An evaluation of Ecological Niche modelling as a predictive management tool to map habitat suitability of seagrasses (Zostera marina) in UK coastal waters

Seagrasses including Zostera marina exist in extensive communities in shallow subtidal zones of UK waters and are therefore subject to intense natural and anthropogenic pressures. As an internationally important species, it is listed under OSPAR (Oslo-Paris Convention), the EC Habitats Directive and designated a ‘priority habitat’ under the UK Biodiversity Action Plan. Zostera spp. communities form fundamental structural and functional components of Large Marine Ecosystems (LMEs) and support a diversity of other fauna. Current distribution is fragmented and has been in decline since the 1930s. Managers of existing Z. marina communities in the UK require innovative tools to empower adaptive management of these natural habitats. This paper contextualises the use of predictive modelling as a decision support tool for Ecosystem based Management (EBM) and evaluates its relative utility amongst the emerging ensemble of contemporary, parsimonious approaches. The use of species distribution models (SDMs) is becoming increasingly widespread in the field of conservation biology, especially with respect to the response of threatened marine species to increased physico-chemical gradients. Using maximum entropy modelling (Maxent), all available records of Zostera marina distribution were modelled in response to a suite of environmental variables. We generated habitat suitability maps and evaluated the model via internal cross-validation and a threshold-independent metric, the Area-Under-Curve (AUC) analysis. External methods of validation were also sourced, including expert interpretation of the suitability maps and data-partitioning. We developed models on both regional and local scales that performed well in terms of cross-validation (Jack-kniving) and the AUC statistic (delivering values of 0.971 and 0.981, respectively) suggesting the use of Maxent is more appropriate on finer spatial scales. The suitability maps identified favourable habitats consistent with expert validation and published species inventories. This study highlights the efficacy of combining ArcGIS systems for data management and modelling species-biogeographic distributions from presence-only data, relevant for habitat suitability management. It thus elucidates habitat areas at risk, both currently and in the future.Faculty of Science and Technolog