A simple model was developed which represented a rectangular, regularly-spaced planting of host trees divided initially into a diseased area and a healthy area. The purpose of the model was to investigate disease dynamics in a plantation over several years, specifically to evaluate the relative importance of dispersal and infection rates, the effect of inoculum source ageing, the impact of different source sizes and the effect of sanitation. Versions of the model were used to simulate gradients of incidence and severity in the horizontal plane as disease developed with time. Only the case of a within-year monocyclic disease was considered, i.e. infections initiated in one year did not produce inoculum until the next year. One such disease is witches' broom of cocoa (an airborne disease caused by the basidiomycete fungus Crinipellis perniciosa (Stahel) Singer). The probability of inoculum arriving at an infection court from a source was assumed to decline exponentially with distance from that source. Infection was modelled stochastically: an infection occurred if a calculated infection probability exceeded a uniformly-distributed random variable, generated for each possible infection event. The shapes of the gradients in both the incidence and severity models were defined by two parameters. An infection parameter determined the position of the disease front and a inoculum dispersal parameter, the slope of the front. The effects of reduced inoculum production as sources age was incorporated into the incidence model but had only a minor impact on disease spread. The size of the source area initially infected was found to be important only when numbers of diseased trees were few. A threshold, above which further increases in source area had little impact on subsequent dynamics, was reached rapidly. Using the severity model, phytosanitation was found to be effective in retarding disease spread only at very high efficiencies, often in excess of 90%.