[Paper Review] The Langevin diffusion as a continuous-time model of animal movement and habitat selection

Théo Michelot, Marie‐Pierre Étienne+1 more|arXiv (Cornell University)|Oct 24, 2018

Marine animal studies overviewEnvironmental Science3 citations

TL;DR

This paper proposes a continuous-time Langevin diffusion model for animal movement that explicitly links habitat selection to space use through a stationary utilisation distribution. By modeling movement as a stochastic process with a potential function derived from spatial covariates, the method enables robust inference of long-term habitat preferences from irregularly sampled telemetry data using pseudo-likelihood estimation, outperforming traditional methods in handling correlated movement data.

ABSTRACT

1. The utilisation distribution describes the relative probability of use of a spatial unit by an animal. It is natural to think of it as the long-term consequence of the animal's short-term movement decisions: it is the accumulation of small displacements which, over time, gives rise to global patterns of space use. However, most utilisation distribution models either ignore the underlying movement, assuming the independenceof observed locations, or are based on simplistic Brownian motion movement rules. 2. We introduce a new continuous-time model of animal movement, based on the Langevin diffusion. This stochastic process has an explicit stationary distribution, conceptually analogous to the idea of the utilisation distribution, and thus provides an intuitive framework to integrate movement and space use. We model the stationary (utilisation) distribution with a resource selection function to link the movement to spatial covariates, and allow inference into habitat selection. 3. Standard approximation techniques can be used to derive the pseudo-likelihood of the Langevin diffusion movement model, and to estimate habitat preference and movement parameters from tracking data. We investigate the performance of the method on simulated data, and discuss its sensitivity to the time scale of the sampling. We present an example of its application to tracking data of Stellar sea lions (Eumetopiasjubatus). 4. Due to its continuous-time formulation, this method can be applied to irregular telemetry data. It provides a rigorous framework to estimate long-term habitat selection from correlated movement data.

Motivation & Objective

• To develop a mechanistic, continuous-time model of animal movement that integrates habitat selection with long-term space use.
• To address the limitations of existing methods that assume independent locations or unrealistic Brownian motion in utilisation distribution estimation.
• To provide a framework that links movement dynamics directly to environmental covariates through a parametric, stationary distribution.
• To enable reliable inference of habitat selection from high-resolution, irregularly sampled telemetry data.
• To offer a robust alternative to discrete-grid or interpolation-based models by using continuous-space formulation.

Proposed method

• The model uses the Langevin diffusion process, whose stationary distribution corresponds to the utilisation distribution, enabling direct linkages between movement and space use.
• The drift term of the diffusion is derived from a potential function that incorporates spatial covariates via a resource selection function.
• Pseudo-likelihood estimation is performed using the Euler discretization scheme to approximate the transition density of the continuous-time process.
• The method handles irregularly spaced telemetry data without interpolation, preserving temporal correlation in movement.
• A two-stage approach is used in the case study: first filtering Argos locations with a continuous-time correlated random walk, then fitting the Langevin model to the filtered tracks.
• The framework allows for hierarchical state-space modeling with measurement error by integrating the observation model directly into the likelihood.

Experimental results

Research questions

• RQ1 How can a continuous-time movement model be constructed such that its stationary distribution represents the utilisation distribution?
• RQ2 Can habitat selection be reliably estimated from irregularly sampled telemetry data using a mechanistic movement model?
• RQ3 How does the performance of the Euler discretization scheme compare to more refined schemes like Ozaki in estimating habitat selection parameters?
• RQ4 To what extent does the Langevin model outperform non-mechanistic methods in capturing true habitat preferences?
• RQ5 Can the model be extended to include measurement error in telemetry data within a unified inference framework?

Key findings

• The Langevin diffusion model provides a closed-form stationary distribution that directly represents the utilisation distribution, enabling a mechanistic link between movement and space use.
• Pseudo-likelihood estimation using the Euler scheme yields reliable parameter estimates, even with irregular sampling, and outperforms the Ozaki scheme in practice due to numerical stability.
• The model successfully infers habitat selection from simulated data, accurately recovering the true resource selection function across various time scales.
• Application to Steller sea lion telemetry data reveals distinct habitat preferences, with the model detecting meaningful spatial use patterns consistent with ecological expectations.
• The two-stage approach (filtering then modeling) is effective but suboptimal; integrating measurement error into a unified state-space model is recommended for improved inference.
• The method is robust to data irregularity and does not require spatial or temporal binning, making it suitable for modern high-resolution telemetry data.