Predicting the spread of the noxious rangeland weed medusahead in California: Importance of dispersal vectors

Background/Question/Methods

Understanding what controls the spread of invasive species is essential to identify areas susceptible to invasion so they can be targeted for early detection surveys and management. Most species distribution models used for this purpose focus only on abiotic factors and ignore dispersal; however, dispersal vectors may be key in promoting or restricting the spread of invaders into certain regions. Here we test whether dispersal factors, in addition to abiotic factors, are important in predicting the spread of a noxious rangeland weed in California. Medusahead (Elymus caput-medusae) is an aggressive annual grass that reduces land value by decreasing forage production, outcompetes native vegetation, and creates a wildfire hazard. Medusahead seeds have long barbed awns, which allow it to attach and disperse via animals including cattle. We hypothesize that adding dispersal parameters to distribution models will increase model fit and alter predictions of where medusahead is likely to invade. Using logistic regression, we first fit an abiotic model including annual precipitation, mean annual temperature, soil texture, slope, and aspect. Then we fit another model adding two dispersal parameters: road density and beef cattle density. We then extrapolated both models to all of California to determine how adding dispersal may alter invasion predictions.

Results/Conclusions

In abiotic models, precipitation, temperature, soil texture, and slope were important for medusahead distribution. Adding dispersal parameters increased model fit (assessed by AIC); beef cattle density was positively associated with medusahead presence and road density exhibited a hump-shaped relationship. However, adding dispersal parameters did not increase model predictive power (assessed by AUC). Both models predict that wetter, hilly regions in Northern California (the North Coast and Sierra Nevada) are likely to be invaded, whereas it is very unlikely for medusahead to invade in hotter, drier regions of Southern California (the Central Valley and desert). The models also highlight some areas in the Central and South Coast with high habitat suitability. Overall, these results suggest that limiting dispersal vectors may help reduce the spread of this invasive grass. For example, when transporting cattle from rangeland that is invaded, measures can be taken to reduce seed transport such as quarantining cattle for a period of time so seeds drop off prior to transport. Prevention is the most cost-effective method of invasive species control; these methods can be easily applied to other invaders and regions to assess invasion risk and inform Early Detection Rapid Response management.