Reliable predictions of the effects changing climate and disturbance regimes will have on forest ecosystems are crucial for effective forest management. Current fire and climate research in forest ecosystem and community ecology offers data and methods that can inform such predictions. However, research in these fields occurs at different scales, with disparate goals, methods, and context. Often results are not readily comparable among studies and defy integration. We discuss the strengths and weaknesses of three modeling paradigms: empirical gradient models, mechanistic ecosystem models, and stochastic landscape disturbance models, and propose a synthetic approach to multi-scale analysis of the effects of climatic change and disturbance on forest ecosystems. Empirical gradient models provide an anchor and spatial template for stand-level forest ecosystem models by quantifying key parameters for individual species and accounting for broad-scale geographic variation among them. Gradient imputation transfers predictions of fine-scale forest composition and structure across geographic space. Mechanistic ecosystem dynamic models predict the responses of biological variables to specific environmental drivers, and facilitate understanding of temporal dynamics and disequilibrium. Stochastic landscape dynamics models predict frequency, extent and severity of broad-scale disturbance. A robust linkage of these three modeling paradigms will facilitate prediction of the effects of altered fire and other disturbance regimes on forest ecosystems at multiple scales and in the context of climatic variability and change.