Prediction of ten seabird species of the Patagonian Shelf: digital baseline with Open Access Machine Learning for a more effective seascape management.

A RAYA REY; F HUETTMANN

Congreso; US-IALE; 2016

Quantitative knowledge about the spatial distribution of seabirds at sea is relevant for conservation. The Southwest Atlantic Ocean and the extended Patagonian shelf in particular, is a highly productive seascape. It is a very complex ecosystem of global relevance (i.e. fisheries, climate regulation) and maintains a great diversity and abundance of marine life. Direct and indirect discharge of chemical pollutants, industrial and expanded cities pollution, by-catch, entanglement, climate change and alien species pose severe threats for seabird populations in the Patagonian shelf. Nevertheless seabird at sea distribution is missing for most species and even less open access efforts. Based on the data mining and machine learning TreeNet (boosting) algorithm, and 10 environmental publicly available Open Source Geographic Information Systems (GIS) layers, we built for the first time 10 predictive seabird models for the Patagonian shelf based on public open access data archives such as the Global Biodiversity Information Facility (GBIF), and Ocean Biogeographic Information System Spatial Ecological Analysis of Megavertebrate Populations database (OBIS-Seamap). Top five explanatory variables for seabird distribution were distances to shelf break, to subantarctic front, to coast, to ports, as well as bathymetry. Our models are the first of its kind for the seascape and have a good accuracy according to two assessment metrics (Receiver Operating Characteristics (ROC-AUC) curves, and alternative field data). Seabird distribution maps for the Patagonian shelf and especially when publicly available for transparency are essential for best practices, marine conservation planning, assessment of conflict with human activities, and forecasting regarding climate change impacts.