Risk analysis for invasive plants in Iowa: Development of risk-assessment models and the perceptions of stakeholders

Kapler, Emily
Major Professor
Janette R. Thompson
Committee Member
Journal Title
Journal ISSN
Volume Title
Research Projects
Organizational Units
Journal Issue
Natural Resource Ecology and Management

Risk analysis is a decision-making framework used to evaluate risk, or the probability of harm given an exposure. Invasive plants pose risks to natural ecosystems because they can significantly alter ecosystem function and decrease native species diversity. Managing these risks comes with many challenges, and may take many forms. This thesis examines two primary aspects of risk analysis: (1) the validation and development of risk-assessment models that can predict the naturalization of non-native woody plants; and (2) the perspectives of stakeholders on invasive plants, risk-assessment models, and nature relatedness.

Good power and accuracy are primary goals of risk-assessment models to predict the naturalization of non-native plants. Testing previously developed models with a new set of species, or external validation, is one way to ensure these goals are met. Validation of four risk-assessment models - previously designed to evaluate the risk of naturalization for woody plants in Iowa - had mixed results when applied to a new selection of species. Classification rates ranged from 62.1 to 93.1%, biologically significant error rates from 11.5 to 18.5%, and horticulturally limiting error rates from 11.1 to 38.5%. Another way to reach the goal of good power and accuracy is to develop new risk-assessment models based on different statistical techniques. Creation of a new risk-assessment model for Iowa using a random forest approach yielded a high initial classification rate (92.0%), no biologically significant errors and 8.7% horticulturally limiting errors. When validated, the random forest model maintained a relatively high classification rate (82.8%), but produced one biologically significant error (4.2%) and more horticulturally limiting errors (29.2%). Differences in performance among the various models were not always significant due to the small sample size of the validating data set (n = 29), but the random forest model shows promise as a new technique to sort benign non-native woody plants from naturalizing or invasive ones.

Implementation of risk-assessment models will depend on the cooperation of diverse stakeholder groups. Addressing their perspectives on invasive plants is therefore an important component of the risk analysis process. Stakeholders in Iowa who will be affected by or involved in implementation of risk-assessment models agreed that invasive plants are a problem that we have a responsibility to manage. Respondents had a strong sense of their personal relatedness to nature, which played some role in shaping their concern about invasive plants. Support for use of risk-assessment models was high, though respondents expressed some concerns about accuracy. Respondents were willing to accept biologically significant error rates of 5 to 10%, and horticulturally limiting error rates of 10 to 20%; overall they found biologically significant errors to be more important than horticulturally limiting errors. Because stakeholders are largely in agreement about invasive plants and their management, risk management efforts in Iowa that incorporate risk-assessment models are more likely to be successful than if stakeholder groups disagreed. Mixed results of efforts to validate existing risk-assessment models suggest the need for further refinement.