Albatrosses respond adaptively to climate variability by changing variance in a foraging trait

Individual variability in diving behavior of the Black-vented Shearwater in an ever-changing habitat

Oceanic mesoscale systems are characterized by inherent variability. Climatic change adds entropy to this system, making it a highly variable environment in which marine species live. Being at the higher levels of the food chain, predators maximize their performance through plastic foraging strategies. Individual variability within a population and the possible repeatability across time and space may provide stability in a population facing environmental changes. Therefore, variability and repeatability of behaviors, particularly diving behavior, could play an important role in understanding the adaptation pathway of a species. This study focuses on characterizing the frequency and timing of different dives (termed simple and complex) and how these are influenced by individual and environmental characteristics (sea surface temperature, chlorophyll a concentration, bathymetry, salinity, and Ekman transport). This study is based on GPS and accelerometer-recorded information from a breeding group of 59 Black-vented Shearwater and examine consistency in diving behavior at both individual and sex levels across four different breeding seasons. The species was found to be the best performing free diver in the Puffinus genus with a maximum dive duration of 88 s. Among the environmental variables assessed, a relationship was found with active upwelling conditions enhancing low energetic cost diving, on the contrary, reduced upwelling and warmer superficial waters induce more energetically demanding diving affecting diving performance and ultimately body conditions. The body conditions of Black-vented Shearwaters in 2016 were worse than in subsequent years, in 2016, deepest and longest complex dives were recorded, while simple dives were longer in 2017-2019. Nevertheless, the species' plasticity allows at least part of the population to breed and feed during warmer events. While carry-over effects have already been reported, the effect of more frequent warm events is still unknown.

A predictive timeline of wildlife population collapse

Contemporary rates of biodiversity decline emphasize the need for reliable ecological forecasting, but current methods vary in their ability to predict the declines of real-world populations. Acknowledging that stressor effects start at the individual level, and that it is the sum of these individual-level effects that drives populations to collapse, shifts the focus of predictive ecology away from using predominantly abundance data. Doing so opens new opportunities to develop predictive frameworks that utilize increasingly available multi-dimensional data, which have previously been overlooked for ecological forecasting. Here, we propose that stressed populations will exhibit a predictable sequence of observable changes through time: changes in individuals' behaviour will occur as the first sign of increasing stress, followed by changes in fitness-related morphological traits, shifts in the dynamics (for example, birth rates) of populations and finally abundance declines. We discuss how monitoring the sequential appearance of these signals may allow us to discern whether a population is increasingly at risk of collapse, or is adapting in the face of environmental change, providing a conceptual framework to develop new forecasting methods that combine multi-dimensional (for example, behaviour, morphology, life history and abundance) data.

The role of individual variability on the predictive performance of machine learning applied to large bio-logging datasets

Animal-borne tagging (bio-logging) generates large and complex datasets. In particular, accelerometer tags, which provide information on behaviour and energy expenditure of wild animals, produce high-resolution multi-dimensional data, and can be challenging to analyse. We tested the performance of commonly used artificial intelligence tools on datasets of increasing volume and dimensionality. By collecting bio-logging data across several sampling seasons, datasets are inherently characterized by inter-individual variability. Such information should be considered when predicting behaviour. We integrated both unsupervised and supervised machine learning approaches to predict behaviours in two penguin species. The classified behaviours obtained from the unsupervised approach Expectation Maximisation were used to train the supervised approach Random Forest. We assessed agreement between the approaches, the performance of Random Forest on unknown data and the implications for the calculation of energy expenditure. Consideration of behavioural variability resulted in high agreement (> 80%) in behavioural classifications and minimal differences in energy expenditure estimates. However, some outliers with < 70% of agreement, highlighted how behaviours characterized by signal similarity are confused. We advise the broad bio-logging community, approaching these large datasets, to be cautious when upscaling predictions, as this might lead to less accurate estimates of behaviour and energy expenditure.