Joint modelling of multi‐scale animal movement data using hierarchical hidden Markov models

How to account for behavioral states in step-selection analysis: a model comparison

Step-selection models are widely used to study animals' fine-scale habitat selection based on movement data. Resource preferences and movement patterns, however, often depend on the animal's unobserved behavioral states, such as resting or foraging. As this is ignored in standard (integrated) step-selection analyses (SSA, iSSA), different approaches have emerged to account for such states in the analysis. The performance of these approaches and the consequences of ignoring the states in step-selection analysis, however, have rarely been quantified. We evaluate the recent idea of combining iSSAs with hidden Markov models (HMMs), which allows for a joint estimation of the unobserved behavioral states and the associated state-dependent habitat selection. Besides theoretical considerations, we use an extensive simulation study and a case study on fine-scale interactions of simultaneously tracked bank voles (Myodes glareolus) to compare this HMM-iSSA empirically to both the standard and a widely used classification-based iSSA (i.e., a two-step approach based on a separate prior state classification). Moreover, to facilitate its use, we implemented the basic HMM-iSSA approach in the R package HMMiSSA available on GitHub.

Incorporating multidimensional behavior into a risk management tool for a critically endangered and migratory species

Conservation of migratory species exhibiting wide-ranging and multidimensional behaviors is challenged by management efforts that only utilize horizontal movements or produce static spatial-temporal products. For the deep-diving, critically endangered eastern Pacific leatherback turtle, tools that predict where turtles have high risks of fisheries interactions are urgently needed to prevent further population decline. We incorporated horizontal-vertical movement model results with spatial-temporal kernel density estimates and threat data (gear-specific fishing) to develop monthly maps of spatial risk. Specifically, we applied multistate hidden Markov models to a biotelemetry data set (n = 28 leatherback tracks, 2004-2007). Tracks with dive information were used to characterize turtle behavior as belonging to 1 of 3 states (transiting, residential with mixed diving, and residential with deep diving). Recent fishing effort data from Global Fishing Watch were integrated with predicted behaviors and monthly space-use estimates to create maps of relative risk of turtle-fisheries interactions. Drifting (pelagic) longline fishing gear had the highest average monthly fishing effort in the study region, and risk indices showed this gear to also have the greatest potential for high-risk interactions with turtles in a residential, deep-diving behavioral state. Monthly relative risk surfaces for all gears and behaviors were added to South Pacific TurtleWatch (SPTW) (https://www.upwell.org/sptw), a dynamic management tool for this leatherback population. These modifications will refine SPTW's capability to provide important predictions of potential high-risk bycatch areas for turtles undertaking specific behaviors. Our results demonstrate how multidimensional movement data, spatial-temporal density estimates, and threat data can be used to create a unique conservation tool. These methods serve as a framework for incorporating behavior into similar tools for other aquatic, aerial, and terrestrial taxa with multidimensional movement behaviors.

Quantifying the behavioural consequences of shark ecotourism

Shark populations globally are facing catastrophic declines. Ecotourism has been posited as a potential solution to many of the issues facing shark conservation, yet increasingly studies suggest that such activity may negatively influence aspects of shark ecology and so further pressure declining populations. Here we combine UAV videography with deep learning algorithms, multivariate statistics and hidden Markov models (HMM) to quantitatively investigate the behavioural consequences of ecotourism in the whale shark (Rhincodon typus). We find that ecotourism increases the probability of sharks being in a disturbed behavioural state, likely increasing energetic expenditure and potentially leading to downstream ecological effects. These results are only recovered when fitting models that account for individual variation in behavioural responses and past behavioural history. Our results demonstrate that behavioural responses to ecotourism are context dependent, as the initial behavioural state is important in determining responses to human activity. We argue that models incorporating individuality and context-dependence should, wherever possible, be incorporated into future studies investigating the ecological impacts of shark ecotourism, which are only likely to increase in importance given the expansion of the industry and the dire conservation status of many shark species.

Crabs ride the tide: incoming tides promote foraging of Giant Mud Crab (Scylla serrata)

BACKGROUND

Effective fisheries management of mobile species relies on robust knowledge of animal behaviour and habitat-use. Indices of behaviour can be useful for interpreting catch-per-unit-effort data which acts as a proxy for relative abundance. Information about habitat-use can inform stocking release strategies or the design of marine protected areas. The Giant Mud Crab (Scylla serrata; Family: Portunidae) is a swimming estuarine crab that supports significant fisheries harvest throughout the Indo-West Pacific, but little is known about the fine-scale movement and behaviour of this species.

METHODS

We tagged 18 adult Giant Mud Crab with accelerometer-equipped acoustic tags to track their fine-scale movement using a hyperbolic positioning system, alongside high temporal resolution environmental data (e.g., water temperature), in a temperate south-east Australian estuary. A hidden Markov model was used to classify movement (i.e., step length, turning angle) and acceleration data into discrete behaviours, while also considering the possibility of individual variation in behavioural dynamics. We then investigated the influence of environmental covariates on these behaviours based on previously published observations.

RESULTS

We fitted a model with two well-distinguished behavioural states describing periods of inactivity and foraging, and found no evidence of individual variation in behavioural dynamics. Inactive periods were most common (79% of time), and foraging was most likely during low, incoming tides; while inactivity was more likely as the high tide receded. Model selection removed time (hour) of day and water temperature (°C) as covariates, suggesting that they do not influence Giant Mud Crab behavioural dynamics at the temporal scale investigated.

CONCLUSIONS

Our study is the first to quantitatively link fine-scale movement and behaviour of Giant Mud Crab to environmental variation. Our results suggest Giant Mud Crab are a predominantly sessile species, and support their status as an opportunistic scavenger. We demonstrate a relationship between the tidal cycle and foraging that is likely to minimize predation risk while maximizing energetic efficiency. These results may explain why tidal covariates influence catch rates in swimming crabs, and provide a foundation for standardisation and interpretation of catch-per-unit-effort data-a commonly used metric in fisheries science.

Leveraging big data to uncover the eco-evolutionary factors shaping behavioural development

Mapping the eco-evolutionary factors shaping the development of animals' behavioural phenotypes remains a great challenge. Recent advances in 'big behavioural data' research-the high-resolution tracking of individuals and the harnessing of that data with powerful analytical tools-have vastly improved our ability to measure and model developing behavioural phenotypes. Applied to the study of behavioural ontogeny, the unfolding of whole behavioural repertoires can be mapped in unprecedented detail with relative ease. This overcomes long-standing experimental bottlenecks and heralds a surge of studies that more finely define and explore behavioural-experiential trajectories across development. In this review, we first provide a brief guide to state-of-the-art approaches that allow the collection and analysis of high-resolution behavioural data across development. We then outline how such approaches can be used to address key issues regarding the ecological and evolutionary factors shaping behavioural development: developmental feedbacks between behaviour and underlying states, early life effects and behavioural transitions, and information integration across development.

Multilevel Hidden Markov Models for Behavioral Data: A Hawk-and-Dove Experiment

Motivated by the analysis of behavioral data taken from an economic experiment based on the Hawk-and-Dove game, this article describes a multilevel hidden Markov model, that includes covariates, autoregression, and endogenous initial conditions under a unified framework. The data at hand are affected by multiple sources of latent heterogeneity, due to multilevel unobserved factors that operate in conjunction with observed covariates at all the levels of the data hierarchy. We fit a multilevel logistic regression model for repeated measurements of player behaviors, nested within groups of interacting players. The model integrates discrete random effects at the group level and Markovian sequences of discrete random effects at the player level. Parameters are estimated by a computationally feasible expectation-maximization algorithm. We model the probability of playing the Hawk strategy, which implies fighting aggressively for controlling an asset, and test the role played by initial possession, property, and other player-specific characteristics in driving hawkish behaviors. The results from our study suggest that crucial factors in determining hawkish behavior are both the way possession is achieved - which depends on our treatment manipulation- and possession itself. Furthermore, a clear time-dependence is observed in the data at the player level as accounted for by the Markovian random effects.

Automatic mapping of multiplexed social receptive fields by deep learning and GPU-accelerated 3D videography

Social interactions powerfully impact the brain and the body, but high-resolution descriptions of these important physical interactions and their neural correlates are lacking. Currently, most studies rely on labor-intensive methods such as manual annotation. Scalable and objective tracking methods are required to understand the neural circuits underlying social behavior. Here we describe a hardware/software system and analysis pipeline that combines 3D videography, deep learning, physical modeling, and GPU-accelerated robust optimization, with automatic analysis of neuronal receptive fields recorded in interacting mice. Our system ("3DDD Social Mouse Tracker") is capable of fully automatic multi-animal tracking with minimal errors (including in complete darkness) during complex, spontaneous social encounters, together with simultaneous electrophysiological recordings. We capture posture dynamics of multiple unmarked mice with high spatiotemporal precision (~2 mm, 60 frames/s). A statistical model that relates 3D behavior and neural activity reveals multiplexed 'social receptive fields' of neurons in barrel cortex. Our approach could be broadly useful for neurobehavioral studies of multiple animals interacting in complex low-light environments.

Classifying behavior from short-interval biologging data: An example with GPS tracking of birds

Recent advances in digital data collection have spurred accumulation of immense quantities of data that have potential to lead to remarkable ecological insight, but that also present analytic challenges. In the case of biologging data from birds, common analytical approaches to classifying movement behaviors are largely inappropriate for these massive data sets.We apply a framework for using K-means clustering to classify bird behavior using points from short time interval GPS tracks. K-means clustering is a well-known and computationally efficient statistical tool that has been used in animal movement studies primarily for clustering segments of consecutive points. To illustrate the utility of our approach, we apply K-means clustering to six focal variables derived from GPS data collected at 1-11 s intervals from free-flying bald eagles (Haliaeetus leucocephalus) throughout the state of Iowa, USA. We illustrate how these data can be used to identify behaviors and life-stage- and age-related variation in behavior.After filtering for data quality, the K-means algorithm identified four clusters in >2 million GPS telemetry data points. These four clusters corresponded to three movement states: ascending, flapping, and gliding flight; and one non-moving state: perching. Mapping these states illustrated how they corresponded tightly to expectations derived from natural history observations; for example, long periods of ascending flight were often followed by long gliding descents, birds alternated between flapping and gliding flight.The K-means clustering approach we applied is both an efficient and effective mechanism to classify and interpret short-interval biologging data to understand movement behaviors. Furthermore, because it can apply to an abundance of very short, irregular, and high-dimensional movement data, it provides insight into small-scale variation in behavior that would not be possible with many other analytical approaches.

Diving deeper into the underlying white shark behaviors at Guadalupe Island, Mexico

Fine-scale movement patterns are driven by both biotic (hunting, physiological needs) and abiotic (environmental conditions) factors. The energy balance governs all movement-related strategic decisions.Marine environments can be better understood by considering the vertical component. From 24 acoustic trackings of 10 white sharks in Guadalupe Island, this study linked, for the first time, horizontal and vertical movement data and inferred six different behavioral states along with movement states, through the use of hidden Markov models, which allowed to draw a comprehensive picture of white shark behavior.Traveling was the most frequent state of behavior for white sharks, carried out mainly at night and twilight. In contrast, area-restricted searching was the least used, occurring primarily in daylight hours.Time of day, distance to shore, total shark length, and, to a lesser extent, tide phase affected behavioral states. Chumming activity reversed, in the short term and in a nonpermanent way, the behavioral pattern to a general diel vertical pattern.

Using acoustic telemetry to quantify potential contaminant exposure of Vermilion Rockfish (Sebastes miniatus), Hornyhead Turbot (Pleuronichthys verticalis), and White Croaker (Genyonemus lineatus) at wastewater outfalls in southern California

Contaminant Exposure Models (CEMs) were developed to predict population-level tissue contaminant concentrations in fishes by pairing sediment-bound contaminant concentrations (DDTs, PCBs) and fine-scale acoustic telemetry data from a habitat-associated species (Vermilion Rockfish, Sebastes miniatus), nomadic flatfish species (Hornyhead Turbot, Pleuronichthys verticalis), and nomadic benthic/midwater schooling species (White Croaker, Genyonemus lineatus) tagged near wastewater outfalls in southern California. Model results were compared to contaminant concentrations in tissue samples. The CEMs developed require further refinement before implementation into management efforts but may act as steppingstones to help shift primary monitoring methods away from the regular field collection of fish for tissue contaminant analyses and towards behavioral modeling and habitat mapping. We also developed Kernel Density Estimates that can be used by managers immediately to identify regions that contribute most to contaminant exposure in species of concern. Prioritizing remediation efforts in these areas are likely to be most effective at improving fish health.

A role for lakes in revealing the nature of animal movement using high dimensional telemetry systems

Movement ecology is increasingly relying on experimental approaches and hypothesis testing to reveal how, when, where, why, and which animals move. Movement of megafauna is inherently interesting but many of the fundamental questions of movement ecology can be efficiently tested in study systems with high degrees of control. Lakes can be seen as microcosms for studying ecological processes and the use of high-resolution positioning systems to triangulate exact coordinates of fish, along with sensors that relay information about depth, temperature, acceleration, predation, and more, can be used to answer some of movement ecology's most pressing questions. We describe how key questions in animal movement have been approached and how experiments can be designed to gather information about movement processes to answer questions about the physiological, genetic, and environmental drivers of movement using lakes. We submit that whole lake telemetry studies have a key role to play not only in movement ecology but more broadly in biology as key scientific arenas for knowledge advancement. New hardware for tracking aquatic animals and statistical tools for understanding the processes underlying detection data will continue to advance the potential for revealing the paradigms that govern movement and biological phenomena not just within lakes but in other realms spanning lands and oceans.

A random walk model that accounts for space occupation and movements of a large herbivore

Animal movement has been identified as a key feature in understanding animal behavior, distribution and habitat use and foraging strategies among others. Large datasets of invididual locations often remain unused or used only in part due to the lack of practical models that can directly infer the desired features from raw GPS locations and the complexity of existing approaches. Some of them being disputed for their lack of biological justifications in their design. We propose a simple model of individual movement with explicit parameters, based on a two-dimensional biased and correlated random walk with three forces related to advection (correlation), attraction (bias) and immobility of the animal. These forces can be directly estimated using individual data. We demonstrate the approach by using GPS data of 5 red deer with a high frequency sampling. The results show that a simple random walk template can account for the spatial complexity of wild animals. The practical design of the model is also verified for detecting spatial feature abnormalities and for providing estimates of density and abundance of wild animals. Integrating even more additional features of animal movement, such as individuals’ interactions or environmental repellents, could help to better understand the spatial behavior of wild animals.

Body language signals for rodent social communication

Integration of social cues to initiate adaptive emotional and behavioral responses is a fundamental aspect of animal and human behavior. In humans, social communication includes prominent nonverbal components, such as social touch, gestures and facial expressions. Comparative studies investigating the neural basis of social communication in rodents has historically been centered on olfactory signals and vocalizations, with relatively less focus on non-verbal social cues. Here, we outline two exciting research directions: First, we will review recent observations pointing to a role of social facial expressions in rodents. Second, we will review observations that point to a role of 'non-canonical' rodent body language: body posture signals beyond stereotyped displays in aggressive and sexual behavior. In both sections, we will outline how social neuroscience can build on recent advances in machine learning, robotics and micro-engineering to push these research directions forward towards a holistic systems neurobiology of rodent body language.

Toward Efficient Bayesian Approaches to Inference in Hierarchical Hidden Markov Models for Inferring Animal Behavior

The study of animal behavioral states inferred through hidden Markov models and similar state switching models has seen a significant increase in popularity in recent years. The ability to account for varying levels of behavioral scale has become possible through hierarchical hidden Markov models, but additional levels lead to higher complexity and increased correlation between model components. Maximum likelihood approaches to inference using the EM algorithm and direct optimization of likelihoods are more frequently used, with Bayesian approaches being less favored due to computational demands. Given these demands, it is vital that efficient estimation algorithms are developed when Bayesian methods are preferred. We study the use of various approaches to improve convergence times and mixing in Markov chain Monte Carlo methods applied to hierarchical hidden Markov models, including parallel tempering as an inference facilitation mechanism. The method shows promise for analysing complex stochastic models with high levels of correlation between components, but our results show that it requires careful tuning in order to maximize that potential.

A bird's-eye view on turbulence: seabird foraging associations with evolving surface flow features

Understanding physical mechanisms underlying seabird foraging is fundamental to predict responses to coastal change. For instance, turbulence in the water arising from natural or anthropogenic structures can affect foraging opportunities in tidal seas. Yet, identifying ecologically important localized turbulence features (e.g. upwellings approximately 10-100 m) is limited by observational scale, and this knowledge gap is magnified in volatile predators. Here, using a drone-based approach, we present the tracking of surface-foraging terns (143 trajectories belonging to three tern species) and dynamic turbulent surface flow features in synchrony. We thereby provide the earliest evidence that localized turbulence features can present physical foraging cues. Incorporating evolving vorticity and upwelling features within a hidden Markov model, we show that terns were more likely to actively forage as the strength of the underlying vorticity feature increased, while conspicuous upwellings ahead of the flight path presented a strong physical cue to stay in transit behaviour. This clearly encapsulates the importance of prevalent turbulence features as localized foraging cues. Our quantitative approach therefore offers the opportunity to unlock knowledge gaps in seabird sensory and foraging ecology on hitherto unobtainable scales. Finally, it lays the foundation to predict responses to coastal change to inform sustainable ocean development.

Hidden Markov models identify major movement modes in accelerometer and magnetometer data from four albatross species

BACKGROUND

Inertial measurement units (IMUs) with high-resolution sensors such as accelerometers are now used extensively to study fine-scale behavior in a wide range of marine and terrestrial animals. Robust and practical methods are required for the computationally-demanding analysis of the resulting large datasets, particularly for automating classification routines that construct behavioral time series and time-activity budgets. Magnetometers are used increasingly to study behavior, but it is not clear how these sensors contribute to the accuracy of behavioral classification methods. Development of effective  classification methodology is key to understanding energetic and life-history implications of foraging and other behaviors.

METHODS

We deployed accelerometers and magnetometers on four species of free-ranging albatrosses and evaluated the ability of unsupervised hidden Markov models (HMMs) to identify three major modalities in their behavior: 'flapping flight', 'soaring flight', and 'on-water'. The relative contribution of each sensor to classification accuracy was measured by comparing HMM-inferred states with expert classifications identified from stereotypic patterns observed in sensor data.

RESULTS

HMMs provided a flexible and easily interpretable means of classifying behavior from sensor data. Model accuracy was high overall (92%), but varied across behavioral states (87.6, 93.1 and 91.7% for 'flapping flight', 'soaring flight' and 'on-water', respectively). Models built on accelerometer data alone were as accurate as those that also included magnetometer data; however, the latter were useful for investigating slow and periodic behaviors such as dynamic soaring at a fine scale.

CONCLUSIONS

The use of IMUs in behavioral studies produces large data sets, necessitating the development of computationally-efficient methods to automate behavioral classification in order to synthesize and interpret underlying patterns. HMMs provide an accessible and robust framework for analyzing complex IMU datasets and comparing behavioral variation among taxa across habitats, time and space.

Uncovering ecological state dynamics with hidden Markov models

Ecological systems can often be characterised by changes among a finite set of underlying states pertaining to individuals, populations, communities or entire ecosystems through time. Owing to the inherent difficulty of empirical field studies, ecological state dynamics operating at any level of this hierarchy can often be unobservable or 'hidden'. Ecologists must therefore often contend with incomplete or indirect observations that are somehow related to these underlying processes. By formally disentangling state and observation processes based on simple yet powerful mathematical properties that can be used to describe many ecological phenomena, hidden Markov models (HMMs) can facilitate inferences about complex system state dynamics that might otherwise be intractable. However, HMMs have only recently begun to gain traction within the broader ecological community. We provide a gentle introduction to HMMs, establish some common terminology, review the immense scope of HMMs for applied ecological research and provide a tutorial on implementation and interpretation. By illustrating how practitioners can use a simple conceptual template to customise HMMs for their specific systems of interest, revealing methodological links between existing applications, and highlighting some practical considerations and limitations of these approaches, our goal is to help establish HMMs as a fundamental inferential tool for ecologists.

Movement patterns of large juvenile loggerhead turtles in the Mediterranean Sea: Ontogenetic space use in a small ocean basin

Mechanisms that determine how, where, and when ontogenetic habitat shifts occur are mostly unknown in wild populations. Differences in size and environmental characteristics of ontogenetic habitats can lead to differences in movement patterns, behavior, habitat use, and spatial distributions across individuals of the same species. Knowledge of juvenile loggerhead turtles' dispersal, movements, and habitat use is largely unknown, especially in the Mediterranean Sea. Satellite relay data loggers were used to monitor movements, diving behavior, and water temperature of eleven large juvenile loggerhead turtles (Caretta caretta) deliberately caught in an oceanic habitat in the Mediterranean Sea. Hidden Markov models were used over 4,430 spatial locations to quantify the different activities performed by each individual: transit, low-, and high-intensity diving. Model results were then analyzed in relation to water temperature, bathymetry, and distance to the coast. The hidden Markov model differentiated between bouts of area-restricted search as low- and high-intensity diving, and transit movements. The turtles foraged in deep oceanic waters within 60 km from the coast as well as above 140 km from the coast. They used an average area of 194,802 km2, where most individuals used the deepest part of the Southern Tyrrhenian Sea with the highest seamounts, while only two switched to neritic foraging showing plasticity in foraging strategies among turtles of similar age classes. The foraging distribution of large juvenile loggerhead turtles, including some which were of the minimum size of adults, in the Tyrrhenian Sea is mainly concentrated in a relatively small oceanic area with predictable mesoscale oceanographic features, despite the proximity of suitable neritic foraging habitats. Our study highlights the importance of collecting high-resolution data about species distribution and behavior across different spatio-temporal scales and life stages for implementing conservation and dynamic ocean management actions.

An application of upscaled optimal foraging theory using hidden Markov modelling: year-round behavioural variation in a large arctic herbivore

BACKGROUND

In highly seasonal environments, animals face critical decisions regarding time allocation, diet optimisation, and habitat use. In the Arctic, the short summers are crucial for replenishing body reserves, while low food availability and increased energetic demands characterise the long winters (9-10 months). Under such extreme seasonal variability, even small deviations from optimal time allocation can markedly impact individuals' condition, reproductive success and survival. We investigated which environmental conditions influenced daily, seasonal, and interannual variation in time allocation in high-arctic muskoxen (Ovibos moschatus) and evaluated whether results support qualitative predictions derived from upscaled optimal foraging theory.

METHODS

Using hidden Markov models (HMMs), we inferred behavioural states (foraging, resting, relocating) from hourly positions of GPS-collared females tracked in northeast Greenland (28 muskox-years). To relate behavioural variation to environmental conditions, we considered a wide range of spatially and/or temporally explicit covariates in the HMMs.

RESULTS

While we found little interannual variation, daily and seasonal time allocation varied markedly. Scheduling of daily activities was distinct throughout the year except for the period of continuous daylight. During summer, muskoxen spent about 69% of time foraging and 19% resting, without environmental constraints on foraging activity. During winter, time spent foraging decreased to 45%, whereas about 43% of time was spent resting, mediated by longer resting bouts than during summer.

CONCLUSIONS

Our results clearly indicate that female muskoxen follow an energy intake maximisation strategy during the arctic summer. During winter, our results were not easily reconcilable with just one dominant foraging strategy. The overall reduction in activity likely reflects higher time requirements for rumination in response to the reduction of forage quality (supporting an energy intake maximisation strategy). However, deep snow and low temperatures were apparent constraints to winter foraging, hence also suggesting attempts to conserve energy (net energy maximisation strategy). Our approach provides new insights into the year-round behavioural strategies of the largest Arctic herbivore and outlines a practical example of how to approximate qualitative predictions of upscaled optimal foraging theory using multi-year GPS tracking data.