Selecting the best individual model to predict potential distribution of Cabomba caroliniana in China.
Uncertainty inherent in ecological niche model predictions, different individual models usually showed variable predictions. Based on niche conservatism hypothesis, classic niche models that calibrated on the native areas and transferred onto introduced areas for evaluation and prediction have advantages. Using spatial independent testing points in introduced areas to select the best individual model to predict potential distribution of invasive species, offering reliable and valuable information for risk analysis. Cabomba caroliniana is an important invasive aquatic weed, it is native to South America, but has established populations in China. Here, we select the best native individual model to predict potential distribution, we first sorted the distributional records of C. caroliniana in chronological order, and investigated climate niche dynamics and niche conservatism of during C. caroliniana invasion across major continents using environmental variables that associated with observed records. 10 individual native models were then calibrated using two environmental datasets and five model algorithms, the best individual mode was used to predict potential distribution, which were selected based on the criteria of low omission and commission errors of independent introduced points. Results showed that present distributions of C. caroliniana were mainly along the eastern coastal areas of China, it dispersal northward in rivers and lakes system along Beijing-Hangzhou Grand Canal or South-to-North water diversion. The climate spaces occupied by different continental populations were overlapped broadly. Climate niche was conserved during its invasion in China, but there were many climate spaces that were unfilled comparing to its native niche spaces, suggesting its high invasion potential in China. Potential distributions were identified by our best native individual model, including Beijing, Shanghai, Shandong, Zhejiang, Jiangsu, Anhui, Hubei and Hunan provinces. These potential areas were mainly distributed in the southeastern rivers, lakes, and canals and channels, all these areas have high human activities and absent the nature enemy which might assist C. caroliniana expansion. Our researches call for rigorous survey in these areas, together with the integrative management action against C. caroliniana. The approaches adopted in present study to select best native individual model for C. caroliniana can be used for predict the potential distribution of other invasive species.