Predicting habitat suitability in temperate seagrass ecosystems

Abstract

The worldwide observed dramatic decline of seagrasses has typically been attributed to multiple stressors such as eutrophication, disease, sedimentation, and toxicity events. Using principal component analysis and (multivariate) logistic regression, we investigated the importance of 30 commonly measured variables in explaining the presence and absence of the temperate seagrass species Zostera marina and Zostera noltii at 84 Western European locations. Although many interrelated variables influence seagrass presence in our dataset, presence or absence of both species could be reliably predicted by using only two easy-to-measure variables. A logistic regression model of Z. marina correctly predicted 77% of all observations by including water column light attenuation and sediment pore-water reduction oxidation potential (RedOx). The Z. noltii model had an 86% accuracy based on only tidal location (intertidal or subtidal zone) and pore-water RedOx. Applying the models to five evaluation sites demonstrated that both models can be usefully applied as tools for seagrass ecosystem restoration and conservation.