Simulation models of disturbance and succession are being increasingly applied to characterize landscape composition and dynamics under natural fire regimes, and to evaluate alternative management strategies for ecological restoration and fire hazard reduction. However, we have a limited understanding of how landscapes respond to changes in fire frequency, and about the sensitivity of model predictions to assumptions about successional pathways and fire behavior. We updated an existing landscape dynamics model (LADS) to simulate the complex interactions between forest dynamics, fire spread, and fire effects in dry forests of the interior Pacific Northwest. Experimental model runs were conducted on a hypothetical landscape at fire rotations ranging from 5 to 50 years. Three sensitivity analyses were carried out to explore the responses of landscape composition to (1) parameters characterizing succession and fire effects on vegetation, (2) the probability of fire spread into different successional stages, and (3) the size and spatial pattern of static fire refugia. The area of old open-canopy forests was highest at the shortest fire rotations, and was particularly sensitive to the probability of stand-replacement fire in open-canopy forests and to the fire-free period required for ingrowth to occur in open-canopy forests. The area of old closed-canopy forests increased with lengthening fire rotation, but always comprised a relatively small portion of the landscape (<10%). The area of old closed-canopy forests increased when fire spread was more rapid in open-canopy forests than in closed-canopy forests, and when the physical landscape incorporated large fire refugia with low fire spread rates. Old closed-canopy forests appear to comprise a relatively minor landscape component in mixed-severity fire regimes with fire rotations of 50 years or less. However, these results are sensitive to assumptions about the spatial interactions between fire spread, landscape vegetation patterns, and the underlying physical landscape.