Geospatial models are used to predict the distribution of terrestrial and marine species, according to their ecological and ethological habits. The bottlenose dolphin is a cosmopolitan marine top predator, inhabiting most of the ocean, with the exception of polar and subpolar waters. This wide distribution is associated with a remarkable plasticity in ecological and behavioural habits, which makes it difficult to model and predict its distribution. This study proposes a ‘multi-type approach’ to predict the presence and distribution of the bottlenose dolphin in the Pelagos Sanctuary, a Specially Protected Area of Mediterranean Importance located in the north-west Mediterranean Sea. A multi-type model based on random forest regression was developed, analysing the distribution habits of two geographical units living in the Pelagos area. When compared with a classical single-type model, the multi-type model performed much better in a prediction test (true skill statistics, TSS = 85% vs. 52%), confirming the value of this experimental approach. This work suggests that wild species should not be considered as one single-type entity, as local specialization may change and shape their distribution habits.
Species-specific distribution model may be not enough: The case study of bottlenose dolphin (Tursiops truncatus) habitat distribution in Pelagos Sanctuary
Vassallo P.;Paoli C.;
2020-01-01
Abstract
Geospatial models are used to predict the distribution of terrestrial and marine species, according to their ecological and ethological habits. The bottlenose dolphin is a cosmopolitan marine top predator, inhabiting most of the ocean, with the exception of polar and subpolar waters. This wide distribution is associated with a remarkable plasticity in ecological and behavioural habits, which makes it difficult to model and predict its distribution. This study proposes a ‘multi-type approach’ to predict the presence and distribution of the bottlenose dolphin in the Pelagos Sanctuary, a Specially Protected Area of Mediterranean Importance located in the north-west Mediterranean Sea. A multi-type model based on random forest regression was developed, analysing the distribution habits of two geographical units living in the Pelagos area. When compared with a classical single-type model, the multi-type model performed much better in a prediction test (true skill statistics, TSS = 85% vs. 52%), confirming the value of this experimental approach. This work suggests that wild species should not be considered as one single-type entity, as local specialization may change and shape their distribution habits.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.