MODELING SPATIAL VARIATIONS OF THE INVASIBILIITY OF SLASH PINE FLATWOODS TO CHINESE TALLOW (Triadica sebifera) INVASION: MECHANISMS AND KEY FACTORS AT THE MICROSCALE

Restoration of endangered slash/longleaf pine flatwood ecosystems is of critical, regional importance to the biota of the Gulf coastal plain. One of the serious threats facing this ecosystem is the invasion of Chinese tallow (Tricdica sebifera), a community-transforming invasive tree species. We developed a conceptual model to decipher the mechanism of Chinese tallow invasion at the microscale based on community assembly theory and modeled the spatial variations of invasibility of slash pine flatwoods to Chinese tallow using a spatially-explicit point process model. The probability of Chinese tallow invasion was predicted through the polynomial functions of a set of identified filters (covariates) that were classified, respectively, as dispersal filters (distances to roads and seed sources), overstory filters (canopy closure, pine: hardwood ratio), and understory filters (grass cover, elevation) based on the roles they play in the invasion processes. To measure these filters (input for the model), 281 contiguous quadrats of 30 m² were set up with a total surveyed area of 0.86-ha. All overstory pine and hardwood species as well as tallow trees including seedlings, saplings and large trees were mapped and tree species and diameter at breast height (dbh, if present) were measured. Understory shrubs and grass/herbaceous species cover (%) and overstory canopy closure (%) were measured in each quadrat. The raster image (1-m resolution) of identified filters were interpolated based on field data or extracted from available DEM data for spatial modeling.

The identified filters vary spatially, and may facilitate or inhibit tallow invasion. Therefore the additive polynomial models of these identified filters provide a viable approach to predict the spatial variations of the probability of tallow invasion (invasibility). The full models including all filters, and the sub-models which included the dispersal filters and either of the overstory and understory filters seemed to have an adequate predictive power based on R², RSE and skewness (the scatter and qq-plots). In contrast, sub-models which included only one of the dispersal, overstory and understory filters failed to precisely predict the observed invasion outcomes. The results support the importance of reducing propagule pressure levels for effective tallow control and slowing of spread.