics model parameterised for the major native species in Spain and tests whether it is able to reproduce observed species-climate distributions. Finally, in Chapter Seven I discuss the main findings of the thesis and avenues for extending this research.