Offshore distribution of yelkouan shearwaters in the north-western Adriatic Sea: insight from machine learning

Authors

DOI:

https://doi.org/10.32582/aa.65.1.8

Keywords:

distribution, Explainable Boosting Machine, machine learning, Mediterranean Sea, occurrence, Procellariidae, shearwaters, visual survey

Abstract

     The yelkouan shearwater Puffinus yelkouan is endemic to the Mediterranean and Black Seas, and classified as Vulnerable in the IUCN Red List. Information on the species’ distribution and habitat use in the eastern Mediterranean is scant, and only a few studies were based on direct visual observations in offshore waters. Here, we provide information on 1) the occurrence of yelkouan shearwaters within a 3000 km2 study area off the region of Veneto, Italy, in the northwestern Adriatic Sea, based on visual surveys conducted from small boats between April and October 2018–2022 (effort: 169 days, 23,836 km), and 2) the geographic, bathymetric and oceanographic variables likely to drive the species’ offshore distribution. Yelkouan shearwaters (238 sightings, 916 individuals) were observed in waters 9–33 m deep, between 2 and 24 km from the coast. Individual counts ranged between 1 and 100, with 95% of the encounters having less than 10 individuals. An Explainable Boosting Machine model – a machine learning technique based on generalized additive models – selected chlorophyll a as the most important variable to explain the species’ occurrence, followed by distance from the coast, and bottom depth. The model indicated a higher occurrence in waters with chlorophyll a less than ~2.3 mg/m3, farther than ~15 km from the coast, and deeper than ~22 m. The effects of SST, salinity, and day of the year were less clear. This study provides insight into the offshore distribution of yelkouan shearwaters, within one of the Mediterranean areas most exposed to cumulative human threats.

Published

24.05.2024

Issue

Section

Original article