Predicting animal behaviour using deep learning: GPS data alone accurately predict diving in seabirds

Deep learning allows computational models that are composed of multiple processing layers to learn representations of data with multiple levels of abstraction. These methods have dramatically improved the state-of-the-art in speech recognition, visual object recognition, object detection and many other domains such as drug discovery and genomics. Deep learning discovers intricate structure in large data sets by using the backpropagation algorithm to indicate how a machine should change its internal parameters that are used to compute the representation in each layer from the representation in the previous layer. Deep convolutional nets have brought about breakthroughs in processing images, video, speech and audio, whereas recurrent nets have shone light on sequential data such as text and speech.

Pelagic marine predators face unprecedented challenges and uncertain futures. Overexploitation and climate variability impact the abundance and distribution of top predators in ocean ecosystems. Improved understanding of ecological patterns, evolutionary constraints and ecosystem function is critical for preventing extinctions, loss of biodiversity and disruption of ecosystem services. Recent advances in electronic tagging techniques have provided the capacity to observe the movements and long-distance migrations of animals in relation to ocean processes across a range of ecological scales. Tagging of Pacific Predators, a field programme of the Census of Marine Life, deployed 4,306 tags on 23 species in the North Pacific Ocean, resulting in a tracking data set of unprecedented scale and species diversity that covers 265,386 tracking days from 2000 to 2009. Here we report migration pathways, link ocean features to multispecies hotspots and illustrate niche partitioning within and among congener guilds. Our results indicate that the California Current large marine ecosystem and the North Pacific transition zone attract and retain a diverse assemblage of marine vertebrates. Within the California Current large marine ecosystem, several predator guilds seasonally undertake north-south migrations that may be driven by oceanic processes, species-specific thermal tolerances and shifts in prey distributions. We identify critical habitats across multinational boundaries and show that top predators exploit their environment in predictable ways, providing the foundation for spatial management of large marine ecosystems. ©2011 Macmillan Publishers Limited. All rights reserved

The management and conservation of the world's oceans require synthesis of spatial data on the distribution and intensity of human activities and the overlap of their impacts on marine ecosystems. We developed an ecosystem-specific, multiscale spatial model to synthesize 17 global data sets of anthropogenic drivers of ecological change for 20 marine ecosystems. Our analysis indicates that no area is unaffected by human influence and that a large fraction (41%) is strongly affected by multiple drivers. However, large areas of relatively little human impact remain, particularly near the poles. The analytical process and resulting maps provide flexible tools for regional and global efforts to allocate conservation resources; to implement ecosystem-based management; and to inform marine spatial planning, education, and basic research.