Modelling the spatial distribution of plaice ( Pleuronectes platessa), sole ( Solea solea) and thornback ray ( Raja clavata) in UK waters for marine management and planning

Species distribution maps are needed for ecosystem-based marine management including the development of marine spatial plans. If such maps are based on predictive models then modelling procedures should aim to maximise validation success, and any uncertainty in the predictions needs to be made explicit. We developed a predictive modelling approach to produce robust maps of the distributions of selected marine species at a regional scale. We used 14 years of survey data to map the distributions of plaice, sole and thornback ray in three hydrographic regions comprising parts of the Irish Sea, Celtic Sea and the English Channel with the help of the hybrid technique regression kriging, which combines regression models with geostatistical tools. For each species–region combination we constructed logistic Generalized Linear Models (GLMs) based on presence–absence data using the environmental variables: depth, bottom temperature, bed shear stress and sediment type, as predictors. We selected GLMs using the mean squared error of prediction (MSEP) estimated by cross-validation then conducted a geostatistical analysis of the residuals to incorporate spatial structure in the predictions. In general, we found that species occurrence was positively related to shallow areas, a bed shear stress of between 0 and 1.5 N/m2, and the presence of sandy sediment. Predicted species occurrence probabilities were in good agreement with survey observations. This modelling framework selects environmental models based on predictive ability and considers the effect of spatial autocorrelation on predictions, together with the simultaneous presentation of observations, associated uncertainties, and predictions. The potential benefit of these distribution maps to marine management and planning is discussed.