Consideration of species community composition in statistical analyses of coral disease risk

Marine Ecology Progress Series

John W. Kern¹  Susan Harrell Yee¹  Sarah Hession¹  Deborah L. Santavy¹ 

关键词: Coral;    Disease;    Bleaching;    Community composition;    Species distribution;    Regression models;   

DOI  :  10.3354/meps09135

学科分类：海洋学与技术

来源: Inter-Research

【 摘 要 】

ABSTRACT: Diseases are increasing in marine ecosystems, and these increases have been attributed to a number of environmental factors including climate change, pollution, and overfishing. However, many studies pool disease prevalence into taxonomic groups, disregarding host-species composition when comparing sites or assessing environmental impacts on patterns of disease presence. We used simulated data under a known environmental effect to assess the ability of standard statistical methods (binomial and linear regression, ANOVA) to detect a significant environmental effect on pooled disease prevalence with varying species abundance distributions and relative susceptibilities to disease. When one species was more susceptible to a disease and both species only partially overlapped in their distributions, models tended to produce a greater number of false positives (Type I error). Differences in disease risk between regions or along an environmental gradient tended to be underestimated, or, even in the wrong direction, when highly susceptible taxa had reduced abundances in impacted sites, a situation likely to be common in nature. Including relative abundance as an additional variable in regressions improved model accuracy, but tended to be conservative, producing more false negatives (Type II error) when species abundance was strongly correlated with the environmental effect. An analysis of field data on bleaching from the Florida Keys, FL, USA, affirmed general results of model simulations. Investigators should be cautious of underlying assumptions about species similarity in susceptibility and species composition when interpreting pooled data on disease risk.

【 授权许可】

Unknown