虚拟物种的基本原理及其在物种分布模型评估中的应用  

The basic principle of virtual species and its application to evaluations of species distribution models

作　　者：黄敏毅[1] 孔晓泉 段仁燕[1] 吴甘霖[1] 张中信[1] 

机构地区：[1]安庆师范学院生命科学学院,安庆246011

出　　处：《生态学报》2016年第9期2460-2470,共11页Acta Ecologica Sinica

基　　金：林业局公益性行业科研专项(201304314);国家自然科学基金(31300342;31100311);安徽省自然科学基金(1408085MC64)

摘　　要：物种分布模型是建立在物种出现或缺失数据的基础上,但可获得的真实分布数据存在着各种各样的缺点(如:物种识别错误、坐标错误、抽样偏差、数据缺失等),影响着物种分布模型的预测性能、稳定性及应用,因此使用物种真实分布数据评估物种分布模型将带来很大的不确定性。为避免这种不确定性,越来越多的研究使用虚拟物种来评价物种分布模型的性能,评估新方法的优劣。虚拟物种是一种建立在真实(或虚拟)地理信息系统下人工生命,是简化和抽象的物种,它通过模拟物种对环境变量的响应关系,评估物种在不同环境变量下的出现概率,人为地给出虚拟的物种分布数据。虚拟物种具有数据容易获得、数据质量可控、避免过度模拟等优势,目前它被广泛用于评估物种特性、抽样偏差、地理信息、出现/缺失标准等对物种分布模型性能的影响。虚拟物种是大尺度研究中不可或缺的重要工具,有利于解决真实数据未能解决的科学问题。常用的构成算法有求和法、求积法和综合法,但这些方法可能存在补偿效应,扩大了物种的分布范围。考虑到虚拟物种的不足,提出了未来虚拟物种可能的发展方向(避免过度脱离真实,完善虚拟物种的构成算法,构建虚拟的模式生物、群落及生态系统等)。为帮助研究者快速构建虚拟物种,基于R环境开发了一个虚拟物种构成软件包(SDMvspecies)。虚拟物种可以与真实物种相结合,通过改进模型的构成方法,有利于解决一些真实数据未能解决的问题;虚拟物种的应用也将导致一些新理论的产生,有利于更好地理解生态学原理。The fascinating scientific questions of how and where species will potentially distribute under current and changing environmental conditions have inspired many biogeographers, ecologists, and managers to predict the potential distributions of plants or animals by quantifying species-environment relationships. The species distribution model （SDM） , an essential modeling tool, has been developed. A key challenge in using real species data （presence-absence data and/or presence-only data） for SDM is the uncertainty about where and how the thousands of species distribution data records are attained. The majority of species distribution data sets are derived from herbaria, university databases, museums, or even amateur field workers. Therefore, attaining a reasonable explanation for species distribution in the wild is often hindered by the problems inherent in these large data sets, including species-specific properties （ e. g., species prevalence, dispersal barriers, interspecific competition, distribution pattern）, biased sampling （e.g., reachability of observation sites, visibility or detectability of observation objects）, variability among observation methods （ e.g., time interval and spatial range）, and habitat types, particularly for data collected over a long time interval and a large spatial range. The use of virtual species could provide a suitable unifying framework to select the most appropriate model for such evaluations, by comparing the predictive accuracy and virtual distributions in a geographic information system model of a real landscape. In recent years, virtual species distribution models have become increasingly important tools to study various problems in the fields of conservation biology, ecology, biogeography, climate change research, and evolution. Virtual species have many advantages, including the ease of attaining a large number of data sets for each scenario, ability to fully control the quality of data, prevention of the over-fitted phenomenon inherent to SDMs, and t

关 键 词：物种分布模型 虚拟物种 生境适宜度 抽样偏差 人工生命 

分 类 号：Q141[生物学—生态学]