Multi-mode movement decisions across widely ranging behavioral processes

An introduction to statistical models used to characterize species-habitat associations with animal movement data

Understanding species-habitat associations is fundamental to ecological sciences and for species conservation. Consequently, various statistical approaches have been designed to infer species-habitat associations. Due to their conceptual and mathematical differences, these methods can yield contrasting results. In this paper, we describe and compare commonly used statistical models that relate animal movement data to environmental data. Specifically, we examined selection functions which include resource selection function (RSF) and step-selection function (SSF), as well as hidden Markov models (HMMs) and related methods such as state-space models. We demonstrate differences in assumptions while highlighting advantages and limitations of each method. Additionally, we provide guidance on selecting the most appropriate statistical method based on the scale of the data and intended inference. To illustrate the varying ecological insights derived from each statistical model, we apply them to the movement track of a single ringed seal (Pusa hispida) in a case study. Through our case study, we demonstrate that each model yields varying ecological insights. For example, while the selection coefficient values from RSFs appear to show a stronger positive relationship with prey diversity than those of the SSFs, when we accounted for the autocorrelation in the data none of these relationships with prey diversity were statistically significant. Furthermore, the HMM reveals variable associations with prey diversity across different behaviors, for example, a positive relationship between prey diversity and a slow-movement behaviour. Notably, the three models identified different "important" areas. This case study highlights the critical significance of selecting the appropriate model as an essential step in the process of identifying species-habitat relationships and specific areas of importance. Our comprehensive review provides the foundational information required for making informed decisions when choosing the most suitable statistical methods to address specific questions, such as identifying expansive corridors or protected zones, understanding movement patterns, or studying behaviours. In addition, this study informs researchers with the necessary tools to apply these methods effectively.

Combined effects of landscape fragmentation and sampling frequency of movement data on the assessment of landscape connectivity

BACKGROUND

Network theory is largely applied in real-world systems to assess landscape connectivity using empirical or theoretical networks. Empirical networks are usually built from discontinuous individual movement trajectories without knowing the effect of relocation frequency on the assessment of landscape connectivity while theoretical networks generally rely on simple movement rules. We investigated the combined effects of relocation sampling frequency and landscape fragmentation on the assessment of landscape connectivity using simulated trajectories and empirical high-resolution (1 Hz) trajectories of Alpine ibex (Capra ibex). We also quantified the capacity of commonly used theoretical networks to accurately predict landscape connectivity from multiple movement processes.

METHODS

We simulated forager trajectories from continuous correlated biased random walks in simulated landscapes with three levels of landscape fragmentation. High-resolution ibex trajectories were reconstructed using GPS-enabled multi-sensor biologging data and the dead-reckoning technique. For both simulated and empirical trajectories, we generated spatial networks from regularly resampled trajectories and assessed changes in their topology and information loss depending on the resampling frequency and landscape fragmentation. We finally built commonly used theoretical networks in the same landscapes and compared their predictions to actual connectivity.

RESULTS

We demonstrated that an accurate assessment of landscape connectivity can be severely hampered (e.g., up to 66% of undetected visited patches and 29% of spurious links) when the relocation frequency is too coarse compared to the temporal dynamics of animal movement. However, the level of landscape fragmentation and underlying movement processes can both mitigate the effect of relocation sampling frequency. We also showed that network topologies emerging from different movement behaviours and a wide range of landscape fragmentation were complex, and that commonly used theoretical networks accurately predicted only 30-50% of landscape connectivity in such environments.

CONCLUSIONS

Very high-resolution trajectories were generally necessary to accurately identify complex network topologies and avoid the generation of spurious information on landscape connectivity. New technologies providing such high-resolution datasets over long periods should thus grow in the movement ecology sphere. In addition, commonly used theoretical models should be applied with caution to the study of landscape connectivity in real-world systems as they did not perform well as predictive tools.

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.

Time-varying habitat selection analysis: A model and applications for studying diel, seasonal, and post-release changes

Resource selection functions are commonly used to evaluate animals' habitat selection, for example, the disproportionate use of habitats relative to their availability. While environmental conditions or animal motivations may vary over time, sometimes in an unknown manner, studying changes in habitat selection usually requires an a priori segmentation of time in distinct periods. This limits our ability to precisely answer the question "When is an animal's habitat selection changing?" Here, we present a straightforward and flexible alternative approach based on fitting dynamic logistic models to used/available data. First, using simulated datasets, we demonstrate that dynamic logistic models perform well in recovering temporal variations in habitat selection. We then show real-world applications for studying diel, seasonal, and post-release changes in the habitat selection of the blue wildebeest (Connochaetes taurinus). Dynamic logistic models allow the study of temporal changes in habitat selection in a framework consistent with resource selection functions but without the need to segment time in distinct periods, which can be a difficult task when little is known about the process studied or may obscure interindividual variability in timing of change. These models should undoubtedly find their place in the movement ecology toolbox. We provide R scripts to facilitate their adoption. We also encourage future research to focus on how to account for temporal autocorrelation in location data, as this would allow statistical inference from location data collected at a high frequency, an increasingly common situation.

A multistate Langevin diffusion for inferring behavior-specific habitat selection and utilization distributions

The identification of important habitat and the behavior(s) associated with it is critical to conservation and place-based management decisions. Behavior also links life-history requirements and habitat use, which are key to understanding why animals use certain habitats. Animal population studies often use tracking data to quantify space use and habitat selection, but they typically either ignore movement behavior (e.g., foraging, migrating, nesting) or adopt a two-stage approach that can induce bias and fail to propagate uncertainty. We develop a habitat-driven Langevin diffusion for animals that exhibit distinct movement behavior states, thereby providing a novel single-stage statistical method for inferring behavior-specific habitat selection and utilization distributions in continuous time. Practitioners can customize, fit, assess, and simulate our integrated model using the provided R package. Simulation experiments demonstrated that the model worked well under a range of sampling scenarios as long as observations were of sufficient temporal resolution. Our simulations also demonstrated the importance of accounting for different behaviors and the misleading inferences that can result when these are ignored. We provide case studies using plains zebra (Equus quagga) and Steller sea lion (Eumetopias jubatus) telemetry data. In the zebra example, our model identified distinct "encamped" and "exploratory" states, where the encamped state was characterized by strong selection for grassland and avoidance of other vegetation types, which may represent selection for foraging resources. In the sea lion example, our model identified distinct movement behavior modes typically associated with this marine central-place forager and, unlike previous analyses, found foraging-type movements to be associated with steeper offshore slopes characteristic of the continental shelf, submarine canyons, and seamounts that are believed to enhance prey concentrations. This is the first single-stage approach for inferring behavior-specific habitat selection and utilization distributions from tracking data that can be readily implemented with user-friendly software. As certain behaviors are often more relevant to specific conservation or management objectives, practitioners can use our model to help inform the identification and prioritization of important habitats. Moreover, by linking individual-level movement behaviors to population-level spatial processes, the multistate Langevin diffusion can advance inferences at the intersection of population, movement, and landscape ecology.

Flexible hidden Markov models for behaviour-dependent habitat selection

BACKGROUND

There is strong incentive to model behaviour-dependent habitat selection, as this can help delineate critical habitats for important life processes and reduce bias in model parameters. For this purpose, a two-stage modelling approach is often taken: (i) classify behaviours with a hidden Markov model (HMM), and (ii) fit a step selection function (SSF) to each subset of data. However, this approach does not properly account for the uncertainty in behavioural classification, nor does it allow states to depend on habitat selection. An alternative approach is to estimate both state switching and habitat selection in a single, integrated model called an HMM-SSF.

METHODS

We build on this recent methodological work to make the HMM-SSF approach more efficient and general. We focus on writing the model as an HMM where the observation process is defined by an SSF, such that well-known inferential methods for HMMs can be used directly for parameter estimation and state classification. We extend the model to include covariates on the HMM transition probabilities, allowing for inferences into the temporal and individual-specific drivers of state switching. We demonstrate the method through an illustrative example of plains zebra (Equus quagga), including state estimation, and simulations to estimate a utilisation distribution.

RESULTS

In the zebra analysis, we identified two behavioural states, with clearly distinct patterns of movement and habitat selection ("encamped" and "exploratory"). In particular, although the zebra tended to prefer areas higher in grassland across both behavioural states, this selection was much stronger in the fast, directed exploratory state. We also found a clear diel cycle in behaviour, which indicated that zebras were more likely to be exploring in the morning and encamped in the evening.

CONCLUSIONS

This method can be used to analyse behaviour-specific habitat selection in a wide range of species and systems. A large suite of statistical extensions and tools developed for HMMs and SSFs can be applied directly to this integrated model, making it a very versatile framework to jointly learn about animal behaviour, habitat selection, and space use.