2020 ESA Annual Meeting (August 3 - 6)

PS 3 Abstract - Modeling plant distribution patterns across Taiwan

Christopher M Jones, INVASIVE PLANT SCIENCE LAB, Researcher, Logan, UT
Modeling plant distribution patterns across Taiwan

Christopher M. Jones, Stephen L. Young, Guo-Zhang M. Song

Background/Question/Methods
Tropical islands contain a diversity of plants that are impacted by environmental and anthropogenic factors. The occurrence and subsequent spread of invasive plants can reduce island biodiversity and lower ecosystem function. Understanding what causes and leads to invasive plant dominance in these systems is important for developing effective conservation practices. We analyzed data from an island wide survey conducted on Taiwan (2009-2012). Vegetation data was collected systematically using a 1 km grid with 10 m2 quadrats. We ran analyses for invasive and non-invasive cover, using survey data (n = 216,065) with assessment of geographic position against road, river/lake, bridge, tunnel, and railway density along with precipitation, elevation, temperature, and soil type as predictor variables (factors). Species distribution models using Random Forests were used to analyze which factors have the most influence on a given species distribution. To build each model, we partitioned data into training and testing sets. We evaluated variables by the Gini Index to determine which variables contributed the most to each model’s predictive power and then tested each model against the remaining data to determine goodness of fit.

Results/Conclusions
We recorded 2,477 plant species of which 1,965 (79.33%) were non-invasive to Taiwan. 512 (20.67%) species were invasive. The most common non-invasive species included Eleusine indica (L.) and Oxalis corniculata (L.). The most common invasive species included Bidens alba (L.) DC. var. radiata (Sch. Bip.) and Chloris barbata Sw. We found the strongest factors for determining species distribution for both invasive and non-invasive species are precipitation, elevation, temperature, and road density. An understanding of the effects of predictor variables, which include a range of environmental and anthropogenic, are important to assessing vegetation changes. As urban areas continue to expand, the threat to natural areas includes loss of biodiversity both by direct pressures and indirect ones, such as invasive plants.