Predicted range expansion of the invasive plant Leucaena leucocephala in the Hengchun peninsula, Taiwan

Abstract

We integrated correlative (species distribution) and process-based (agent-based) modeling approaches—to predict future range expansion of an invasive plant (Leucaena leucocephala) in the lowland rainforest and tropical coastal forest ecosystems of the Hengchun peninsula, Taiwan. We simulated range expansion to the year 2027 using a spatially-explicit, agent-based model representing invasion rate as a function of habitat quality and propagule pressure. We developed an index of habitat quality by relating 1988 plant distribution data to geo-referenced data on climatic conditions, landscape features, and anthropogenic factors via logistic regression. We represented propagule pressure using a lognormal dispersal kernel. We evaluated model performance by simulating range expansion from 1988 to 2007 and comparing simulated distribution patterns to those observed in 2007. Results of logistic regression indicated that L. leucocephala is more likely to occur in warm, dry areas containing a higher percentage of natural landscape (forest or grassland), areas adjacent to the forest edge, and areas disturbed by human activities such as a main road or an abandoned sisal plantation. Model evaluation showed that the prediction performance is excellent with AUC > 0.9 and Mantel’s r = 0.77. Our results indicated that L. leucocephala will continue from the western portion toward the southern and central portions of Checheng township, and throughout the southern portion of Hengchun and Manjhou townships, with about 2500 ha of new area invaded within the next 20 years. Our predictions should allow managers to develop proactive management plans for the areas most likely to be invaded.