Changes in version 4.5.0 (2026-09-17)

• Functionality change - OSCARS Optimization: Fitting non-smooth distance functions (oneStep, triangle, and huber) is now accomplished by OSCARS::oscars. OSCARS is a true global maximizer and hence cannot absolutely guarantee that the global maximum has been found. However, OSCARS now starts at parameters reasonably close to the likelihood’s maximum and performs a maximum of 10,000 iterations (the default, see options("Rdistance_oscarEvals")). While slower than other methods, this method usually finds a higher maximum of the likelihood than other methods. Due to the significant slow down using OSCARS, estimated run times are reported on the command line.
• Functionality change - Transect Survey Design Routines: Added a family of functions for designing random line-transect surveys inside study-area polygons. Routine findSpacing() computes transect spacing that yields a target survey length. Routine makeLines() places transects with a random start. calcLineLength() converts a target number of detected groups into a target transect length. The main function called by users, drawTransects(), is a wrapper that calls findSpacing() followed by makeLines(). Both parallel (“rectangular”, or “back and forth”) and “zigzag” transects are implemented. Transects are applied over one or several polygons, with an option to generate random replicates (parameter R). Transects can be returned as one continuous route (combine = TRUE) or as individual legs (combine = FALSE), and targetLength can refer to the total (target = "total") or on-effort length (target = "onEffort"). Computation of the spacing is an optimization problem performed by OSCARS::oscars. Some notes:
  • “zigzag” routes: Constructed by placing a series of line segments perpendicular to a baseline that are regularly spaced with a random start. Transect “legs” run from polygon edge to polygon edge and cross the baseline at the points separated by the specified spacing. Locations where “legs” meet on the polygon’s boundary are the transect’s “pivots”.
  • Spacing: Spacing is the distance between adjacent crossings of the baseline under both the rectangular and zigzag layouts.
  • Combine: Parameter combine controls clipping and shape of the returned object. combine = TRUE returns the full route, unclipped, including on-transect and off-transect transit segments. combine = FALSE returns one row per “leg”, clipped to the polygon. The returned lengths sum to on-effort length. Legs that are broken by a concavity is divided into pieces and is returned as a MULTILINESTRING geometry.
  • Zigzag baselines: Unless the user supplies a baseline, zigzag transect baselines are straight. When no baseline is supplied, the polygon’s general centerline is approximated by the midpoints of polygon cords. This initial baseline is then straightened by regressing vertical coordinates onto horizontal coordinates and translating the estimated line to pass through the polygon’s centroid.
    
  • Convex Partitioning: Routine convexPartition splits a strongly concave polygon into a number of (hopefully) more-convex pieces using Approximate Convex Decomposition (Lien & Amato 2006). Splitting improves the coverage of zigzag transects on bent, arc-, or L-shaped polygons. Breaking polygons into smaller more-convex pieces is not necessary for rectangular transects. Parameter nPieces is either an integer, in which case exactly that many pieces are returned, or "optimum" (the default), in which case the count is chosen automatically. method = "fast" (the default) cuts greedily at the most concave vertex. method = "optimum" searches the cut vertices with OSCARS::oscars to maximize the minimum solidity of the pieces.
• New data sets:
  • exampleSurveyPoly: two real-world concave survey strata in Alaska (in Cook Inlet) used to demonstrate the survey-design functions.
  • pronghornDf: Data from an aerial line-transect survey for pronghorn (Antilocapra americana) collected by the Wyoming Game and Fish Department in southeast Wyoming, 2012-2019.
  • pronghornAreas: The study-area (herd-unit) polygons associated with the pronghorn line-transect data.
• New dependencies: Added sf, grDevices, and OSCARS to Imports, required by the new survey-design functions.
• Functionality change: abundEstim() now accepts a previously fitted abundance object in addition to a distance function. This lets users add bootstrap iterations to an existing fit, for example by fitting with ci = NULL first and running (or extending) the bootstrap later.
• Documentation Updates: Many documentation updates to (hopefully) clarify the routines and methods.

Changes in version 4.4.3 (2026-05-13)

• Bug Fix: Bug fix in ESW when likelihood is oneStep and expansions > 0. Before: integrateOneStepNumeric returned areas without units, which was okay for optimization, but not okay for ESW. integrateOneStepNumeric now returns a vector with the correct units and is consistent with the other integrate??? functions.

Changes in version 4.4.2 (2026-05-08)

• Bug Fix: ‘outArea’ was undefined when likelihood returned NA or Inf or NaN and this stopped execution. ‘outArea’ is now always defined. This bug only effected the non-smooth likelihoods ‘oneStep’, ‘triangle’, and ‘huber’.
• Bug Fix: Fixed failing bootstraps. Base function ‘rbind’ mysteriously began stripping attributes from its arguments. Rdistance relies on these attributes for bootstrapping. Attributes restored now and bootstraps run.
• Bug Fix: Fixed the width vector for smooth likelihoods that was too small when expansions were present. Ill-sized W vector caused failure when expansions > 0 and likelihood was halfnormal, hazrate, negexp, or Gamma.
• Documentation Updates: Updated and clarified sections of the documentation of abundEstim and RdistanceControls.

Changes in version 4.4.0 (2026-04-22)

• Functionality change: Added triangle distance function, a mixture of triangle and uniform distributions. See help(triangle.like) examples.
• Functionality change: Added huber distance function, a mixture of an inverted version of Huber loss and a uniform distribution. See help(huber.like) examples.
• Functionality change: Changed bounds on the oneStep Theta parameter to the minimum and maximum of distances, thus ensuring the ledge lies within the range of observed distances.
• Functionality change: Implemented use of stats::optim for optimization. Prior to this, the only non-gradient optimizer was HookeJeeves. Default optimizer for oneStep, triangle, and huber is now the Nelder-Mead implementation in optim.
• Update: Minor documentation updates and clarifications.
• Note: No bugs found that effect results produced by prior versions.

Changes in version 4.3.1 (2026-02-10)

• Version v4.3.1 contains a critical bug fix. The prior version, 4.3.0, incorrectly reported ESW when scaling factors (g.x.scl) were less than 1.0. Versions <=4.1.0 reported correct ESW, and ESW reported by 4.3.0 was correct when g.x.scl equaled 1.0 (the default).
• Default value for w.lo changed in parseModel to suppress warning re x.scl < w.lo.

Changes in version 4.3.0 (2026-01-06)

• Functionality change: Added parallel bootstrap processing for confidence interval estimation. Bootstrap resampling to compute confidence intervals can take considerable time and running in parallel, on multiple CPU cores, speeds things up. Parallel processing is handled by multidplyr. Default is to run bootstraps in parallel on n-1 CPU cores, where n is the number of cores available on the local machine . No progress bar is produced during parallel processing. Disable parallel processing by setting parallel = FALSE in call to abundEstim. Run on a specific number of cores by setting parallel to that number (e.g., parallel = 3 uses three cores).
• Functionality change: Added the Gamma likelihood back to the suite of likelihoods available in Rdistance. The Gamma was removed during the big re-vamp of version 4.0.3. The Gamma likelihood works for both lines and points. It works with both covariates and expansions.
• Update: Substantial documentation updates and clarifications.

Changes in version 4.1.0 (2025-11-29)

• Functionality change: Added oneStep.like, a mixture of non-overlapping uniform densities, as a distance function. Included associated print, plot, summary, and expansion methods. This required inclusion of a new non-gradient based optimizer.
• Functionality change: Implemented reporting of bootstrap coefficient standard errors, after bootstrap resampling is complete. Prior to bootstrapping, asymptotic se’s are reported when known.
• Functionality change: Included unit assignment helpers. Units can now be assigned with the %#% operator (e.g., 3 %#% “m”), which makes unit assignment easier than in prior versions (which used units::set_units). Fixed unit assignment operators are included for all popular linear and squared units (e.g., 3 %m%. assigns meters to 3). See help(unitHelpers).
• Functionality change: Included b-spline expansion factors for distance functions. Expansion factors are now cosine, hermite, simple, and bspline.
• Functionality change: Added verbosity option. This prints intermediate parameter estimates and likelihoods during maximization. Set options(Rdistance_verbocity = 1) or higher to see progressively more detailed intermediate output.
• Update: Likelihood computation time was significantly decreased relative to the prior version, greatly increasing speed of maximizations. Now, all likelihoods for both lines and points that do not contain expansions use exact integration, which is much quicker than numerical integration. Numeric integration is used only when expansion terms are included.
• Update: Substantial documentation updates and clarifications.
• Bug Fixes: . Fixed bug in plot.dfunc.para when w.lo > 0 . Fixed bug in point transect methods resulting in incorrect likelihood scalings . Fixed expansions hermite and simple that was causing non-convergence issues

Changes in version 4.0.5 (2025-04-10)

Bug fixes:

• Fixed bug in predict method when type = "density" causing NaN estimates on transects with observations outside the strip.
• Fixed bug in RdistDf when merge parameter by was named. When by was named and merge was on different named variables, names and values in by were reversed prior to the fix due to first nesting then merging. This bug did not affect merges on same-named variables.
• Non-bug change: Moved all the examples from README to tutorials on the McDonald Data Science website because we have more control, they are indexed by date and keyword, and they look better.

Changes in version 4.0.3 (2025-03-28)

Methods and workflow in Rdistance versions >4.0.0 are substantially different from prior versions.

• BIGGEST CHANGE: Input data frames are now nested tibbles with one row per transect, and detection information in a data frame contained within a list column. Technically, the new data frame format is a grouped tibble with one row per group, and each group is a transect. Survey type, observer system, and name of the effort column are recorded as attributes in the new data frame.
  • RdistDf constructs the new Rdistance data frames from separate transect and detection data frames formatted for use in prior versions. Use this function on old sets of site and detection data frames to construct the new nested data frames. See examples in ?RdistDf.
  • is.RdistDf checks the validity of the new Rdistance data frames.
    
  • summary prints a summary of number of transects, number of groups seen, number of individuals, etc.
• Temporary Functionality Removal: Versions >4.0.0 are a near complete re-write of versions <3.1.3. I felt the re-write was necessary to improve stability of the code and lay the ground work for future functionality improvements. But, to keep from going insane, I am rolling out new versions that build on one another, with core functions first (in 4.0.* versions). As such, I have temporarily removed the following functionality from versions with numbers 4.0.*:
  • Logistic distance function form.
  • Smoothed distance functions.
  • Gamma distribution distance functions.
  • Double-observer methods.
  • All vignettes.
• print, summary, and plot methods have been improved.
• Documentation updated and improved.
• Rdistance options can now be assessed by the regular options function. All Rdistance options are prefixed with ‘Rdistance_’ to distinguish them from other options.
• New data frame format is pipe-able through the estimation workflow. For example, the basic workflow is df |> dfuncEstim(dist~1) |> abundEstim(), which will estimate a distance function and density in one go.

Changes in version 3.1.3 (2025-03-11)

Version 3.1.3 contains three patches. Several updates to documentation and one bug fix precipitated by changes to the ‘units’ package.

Changes in version 3.1.0 (2023-08-28)

Version 3.1.0 primarily addresses GitHub issues.

• Summary Methods Implemented: Issue #57. Rdistance did not implement summary methods for dfunc and abund objects prior to version v3.1.0. This version implements both summary and print methods for the main outputs. print.dfunc and print.abund are modeled on print methods for lm and glm objects. New methods summary.dfunc and summary.abund are modeled on the summary methods for lm and glm and will produce fuller (relative to print) outputs.
• Documentation fixes:
  • Added reference to general method in DESCRIPTION as suggested by CRAN editors.
  • Updated and clarified Description field of DESCRIPTION file.
  • Updated numerous links in README and NEWS.
• Bug Fixes:
  • Fixed bug when abundEstim was called with tibbles.

Changes in version 3.0.0 (2023-06-12)

Version 3.0.0 is a substantial change and upgrade.

• BIGGEST CHANGE: Measurement units are now required on all distances (e.g., off-transect distances, strip widths, study area size, etc.), and it is possible to specify output units. Unit conversion makes use of the units package, is internal, and automatic.
• One-sided Transects: Added the ability to analyze single-sided transects by setting singleSided = TRUE in call to abundEstim.
• User Requested: Added computation of bootstrap confidence intervals for effective sampling distances (ESW and EDR).
• NA Length Transects: NA length transects are now allowed. Observations on NA length transects are used to estimate distance functions, but are dropped when estimating density. This allows, for example, off-transect distance observations from one study area or year to estimate a distance function which is then applied to the observations from another study area or year.
• Added calculation of target density (number per unit area) to abundEstim, and it is included in bootstrapping.
• Allowed expansion terms and covariates in the same distance function.
• Renamed uniform likelihood to logistic. Uniform is now deprecated.
• Added covariates to Gamma likelihood estimation.
• Change in default parameter: Changed default value of area in abundEstim from 1 to NULL. NULL now translates to 1 square output unit.
  
• Added color to print methods. Results are in green, if the R terminal allows it (i.e., in Rstudio, but not R gui). Errors are in red.
• Reformatted default output to include target density and abundance when printing an abund object.
• Fixed bug in F.gx.estim that occasionally popped up when sighting function was monotonically decreasing.
• Updated starting values for faster estimation.
• Decreased lower limit of negexp likelihood parameter to achieve more valid fits.
• Warning messages now report which parameters are at their boundary, and which boundary (lower or upper).
• Bug fixes:
  • Fixed issue with contrasts in model.matrix
  • Fixed issue of no observations during bootstrapping
  • Fixed issues with changing scale locations (i.e., x.scl and g.x.sxl)
  • Fixed ESW for w.lo > 0
  • Restricted likelihood values to be positive
• Substantial documentation updates
• Substantial testing facilities added.

Changes in version 2.1.5 (2020-06-17)

• Fixed bug when bootstrap iteration selects zero targets
• Many updates to vignettes
• Updates to testthat files
• Added descriptive error for case when bySite==T and detection level covariates are present (Issue #72)

Changes in version 2.1.4 (2019-06-11)

• Fixed bug caused by empty factor levels
• Included automated testing (testthat)

Changes in version 2.1.3 (2019-01-02)

• Fixed scaling error in the Gamma likelihood causing mode to be less than 1.0
• Fixed inverted plots that occurred when w.lo > 0 (left-truncation)
• Fixed bug in ‘autoDistSamp’ when ‘bySite=TRUE’
• Fixed bug causing bootstrap failure of point transect surveys when ‘siteData’ contained only one column

Changes in version 2.1.2 (2018-08-23)

• Included methods for point transects. Previous version (v1.3.2) allowed only line transects.
• Implemented covariates in the detection functions.
• Implemented non-parametric (smoothed) distance functions
• Enhanced plotting abilities
• Enhanced vignettes
• Many bug fixes