Exploring plant characteristics for constructing a pre-border weed risk assessment for China.

Biological invasions has caused significant damage to the ecological environment and economy of the world. Pest risk assessment is the most cost-effective means of preventing biological invasions to identify potentially suitable indexes for constructing pre-border weed risk assessment methods for China, we screened 80 metrics derived from 53 plant characteristics known to be related to invasive alien plants in other parts of the world and tested whether these metrics differed significantly between two groups of 103 invasive alien plants and 107 non-invasive plants in China. The results showed significant differences in 30 characteristics between invasive and non-invasive plants in China. Compared to the non-invasive plant group, the invasive plant group in China had a greater proportion of (1) plants native to the Americas, (2) plants belonging to the Asteraceae family, (3) polyploid plants, and had (4) a smaller proportion of plants propagated asexually only. The 30 metrics with significant differences were selected for LASSO regression to develop a predictive model to determine how well the metrics could distinguish between invasive and non-invasive alien plants already present in China. Finally an optimal model with 18 metrics was screened out. The optimal model was able to accurately discriminate 75% of non-invasive plants and 90% of invasive plants on the test set. Therefore the present study screened a range of useful metrics for the identification of invasive plants in China, and the high discriminative power of our models indicates that the subset of 18 variables retained in the final model could be useful for establishing a pre-border invasive plant screening tool for China in the future. [ABSTRACT FROM AUTHOR]