Source-linked AI summary
Animal Movement Tools (amt): R-Package for Managing Tracking Data and Conducting Habitat Selection Analyses
Johannes Signer, John Fieberg, Tal Avgar
TL;DR
Growing telemetry datasets create data-management and analysis challenges in movement ecology. This paper presents amt, an R package for managing tracking data, fitting step-selection functions, and simulating space use, illustrated through a four-step fisher analysis.
Problem
Growing fine-scale animal location datasets create challenges requiring efficient, straightforward tools for movement-data management and analysis.
Method
The amt package supports telemetry-data preparation, exploratory analysis, step-selection-function fitting, and simulation of space use from fitted models.
Results
Using fisher data, the authors demonstrate a four-step workflow spanning data management, exploration, model fitting, and simulation from fitted models.
Takeaways & Limitations
amt provides functions intended to support reproducible telemetry workflows and broader use of integrated step-selection functions and predicted space-use patterns.
Takeaways & Limitations
Currently, amt cannot simulate utilization distributions from models including interactions between movement and other covariates.
Abstract
from arXiv · showhide
1. Advances in tracking technology have led to an exponential increase in animal location data, greatly enhancing our ability to address interesting questions in movement ecology, but also presenting new challenges related to data manage- ment and analysis. 2. Step-Selection Functions (SSFs) are commonly used to link environmental covariates to animal location data collected at fine temporal resolution. SSFs are estimated by comparing observed steps connecting successive animal locations to random steps, using a likelihood equivalent of a Cox proportional hazards model. By using common statistical distributions to model step length and turn angle distributions, and including habitat- and movement-related covariates (functions of distances between points, angular deviations), it is possible to make inference regarding habitat selection and movement processes, or to control one process while investigating the other. The fitted model can also be used to estimate utilization distributions and mechanistic home ranges. 3. Here, we present the R-package amt (animal movement tools) that allows users to fit SSFs to data and to simulate space use of animals from fitted models. The amt package also provides tools for managing telemetry data. 4. Using fisher (Pekania pennanti ) data as a case study, we illustrate a four-step approach to the analysis of animal movement data, consisting of data management, exploratory data analysis, fitting of models, and simulating from fitted models.
Author’s Affiliations · Summary
The amt R package addresses growing challenges in managing and analyzing animal tracking data. It supports step-selection analyses, telemetry-data management, and simulation-based investigation of animal space use through a four-step workflow.
- Summary: Tracking technology has exponentially increased animal location data, expanding movement-ecology opportunities while creating data-management and analysis challenges.This growth motivates tools for organizing and analyzing telemetry data.
- Summary: Step-Selection Functions link fine-resolution animal locations with environmental covariates by comparing observed steps with random steps.Their likelihood is equivalent to that of a Cox proportional hazards model.
- Summary: Common distributions for step lengths and turn angles support modeling movement processes alongside habitat-related covariates.Covariates can include functions of distances between points and angular deviations.
- Summary: The fitted step-selection model can support inference about habitat selection and movement processes while controlling one process to investigate the other.It can also be used to estimate utilization distributions and mechanistic home ranges.
- Summary: The amt R package enables users to fit Step-Selection Functions and simulate animal space use from fitted models.It also provides tools for managing telemetry data.
- Summary: A fisher (Pekania pennanti) case study illustrates a four-step animal-movement analysis workflow.The steps are data management, exploratory data analysis, model fitting, and simulation from fitted models.
Introduction
Fine-scale biotelemetry has expanded movement-ecology research while making data preparation and analysis more challenging. The amt R package addresses this need with a coherent workflow for managing tracking data and fitting step-selection models.
- Advances in technology have produced large collections of fine-scale animal biotelemetry data and spurred new quantitative methods for studying movement.
- Resource Selection Functions compare environmental covariates at observed animal locations with covariates at random locations within an availability domain.
- Integrated step-selection functions support simultaneous inference on habitat selection and movement processes by sampling random steps from analytical distributions and incorporating movement-related covariates.
- The amt package provides a flexible, coherent R workflow for animal tracking data, addressing the complexity of data preparation before fitting SSFs and iSSFs.
Functionality
amt organizes animal-tracking analysis into four steps: data preparation and management, exploratory analysis, modeling, and simulation. The workflow supports quality control, movement and space-use summaries, hypothesis testing, and simulation of derived quantities from fitted models.
- Functionality: The workflow comprises four steps: data preparation and management, exploratory and descriptive analysis, modeling, and simulation.These steps are presented as the typical workflow for analyzing animal tracking data.
- Data preparation, inspection, and management: Data preparation includes inspecting gaps, resampling tracks when needed, and adjusting coordinate reference systems.Users also perform data-quality checks, inspect sampling-rate variability and outliers, remove capture-effect periods, and form regular bursts.
- Exploratory data analysis and descriptive analyses: Exploratory analysis examines movement characteristics and trajectory or space-use summaries across animals or time periods.Examples include step-length and turning-angle distributions, net squared displacement, path sinuosity, and home-range area.
- Modeling: Modeling fits models to test hypotheses about animal movement, habitat selection, and space use.amt supports generating random points or steps and extracting environmental covariates for observed and random steps before fitting RSFs and (i)SSFs.
- Simulation: Simulation uses fitted models to estimate space use, identify high-use corridors, and assess model power or fit.amt specifically provides means to simulate space use from fitted SSFs.
Case study
The case study demonstrates managing and resampling fisher tracking data, constructing steps, fitting an integrated step-selection function, and interpreting individual and population-level habitat selection. Ricky T preferred forested wetlands, while analyses extended to several animals revealed general wetland-forest and natural-area preferences with variability for crops.
- Data management: The analysis loaded all fisher data, removed locations missing longitude or latitude, and selected Ricky T (id 1016) for individual analysis.The workflow then extended similar analyses to several animals for population-level inference.
- Data management: The track was converted from geographic coordinates (EPSG 4326) to projected NAD83 coordinates (EPSG 5070) before analysis.amt’s mk_track creates the track from coordinate and time columns and allows a coordinate reference system to be set.
- Exploratory data analysis: The dataset contained 8957 locations, with a median interval of roughly 2 minutes, but the track was resampled to 10 minutes with a 1-minute tolerance.Observed intervals ranged from 0.1 to 1208 minutes.
- Step construction: The resulting steps included 1501 rows and 11 attributes, including coordinates, times, step lengths, turning angles, and time differences.Turning angles cannot be calculated for steps not preceded by a valid step.
- Model fitting: The fitted model indicated preference for forested wetlands, no day–night difference in habitat preference, and day- and night-specific modification of the gamma step-length shape.The integrated step-selection function used gamma step lengths, von Mises turn angles, and nine random steps per observed step.
- Population-level inference: At the population level, all fishers appeared to prefer wetland forests and natural areas over developed areas, while crop coefficients showed considerable among-animal variability.The fitted model could be inspected across several animals to assess population-level effects.
Discussion and Outlook
The discussion presents amt as a reproducible toolkit for fitting and applying iSSFs, managing telemetry data, and expanding movement-ecology analyses. It also identifies current simulation limitations and future development priorities.
- Applications: amt fits SSFs and uses individual iSSFs to simulate utilization distributions as stochastic approximations of mechanistic home-range models.The package supports analyses of temporal movement and habitat-selection patterns at individual and population levels.
- Expected contributions: amt is expected to support more reproducible data and analysis workflows while widening iSSF use and enabling joint modeling of movement and habitat selection.The discussion highlights modeling how landscape features influence both processes.
- Capabilities and development: Beyond SSF fitting, amt calculates home ranges, estimates RSFs, supports telemetry-data utilities, interfaces with other packages, and plans more data-quality-assurance functions.Future development is explicitly focused on increased functionality for data quality assurance.
- Limitations: Currently, amt cannot simulate utilization distributions from models including interactions between movement, such as speed or turning angles, and other covariates.The authors hope to implement utilization-distribution estimation for more sophisticated models.
Figures
The figures illustrate exploratory distributions of movement behavior, kernel-based simulation of utilization distributions, and relative selection-strength estimates across land-use classes. They also show how day–night movement-speed differences affect transient and steady-state utilization distributions.
- Exploratory data analysis: Exploratory analysis compares empirical step-length and turning-angle distributions across habitats and between day and night for one fisher.The example individual is Ricky T (id: 1016), with forested wetland and other habitats shown separately.
- Simulated utilization distributions: Simulated utilization distributions combine a movement kernel centered on the animal’s current position with a habitat kernel to sample the next step.The next step is sampled with probability proportional to the product of the two kernels.
- Simulated utilization distributions: Day–night differences in movement speeds appear more strongly in transient than steady-state utilization distributions.These differences are shown in transient UD panels C and E and, to a lesser extent, steady-state UD panels D and F.
- Relative selection strength: Relative selection-strength estimates are presented with 95% confidence intervals for land-use classes, using wetland forests and wet areas as the reference class.Animal identity and sex are distinguished visually, while population-level estimates and confidence intervals are also shown.
Tables
Table 1 presents the coefficients of the fitted integrated Step Selection Function.
- Table 1 reports coefficients for the fitted integrated Step Selection Function.
- The table concerns an integrated Step Selection Function.
- The reported coefficients are associated with a fitted model.
Supplement S1
Supplement S1 provides reproducible amt workflows for managing telemetry data, preparing movement steps, fitting integrated step-selection models, and simulating utilization distributions. It also demonstrates applying these analyses across multiple fisher individuals and summarizing individual-level coefficients by term.
- Movement-data preparation: Movement steps are filtered by burst size, generated or resampled, associated with land-use covariates, and augmented with time-of-day and movement variables.The workflows extract covariates at step starts, ends, or both, and derive log step length and cosine turn angle for modeling.
- Exploratory data analysis: Exploratory plots compare step-length and turn-angle densities across land-use classes and times of day.The code produces faceted density plots for step length and turn angle, grouped by time of day and faceted by land use.
- Model fitting: Integrated step-selection models include habitat, step-length, and time-of-day effects with interactions, using step-stratified conditional likelihoods.The examples fit ISSFs with alternative step-length specifications and apply bootstrap resampling of step strata before refitting conditional logistic models.
- Utilization-distribution simulation: Fitted movement and habitat kernels generate transient and steady-state utilization distributions for day and night conditions.The simulations use n = 1e7 steps for steady-state distributions and repeated n = 72-step simulations for transient distributions.
- Data management: The examples load fisher telemetry data, retain selected tag identifiers, nest tracks by individual, assign sex, transform coordinates, and summarize sampling rates.The multi-animal workflow retains six identifiers and assigns three females and three males before unnesting sampling-rate summaries.