Navigating through the R packages for movement

A Movement Ecology background

• Tracking packages are defined as R packages that create, transform, or analyze data with at least two spatial coordinates and a time index, supporting a workflow from biologging data to analysis.Packages analyzing accelerometry series for behavioral change alone fall outside this definition.Source: A Movement Ecology background · p. 4
• The review organizes package functions into pre-processing, post-processing, and analysis, covering visualization, track description, path reconstruction, behavior, space use, and simulation.Pre-processing converts data to (x, y, t); post-processing can address errors and derive variables before analysis.Source: A Movement Ecology background · pp. 4–5
• The review assesses documentation quality through a user survey and evaluates package links and compatibility.The paper treats documentation as important for making packages accessible to users.Source: A Movement Ecology background · p. 5
• The review aims to give movement ecologists package-selection criteria and developers recommendations to improve package utility and strengthen community links.Source: A Movement Ecology background · p. 5

Data sources

• Package discovery combined CRAN task-view listings, package suggestions, web searches, and exchanges with ecologists.Search terms included trajectory, movement, or spatiotemporal with package and R.Source: Data sources · p. 6
• Packages were retained only if they matched the paper’s definition of a tracking package.Packages removed from CRAN or described as very early versions on GitHub were discarded.Source: Data sources · p. 6
• Information gathering ran from March to August 2018, so newer packages and later package versions may not be represented.The repository includes the reviewed versions and package characteristics.Source: Data sources · p. 6
• Package citations refer to the output of R’s citation(package) at manuscript preparation, and may not match the version studied.Source: Data sources · p. 7

The R packages

• 58 packages were identified for tracking-data processing and analysis, with some spanning multiple workflow stages and others focusing on one.Functions were classified mainly from package documentation, supplemented by other sources.Source: The R packages · p. 7

Pre-processing

• Pre-processing depends on biologging device type; the review groups light-based geolocation methods as threshold, template-fitting, or twilight-free.Threshold methods use chosen irradiance levels; template-fitting models observed against theoretical light; twilight-free methods avoid identifying twilight events.Source: Pre-processing · p. 8, GLS data pre-processing · pp. 8–10
• Dead-reckoning combines sensor-derived speed and direction with a known starting location, but vector calculations accumulate error and require independent locations for correction.The reviewed DR packages use different methods, and the paper advises users to understand those methods before use.Source: Dead-reckoning using accelerometry and… · pp. 12–13

Post-processing

• Post-processing includes filtering implausible PTT locations, removing duplicates or unrealistic speeds, and rediscretizing or interpolating tracks to regular steps.argosfilter and SDLfilter target PTT locations; several other packages support cleaning and track compression.Source: Data cleaning · pp. 14–15, Data compression · p. 15
• Packages can compute movement metrics such as speed, step length, and orientation, either automatically during data-class conversion or when users request them.Source: Computation of metrics · p. 16
• Some tools transform device-specific records into individual movement data, including RFID detection events and acoustic telemetry detections.feedr derives station arrival, departure, and visitation-time statistics; VTrack handles acoustic telemetry records.Source: Computation of metrics · pp. 16–17

Visualization

• anipaths and moveVis animate tracks, with options for timing, interpolation, background maps, and displaying individual covariates or environmental statistics.moveVis can preserve supplied timestamps or display all individuals as if their tracks began at time zero.Source: Visualization · pp. 17–18

Track description

• Track-description packages summarize movement with metrics such as distance and straightness, while trackeR reports workout effort and activity measures for human GPS tracks.movementAnalysis depends on adehabitat, which was removed from CRAN in 2018 and superseded by adehabitatLT, adehabitatHR, and adehabitatMA.Source: Track description · p. 18

Path reconstruction

• Path reconstruction packages span GPS–dead-reckoning correction, general continuous-time movement models, and location estimation tailored to GLS or PTT data.TrackReconstruction forces paths through GPS fixes, whereas BayesianAnimalTracker estimates a posterior using a Brownian Bridge prior; other packages fit movement models or improve estimated locations.Source: Path reconstruction · p. 19, Improving location estimation from GLS… · p. 19, Combining dead-reckoning and GPS data · pp. 20–21

Behavioral pattern identification

• Behavioral methods include independent classification, temporal segmentation, and hidden Markov models, with packages offering different model structures and inputs.Examples range from supervised random forests and migration-focused segmentation to HMMs using movement variables or net squared displacement.Source: Behavioral pattern identification · p. 22, Non-sequential classification or… · p. 22, Segmentation methods · p. 23

Home-range estimation

• Home-range packages offer methods from minimum convex polygons to utilization distributions, including approaches that account for temporal autocorrelation.Methods include local convex hulls, Brownian-bridge and biased-random-bridge kernels, and product-kernel algorithms; amt also estimates movement-based UDs from fitted step-selection functions.Source: Home range · pp. 25–26
• T-LoCoH constructs home-range hulls using time-scale distance and neighbor criteria, and computes revisitation counts and durations as space-use metrics.The package was originally implemented for GPS data but can be used with tracking data generally.Source: Home range · p. 26

Habitat selection

• Habitat-selection tools include step-selection functions that compare observed steps with alternative steps from the same starting point, and integrated versions that model movement and resource selection together.amt supports SSFs and iSSFs; ctmcmove uses a generalized linear model framework with rasterized features and an imputed, discretized track.Source: Habitat use · pp. 27–28

Other approaches

• Other space-use approaches include moveNT’s network analysis of pixel transitions and rsMove’s random-forest identification of feeding sites from environmental data.recurse calculates revisitation counts and durations for user-defined areas; feedr and VTrack summarize visits to logger stations or receivers.Source: Non-conventional approaches for space… · pp. 28–29

Trajectory simulation

• Trajectory simulation packages generate tracks from fitted or specified movement models, with SiMRiv also allowing user-defined environmental resistance to constrain movement.Available models include hidden Markov models, correlated random walks, Brownian motion, Lévy walks, and Ornstein-Uhlenbeck processes; simulation can support hypothesis comparisons or parametric bootstrapping.Source: Trajectory simulation · pp. 29–30

Other analyses of tracking data

• Other analyses include interaction metrics, trajectory similarity measures, caribou population-size estimation, wind-speed inference from circling bird segments, and trajectory database management.The caribou method requires user-specified group-size, collar-count, and detection-model parameters.Source: Interactions · p. 31, Movement similarity · p. 31, Population size · pp. 31–32

Analysis of biologging but not tracking data

• For biologging data without tracking locations, diveMove and rbl analyze TDR data, while six packages focus on human accelerometry, including activity classification and summaries.diveMove identifies dives and their phases; rbl supports prey-catch and swimming-effort analyses, while acc, GGIR, and pawacc classify activity levels using user-provided thresholds.Source: Analysis of biologging but not… · pp. 32–33

Survey method

• The survey was completed by 225 people between August and October 2018 and asked about documentation helpfulness and package-use frequency.The authors direct readers to supplementary material for question wording, detailed results, and discussion of survey representativity.Source: Packages documentation · p. 34

Survey results

• 10 of the 26 packages with at least 10 survey respondents had adequate documentation, defined as more than 75% rating it good or excellent.momentuHMM received a 93.8% good-or-excellent rating.Source: Packages documentation · pp. 34–35

Implications

• The authors suggest developers use the survey results to decide whether documentation improvements could make packages usable by more researchers.They describe manuals, vignettes, tutorials, and published articles as documentation that can guide package use and explain methods.Source: Packages documentation · pp. 34–35

• Package links include dependencies on functions from other packages and suggestions to use another package for a task such as post-processing.The authors group Depends on, Imports, and Linking to under the concept of dependency.Source: Links between the packages · p. 35
• Developers define data classes to specify function-compatible formats, while coercion methods allow compatibility with other packages’ data classes.amt provides coercion methods for data classes from packages it suggests.Source: Links between the packages · pp. 35–36
• Thirty-nine packages had some connections, while 19 (32%) worked in isolation; adehabitatLT and move were the most suggested or depended-on packages.The connected packages formed one large group of 30 and three smaller groups.Source: Links between the packages · p. 36

Discussion

• The 58 reviewed packages cover movement-ecology workflow stages, but collective-motion analysis beyond descriptive metrics is poorly covered.The review highlights the abundance of tools and prioritizes improving existing packages’ visibility and accessibility over developing new ones.Source: Discussion · pp. 36–37

Integration over proliferation

• Independent proliferation makes package functionality harder to overview, and developers are urged to create new packages only for complementary contributions.The authors expect the number of packages to continue increasing, with greater possibility of redundancy and disconnection.Source: Integration over proliferation · pp. 37–38

Recommendations

• Researchers are advised to favor packages that are well documented, actively maintained, and widely used, while selecting methods they understand and reporting bugs.The authors associate maintenance with the likelihood of errors being fixed and wider use with opportunities to identify bugs and obtain guidance.Source: Recommendations · p. 39
• Developers should assess whether a function fills a gap or belongs in an existing package, and consider integrating new functions with maintained packages.The authors also encourage existing package authors to be open to integration and new collaborators.Source: Recommendations · p. 40
• Package developers are encouraged to support compatibility through data-class coercion and to discuss a shared trajectory class.At the time of the research, 40 of 58 packages used sp classes, while only crawl was compatible with sf.Source: Recommendations · pp. 40–41
• Documentation should explain functions and methods, provide references, and use examples or vignettes to help researchers navigate packages.The authors say method descriptions and references are especially important when a package implements a new analysis method.Source: Recommendations · p. 41
• Developers should identify people responsible for maintaining a package over time.The authors suggest lab PIs or working-group members as possible maintainers.Source: Recommendations · p. 41

• The review serves as a road map of tools implemented by packages for movement-ecology data analysis.The authors report that programmers had developed at least 58 R tracking packages.Source: Conclusions · p. 41
• The authors identify documentation as fundamental to improving the accessibility and understanding of existing packages.Source: Conclusions · p. 41
• The authors conclude that improved accessibility and more integrated package development will help advance research in the field.Source: Conclusions · p. 42

• RJ conceived the review, reviewed the R packages, led the survey and its analysis, and led manuscript writing.Source: Authors’ contributions · p. 43
• RJ and MB developed the manuscript workflow; RJ and MEB gathered package information, and MEB and MB implemented the survey online.RJ, MEB, and MB also worked on the Zenodo repository.Source: Authors’ contributions · p. 43
• All authors critically contributed to manuscript drafts and reviewer responses and gave final approval for publication.Source: Authors’ contributions · p. 43

Table 1

• The table summarizes tracking-package functionality across data-processing and analysis stages.Its caption identifies the table as a functionality summary; the paper reports 58 packages in the review.Source: Table 1

Data processing

• Data-preparation entries include GLS processing, location cleaning, duplicate removal, interpolation, and track compression.The listed methods include threshold, twilight-free, and triangulation approaches for GLS data.Source: Table captions · p. 70

Data analysis

• Analysis entries cover visualization, trajectory description and similarity, behavioral models, and space-use estimation.Listed methods include track animations, hidden Markov models, state-space models, and utilization distributions.Source: Table captions · pp. 70–71, Path reconstruction · pp. 71–72

Other functions

• The inventory also includes specialized tools for interactions, movement simulation, and database-linked workflows.Examples listed include dyad interaction metrics, movement-model simulation, and PostgreSQL/PostGIS support.Source: Path reconstruction · p. 72

Figure 1

• Workflow counts in parentheses indicate packages per category, and packages may appear in multiple categories except visualization.Figure 1 presents the movement-ecology data-processing and analysis workflow.Source: Figure captions · p. 73, Figure 1 · p. 74

Figure 2

• Packages were reviewed between March and August 2018, but 2018 was incomplete and therefore was not included in the graph of package counts by publication year.Source: Figure captions · p. 73, Figure 2 · p. 75

Figure 3

• Documentation colors distinguish standard documentation, vignettes, and packages with peer-reviewed articles; only packages with at least 10 respondents are shown.Green denotes standard documentation only, blue denotes vignettes, and purple denotes peer-reviewed articles as well.Source: Figure captions · p. 73, Figure 3 · p. 76

Figure 4

• Network arrows point to suggested or depended-on packages, and circle size reflects how many packages link to each package.Dashed arrows indicate suggestions and solid arrows indicate dependencies; bold type marks active packages.Source: Figure captions · p. 73, Figure 4 · p. 77