Modeling a Disease‐Caused Patch Disturbance: Oak Wilt in the Midwestern United States

Abstract

Sparse data and inadequate methodology have limited our ability to characterize disease—caused forest disturbances and their effects on plant communities. We describe the formulation of an empirically derived, stochastic simulation model that predicts tree—by—tree mortality due to oak wilt, an important disease in central United States hardwood forests. Oak wilt has two means of spreading: through root grafts between adjacent trees, and by insect vector over longer distances. The model describes the level of spread from outside the woodlot, the level of within—plot vector spread, and the probabilities of root graft transmission between pairs of trees. Also included are the time between consecutive tree deaths and the spatial characteristics of patch spreading. The model generates stochastic mortality which is highly variable even in runs having identical parameters. As parameters are varied, average mortality is found to be highly sensitive to changes in parameters specifying insect vectors spread in general, and new introductions to a woodlot in particular. This supports the hypothesis that oak wilt dynamics are constrained by the rare occurrences of woodlot introductions. Highly variable damage, as seen both in the field and in model results, follows from this constraint and the positive feedback between the two types of local spread. The model results also explain regional differences in the amount of damage to oaks and average size of disease pockets (foci). Simulations using data from southern Wisconsin stands predict serious damage in both dry and dry—mesic forests, but in view of the advance reproduction, oak wilt disturbances seems likely to shift species in only dry—mesic forests.