A species distribution model for the shade-tolerant, invasive plant Japanese stiltgrass

Background/Question/Methods

      Exotic plant invasions have huge ecological and economic ramifications. Our understanding of exotic plant invasions, however, is mostly based on shade-intolerant species. We studied the invasion patterns of Microstegium vimineum (Trin.) A. Camus (Japanese stiltgrass), a widespread shade-tolerant exotic plant species found throughout much of the eastern United States.  Where it spreads, Microstegium profoundly affects ecological functions, altering soil chemistry and hydrology, displacing native flora, and thus reducing biodiversity. Successful control of this noxious species is highly contingent on early detection before large seed banks are established.  As such, identifying areas at risk for invasion would allow conservation managers to better apply resources for maintenance and control.  In this study, we developed a spatially applicable, species distribution model for Microstegium based on field surveys of 160 points throughout the Chesapeake Bay lowlands. Within a geographic information system (GIS), we assessed landscape features within three categories: abiotic, habitat, and anthropogenic.  Variables selection was based on life-history traits of Microstegium as well as the literature on Microstegium invasions. In order to capture how Microstegium responds to the landscape, we considered different extents for each variable (i.e. a radius of 90-m, 270-m, 540-m, 1-km, 2-km, 3-km), which allowed us to assess whether a feature was most relevant at a local or landscape level. Using an information-theoretic modeling approach, we assessed all model combinations within each category. 

Results/Conclusions

     Ultimately our final model consisted 11 landscape features and metrics. Some of the most predictive variables were proportion of clear-cut lands within a 2-km radius, Atlantic mesic forests within a 270-m radius, and dry-mesic oak forests within a 2-km radius, distance to roads, distance to water bodies and hunting. Using independent survey data for validation, we found our model to have excellent predictive value for sites unoccupied by Microstegium, but only marginal predictive value invaded sites.  Our study adds to a growing body of evidence that human features within the landscape influence the distribution of invasive exotic plants. Furthermore, our findings illustrate that large scale processes may ultimately be more relevant to predicting invasions. We will sample additional sites this summer to improve model predictions.