Predictive Spatial Model of Japanese Stiltgrass Spread
Management of invasive species is most effective when new colonizers can be detected and eliminated early, and when elimination of existing populations targets patches that have the greatest impact on the spread of the species. Because property managers have limited resources, we need more tools to predict where to search for new patches and to determine which patches are most important for spread dynamics. I am developing a GIS-based model to predict the spread of the highly invasive forest grass, Microstegium vimineum of Japanese stiltgrass. I have collected three years of data on the distribution and abundance of Microstegium at ten sites, surveying a total of 275 ha in south central Indiana. These sites include undisturbed, naturally-disturbed, and harvested forests, allowing us to compare spread rates across different disturbance types. I have determined correlations between Microstegium occurrence and various environmental factors (slope, aspect, light) and distance to dispersal corridors (roads, trails, streams) and disturbances. I then use Bayesian statistical analysis to determine the probability of Microstegium occurrence at a site, dependent on these variables. Slope and light are important determinants of Microstegium site suitability. Distance from roads, streams, and harvests very strongly affect the probability of invasion. On average, spread rate was higher at harvested sites than naturally-disturbed or undisturbed sites, with one harvested site having a 388% increase in cover over two years. However, the highest rate of spread at a single site was at a naturally-disturbed site where Microstegium cover increased by 474% over two years due to flooding. I am working on integrating these predictions with a GIS system to generate predictive spatial maps of likely Microstegium invasions. This model should permit property managers to better predict sites that are likely to be invaded and which patches they should target for control.