A statistical platform for evaluating coral health in an era of changing global climate
By Anderson Mayfield, Alexandra Dempsey, Javed Inamdar, and Chii-Shiarng Chen
Given the significant threats against coral reef ecosystems, there is an urgent need to develop the capacity to make predictions as to which coral reefs are most stress-susceptible, as well as which are most resilient. However, there is such extensive variation in coral physiology, even in conspecifics reared in the same laboratory tank, that prior works have been characterized by too low statistical power to even explain previously obtained datasets with confidence, let alone predict the behavior of to-be-sampled corals. To obtain a better grasp of the environmental and organismal factors that contribute to variation in coral physiology, a published coral reef dataset from Fiji’s remote, understudied Lau Archipelago was re-analyzed herein with a variety of both univariate and multivariate statistical approaches. Of the 12 environmental parameters hypothesized to influence reef coral physiology, only two both significantly drove variation in the multivariate coral physiological response and featured readily in the best-fit models produced by stepwise regression and partial least squares analyses: island and host coral species. That being said, the majority of the models were characterized by low predictive capacity; more data are clearly needed to generate statistical algorithms capable of forecasting coral behavior with confidence in this era of rapidly changing global climate.