Exploratory analysis for complex-life-cycle amphibians: Revealing complex forest-reproductive effort relationships using redundancy analysis

详细信息	查看全文 | 推荐本文 |

• 作者：Joanna Hawley Howard ; ^{Jhawley504@gmail.com} ; Robert F. Baldwin ^{Baldwi6@clemson.edu} ; Bryan L. Brown¹ ; ^{Stonefly@vt.edu}
• 关键词：Species&ndash ; environment relationships ; Statistical analysis ; Amphibian ecology ; Redundancy analysis
• 刊名：Forest Ecology and Management
• 出版年：2012
• 期刊代码：54_03781127
• 类别：abs
• 出版时间：15 April, 2012
• 卷：270
• 期：Complete
• 页码：175-182
• 文件大小：298 K

摘要

The relationships among complex life-cycle (CLC) amphibians and their habitats involve interaction of biotic and abiotic variables across multiple spatial and temporal scales. Most studies of species-environment relationships utilize classic distributions and regression methods which may oversimplify results, as they often involve initial reduction of data, a priori model specification, and parsimonious model selection. We re-analyzed a dataset collected in a study of environment-reproductive effort relationships for two amphibian species. This data was originally analyzed using negative binomial regression and AIC model selection; we applied a multivariate analysis, a variant of Redundancy Analysis (RDA), to the same dataset to examine if additional information was revealed. The multivariate analysis identified the same primary drivers as did model selection: in the context of relatively intact forests, hydroperiod was the most important factor contributing to reproductive effort. However, the RDA elucidated what may be new primary and secondary drivers related to forest structure and composition. Reproductive effort was strongly associated with tree species richness, canopy cover, and presence of individual tree species; the strength of these relationships varied between amphibian species and across spatial scales. Spotted salamanders (Ambystoma maculatum) were strongly influenced by microclimate conditions and wood frogs (Lithobates sylvaticus) by pool canopy cover and presence of wetland tree species. A priori, parsimonious model selection assists managers by delivering biologically defensible, simple explanations. However the process reduces hypothesis-generating potential and omits variables that may become important as environments change.