Use of simulation-based statistical models to complement bioclimatic models in predicting continental scale invasion risks

Authors: MUTHUKRISHNAN, R - University Of Minnesota; JORDAN, NICHOLAS - University Of Minnesota; Davis, Adam; FORESTER, JAMES - University Of Minnesota

Submitted to: Diversity and Distributions
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 9/10/2018
Publication Date: 3/1/2019
Citation: Muthukrishnan, R., Jordan, N.R., Davis, A.S., Forester, J.D. 2019. Use of simulation-based statistical models to complement bioclimatic models in predicting continental scale invasion risks. Diversity and Distributions. 21(3):847-859. https://doi.org/10.1007/s10530-018-1864-3.
DOI: https://doi.org/10.1007/s10530-018-1864-3

Interpretive Summary: Efforts to reduce potential invasion risks associated with novel crop introductions can benefit from quantifying where the risks are greatest. We present a case study, with a major biomass crop, Miscanthus x giganteus, for how landscape characteristics can be factored into large-scale analyses of invasion risk. We do this by extending simulation models with statistical modeling tools to the scale of the entire United States, then combine simulation-based predictions with those from bioclimatic methods. Lastly, we evaluate how potential risks change for other potential invasive crops that differ in the strength of invasiveness traits. We produced a map of simulation-based invasion risk for M. x giganteus that overall shows the highest risks in the Great Plains region and around California’s central valley. The regions with the highest simulation-based estimates of invasion risk generally do not overlap with areas of high bioclimatic invasion risk except in California's central valley and the southeastern edge of the Great Plains. Simulation-based estimates of risk largely align with the habitat composition of landscapes but spatial structure can also have subtle influences and certain landscapes are more susceptible than others to particular invasiveness traits. Combining risk analyses based on biology and climate can offer more detailed invasion predictions than either approach alone.

Technical Abstract: Invasive species represent one of the greatest risks to global biodiversity and economic productivity of agroecosystems. The development of certain novel crops—e.g., herbaceous perennial biomass crops—may create a risk of novel invasions by these crops. Therefore, potential benefits and risks need to be weighed in making decisions about their introduction. Ideally, such a weighing will be based on good estimates of invasion risks in realistic scenarios that incorporates both climatic and landscape level information. We present a case study, with a major biomass crop, Miscanthus x giganteus, for how landscape level spatial patterns and biotic interactions can be efficiently incorporated in large-scale analyses of invasion risk. We do this by extending simulation models with statistical modeling tools to the scale of the entire United States, then combine simulation-based predictions with those from bioclimatic methods. Lastly, we evaluate how potential risks change for other potential invasive crops that differ in the strength of invasiveness traits. We produced a map of simulation-based invasion risk for M. x giganteus that generally shows the highest risks in the Great Plains region and around California's central valley. These patterns are generally consistent across response metrics and with enhanced invasiveness traits, though there are subtle geographic differences. The regions with the highest simulation-based estimates of invasion risk generally do not overlap with areas of high bioclimatic invasion risk except in California's central valley and the southeastern edge of the Great Plains. Simulation-based estimates of risk largely align with the habitat composition of landscapes but spatial structure can also have subtle influences and certain landscapes are more susceptible than others to particular invasiveness traits. The combination of bioclimatic and more mechanistic approaches can offer more nuanced predictions of invasion risk than either approach alone.