基于生态位模型的石蒜适生区预测

The prediction on potential geographic distribution of Lycoris radiata based on ecological niche models

运用预设预测规则的遗传算法(GARP)和最大熵(Max Ent)两种生态位预测模型,以及受试者工作曲线(ROC)分析方法,预测石蒜属石蒜潜在适生区。结果表明,GARP和Max Ent模型ROC曲线下面的面积AUC(area under the ROC curve)均值分别为0.910和0.988,Max Ent模型的AUC值更大,预测结果更准确,运行速度更快,更适合用于石蒜的适生区预测。对环境变量进行刀切法表明,在所有环境变量中,最冷季度平均温度对于石蒜分布的影响(贡献)最大,其次是年均温、最冷月的最低温度和最暖季度降水量,而海拔、降水量变化方差对石蒜分布的影响比较小。预测结果显示,石蒜在世界范围内主要分布在亚洲东部以及亚洲中部一小部分,另外北美洲的东部,欧洲南部一小部分地区也适合其生长。在中国范围内主要分布在云南、贵州、福建、江苏、浙江、安徽、江西、重庆、湖北、湖南等省,以及山东、河南、陕西、甘肃等省南部;四川东部和广东、广西(除了南部沿海地区)均预测为适生区,海南、台湾、西藏部分地区也是适生区。

Lycoris radiata is a perennial ornamental plant. In this paper,the potential geographic distribution of L. radiata was predicted on by using the Genetic Algorithm for Rule-set Production（ GARP） and Maximum Entropy（ Max Ent） models,and the better model was determined by ROC curve. The results showed that the average AUC（ areas under ROC curves） with GARP model and Max Ent model were 0. 910 and 0. 988 respectively,which meant that Max Ent model was better than GARP model in predicting the potential geographic distribution of L. radiata. The results obtained from jackknife method indicated that the average temperature in the coldest season,the annual average temperature,the lowest temperature in the coldest month and the precipitation in the warmest season would influence on the distribution of L. radiata greatly,while the altitude and the variance of precipitation range would influence on the distribution of L. radiata less. According to the prediction of Max Ent model,the potential geographic distribution within the world was mainly in the eastern Asia,a small part in the southern Europe,eastern North America and central Asia. And the potential geographic distribution within China mainly concentrated in the regions as Yunnan,Guizhou,Fujian,Anhui,Jiangsu,Zhejiang,Jiangxi,Hubei,Hunan,Guangdong and Guangxi,southern Shandong,Henan,Shaanxi,Gansu and the eastern Sichuan,and a small part in Hainan as well as Taiwan and Tibet.