北美银钟花在中国的适宜栽培区研究

Prediction of suitable cultivation area for Halesia carolina L. in China

【目的】北美银钟花(Halesia carolina L.)是优良的园林景观树种,利用最大熵(Maxent)模型预测其在中国适宜引种栽培区,为北美银钟花合理引种栽培提供科学依据。【方法】通过ENMeval数据包优化Maxent模型参数设置,基于57条现代地理分布记录和7个生物气候变量模拟和预测北美银钟花在中国的适宜栽培区,综合使用百分比贡献率、置换重要值和多元方差分析等方法评价中国、美国适宜分布区的环境差异和主要气候限制因子。【结果】当特征组合为线性、二次型、乘积型和片段化,调控倍频为1. 5时,模型复杂度和过拟合程度较低,此时测试集受试者工作特征曲线下的面积值达到0. 956 7±0. 032 9,模拟准确度极高。北美银钟花在中国的适宜栽培区主要包括台湾岛东部和北部、海南岛东部、云南西部和西南部、中国大陆中东部4个区域。多元方差分析和稳健多元方差分析显示,中国和美国的适宜区之间具有显著的气候差异,其中,指示降水和气温变异程度的指标差异最大,指示气温极端特征的指标差异最小。【结论】降水因子与北美银钟花潜在地理分布的关联性更强,中国适宜区内不均匀的降雨季节性分配可能是限制其引种栽培的主要气候因素。

【Objective】Halesia carolina L.,commonly called the snowdrop-tree,is a small understory tree of the genus Halesia in the family Styracaceae. This species has high ornamental value for its spell-binding beauty when in blossom.Here,we used the Maxent model to predict the potential suitable cultivation distribution of H. carolina in China. 【Method】We used the ENMeval package in R software to optimize parameters of the Maxent model,and simulated suitable cultivation regions of H. carolina based on 57 geographical distribution records and 7 bioclimatic variables. Percent contribution,permutation importance,and multivariate analysis of variance were applied to evaluate the limiting variables and environment differences between China and America. 【Result】Feature combination of linear,quadratic,product and hinge and the regularization multiplier of 1. 5 significantly decreased model complexity and the degree of overfitting. Accuracy of the Maxent model was very high,as indicated by the area under the receiver operator characteristic curve for a test dataset of 0. 956 7 ± 0. 032 9. The suitable cultivation regions in China cover eastern and northern Taiwan,eastern Hainan Island,western and southwestern Yunnan,and the central and eastern mainland of China. Multivariate analysis of variance and robust multivariate analysis of variance showed that differences in climate between China and America were significant. Variables representing variation in precipitation and temperature had greater differences than variables indicative of extreme temperatures.【Conclusion】Precipitation factors are strongly related to the potential geographical distribution of H. carolina,seasonal precipitation may be the main factor restricting the suitable cultivation regions in China.