Haefner J W
Department of Biology, Utah State University, 84322-5305, Logan, UT, USA.
Ecology Center, Utah State University, 84322-5305, Logan, UT, USA.
Oecologia. 1988 Jan;74(4):551-565. doi: 10.1007/BF00380053.
Using previously published data, several models were constructed to predict the distribution of Anolis lizard species on a set of sites on Puerto Rico and Jamaica. The models form a series with increasing ecological detail. The simpler "null" models are based on randomly created species-site matrices using progressively greater dependency on the observed matrix. The remaining models hypothesize that competition is the most important biotic interaction determining the intra-island distribution of the lizards. "Simple" competition models test the predictive power of simple statistical descriptions relating intensity of competition and ecological variables such as niche overlap and body size ratios. More complicated models are based on the ecomorph model of Williams (1972) and use the lizard resourceuse data of several niche dimensions (e.g., perch diameter and height). These models are derived from Puerto Rican data and tested against Jamaican data. The primary statistical tool used to test the accuracy of these models in the kappa statistic (Fleiss 1973) which assesses the degree of agreement in a contingency table relative to that expected by chance. The model structure is based on generative grammars (Haefner 1981), but is also related to artificial intelligence expert systems. Model comparisons indicate the following. (1) Only those null models constrained by the marginals of the observed species-site matrix agree with observed data. (2) Simple competition models based on fixed size ratios and/or fixed levels of allowable overlap do not agree well. (3) A complex competition model developed for Puerto Rico also shows significant agreement with lizard distributions on Jamaica, but is not better than a constrained null model. (4) If allowance is made for the restricted distribution of A. sagrei, a recent colonist of Jamaica, agreement of the competition model increases dramatically. It is predicted that A. sagrei would persist following an experimental transplant to eastern Jamaica.
利用先前发表的数据，构建了几个模型来预测波多黎各和牙买加一组地点上安乐蜥物种的分布。这些模型构成了一个生态细节不断增加的系列。较简单的“空”模型基于随机创建的物种 - 地点矩阵，对观察到的矩阵的依赖程度逐渐增加。其余模型假设竞争是决定蜥蜴岛内分布的最重要生物相互作用。“简单”竞争模型测试简单统计描述的预测能力，这些描述涉及竞争强度与生态变量，如生态位重叠和体型比例。更复杂的模型基于威廉姆斯（1972）的生态形态模型，并使用几个生态位维度（如栖息直径和高度）的蜥蜴资源利用数据。这些模型源自波多黎各的数据，并以牙买加的数据进行检验。用于检验这些模型准确性的主要统计工具是kappa统计量（弗莱斯，1973），它评估列联表中的一致程度相对于偶然预期的程度。模型结构基于生成语法（海夫纳，1981），但也与人工智能专家系统相关。模型比较表明如下情况。（1）只有那些受观察到的物种 - 地点矩阵边缘约束的空模型与观察数据一致。（2）基于固定体型比例和/或固定允许重叠水平的简单竞争模型与数据拟合不佳。（3）为波多黎各开发的复杂竞争模型也与牙买加的蜥蜴分布显示出显著一致性，但并不比受约束的空模型更好。（4）如果考虑到牙买加近期殖民者沙氏安乐蜥的分布受限情况，竞争模型的一致性会大幅提高。预计将沙氏安乐蜥实验性移植到牙买加东部后它将持续存在。
