样本量及其空间分布对物种分布模型的影响  

Impact of sample size and spatial distribution on species distribution model

作　　者：赵晓冏 巩娟霄 赵莎莎 孟浩贤 闫伯前[4] 齐欣宇 Zhao Xiao-jiong1,2,3, Gong Juan-xiao1,2,3, Zhao Sha-sha2, Meng Hao-xian2, Yan Bo-qian4, Qi Xin-yu5(1. Northwest Institute of Eco-environment and Resources, Chinese Academy of Sciences, Lanzhou 730000, China ;2. Gansu Academy of Environmental Sciences, Lanzhou 730000, China ;3. University of Chinese Academy of Sciences, Beijing 100049, China ;4. Food Science, Beijing Agriculture University, Beijing 102206, China; 5. College of Earth and Environmental Sciences, Lanzhou University, Lanzhou 730000, Chin)

机构地区：[1]中国科学院西北生态环境资源研究院,兰州730000 [2]甘肃省环境科学设计研究院,兰州730000 [3]中国科学院大学,北京100049 [4]北京农学院食品科学系,北京1022065 [5]兰州大学资源环境学院,兰州730000

出　　处：《兰州大学学报（自然科学版）》2018年第2期208-215,共8页Journal of Lanzhou University(Natural Sciences)

基　　金：国家自然科学基金项目(31370587);甘肃省青年科技基金项目(1506RJYA154)

摘　　要：在大量野外调查的基础上,结合SPOT遥感影像解译的太白红杉分布范围,采用从解译的太白红杉分布数据中提取的6种不同样本量、在不同海拔梯度和坡向的样本量分布以及3种不同的评价方法,测试了两种模型(MaxEnt,Garp)下太白红杉预测精度的变化.结果表明:随着样本量的增加,模型精度随之增大,预测结果趋于稳定.大样本量时,Garp模型的性能优于MaxEnt模型;小样本量时,MaxEnt模型预测结果最优.样本在不同海拔梯度和坡向分布时,两种模型样本量不均等分布的预测性能明显优于均等分布;相同样本分布下,两者预测能力没有明显差异.在保证一定样本量的前提下,为提高模拟能力,选择样本时应尽可能按照物种分布地的数量依比例取样.结果对合理采样和提高物种分布模拟精确度有重要意义,可用于指导物种分布模型的选择及其最优样本的获取.On the basis of a great deal of field investigation, combined with high resolution spot remote sensing imaging, the distribution range of Larix chinensis in Mount Taibai was interpreted, using six different sample sizes from L. chinensis distribution data that were extracted from different elevation gradients and slopes and directions, three different evaluation methods, changes in prediction accuracy were tested under the two models（MaxEnt and Garp）. The results showed that, with the increase of the sample size, the accuracy of the model increased, the prediction results tended to be stable. With a larger sample size, the Garp model performed better than the MaxEnt model, and the MaxEnt model exhibited a lower sensitivity to the sample size, the MaxEnt model in all sample sizes showed the prediction ability better than Garp; the smaller the sample size, the MaxEnt model produced a better forecasting result. The samples were distributed in different elevation gradients and slope directions for MaxEnt and Garp models and, for an unequal distribution of sample sizes, the prediction performance was significantly better than the equal distribution. For the same sample distribution, the two models＇ predictive ability had no significant difference. In order to ensure a certain sample volume, the sample selection was made as soon as possible in accordance with the distribution according to the proportion of sampling. The results can be used to guide the selection of a species distribution model and optimal sample acquisition.

关 键 词：MaxEnt模型 Garp模型 SPOT影像 太白红杉 

分 类 号：Q948[生物学—植物学]