NEWS

---

aniMotum 1.3-0

• adds a joint (hierarchical) move persistence SSM via fit_ssm(model = "jmp"), fitting the time-varying move persistence model to all tracks at once (Jonsen et al. 2019). By default, sigma_g and the measurement error parameters are pooled, the process variance is hierarchical, and rho_p is estimated separately for each individual
• adds a rotation-invariant joint correlated random walk SSM via fit_ssm(model = "jcrw"). Movement scale and anisotropy are shared among individuals, while each individual keeps its own orientation. Anisotropy is set via share_control(aniso = )
• adds share_control() to specify which parameters are pooled, hierarchical or individual-specific in joint fits. Combinations that are not identifiable are refused
• adds sim_jcrw() to simulate tracks from the jcrw process model with known parameters
• joint fits have class jssm_df. summary() and print() report convergence and AICc once for the joint fit, and separate parameter estimates shared by all individuals from those estimated per individual
• adds a projection argument to fit_ssm(). The default "mercator" is unchanged, "auto" chooses a conformal projection suited to the data extent via the new auto_crs(), or a proj4 string can be supplied. The projection is chosen once for all individuals
• fit_ssm() now always respects the projection of an input sf object and never re-projects it
• adds a Projections vignette, describing how the choice of projection affects parameter estimates
• adds smru_haulout() to flag haulout periods (ho = 1) in tracking data, using an SMRU haulout data file
• adds haulout, ho.ref and ho.id_fun arguments to fit_ssm(), calling smru_haulout() before fitting. An existing ho column in the input data is also used
• the rw, crw and mp models now keep locations nearly stationary during haulouts, by scaling the process variance with ssm_control(ho_scale = ) (default 0.01)
• adds ssm_control(gap.thresh = ) to flag data gaps longer than gap.thresh hours. The crw and mp models break directional persistence across flagged gaps, which stops implausible looping by crw tracks across large gaps. No gaps are flagged by default
• route_path() no longer moves locations known to be on land: haulout locations (ho = 1), locations within intervals supplied via the new onland argument, and on-land locations at track ends. All locations are returned, with on_land = TRUE where kept on land
• sim_fit() now simulates tracks in the projection used to fit the SSM, rather than assuming Mercator. This fixes simulations from fits to projected sf data and from tracks that cross -180,180 (#76)
• sim_post() now samples more efficiently, using far less memory for long tracks (Jan-Ole Koslik, #85), supports joint fits, and no longer assumes a Mercator projection
• fixes issue with osar() where residuals were assigned to the wrong individuals when they could not be calculated for some individuals (#91)
• fixes issue with fit_mpm(model = "jmpm") where tracks with numeric ids and unequal lengths resulted in an error. Also fixes fit_mpm() dropping id and date when coordinates are named x, y, and grab() failing on crashed jmpm fits (#90)
• fixes issue with fit_ssm(model = "mp") where tau was estimated for "GL" locations (#82)
• fixes issue with fit_ssm() where all columns but geometry were dropped from sf input data when spdf = FALSE
• fixes issue with ellp.par() where simulated location errors did not match their error ellipses
• fixes issue with sim_fit() where simulations used fitted locations when what = "predicted" was specified, and rw tracks did not return to their start when cpf = TRUE
• removes ggplot2 warnings about empty axis labels in plot functions

aniMotum 1.2-06

• various minor issues fixed including: build against TMB 1.9.11, fixes error when grad specified in sim_fit().

aniMotum 1.2

• addition of sim_post(), a function for posterior simulations from SSM fits, conditional on data and movement parameters
• addition of plot method for sim_post objects
• sim_filter() can now use arbitrary variables, including user-appended environmental variables, for filtering tracks simulated with sim_fit()
• sim_fit() now simulates tracks from user-specified start and end locations that differ from the estimated track start and end.
• fit_ssm() now handles "generic location" data provided the locations have x and y standard errors. These data can be light-level geolocations, acoustic telemetry positions, or other location data. Input data should have lc = "GL" for all generic locations.
• a change to the min.dt argument in fit_ssm(), the default is now min.dt = 0 (no minimum time interval between observations, but any subsequent observations that occur at the same time are ignored when fitting an SSM).
• fixed an issue with high temporal resolution data, where prediction times exactly match observation times when observations occur 1 s apart. This caused an error when fitting SSM's.

aniMotum 1.1-06

• fixes issue with route_path() where simulated tracks that are entirely on land resulted in an error

aniMotum 1.1-04

• fixes issue with grab() where multiple data sets with lon modulo 0,360 resulted in an error
• fixes issue with route_path() where rerouting tracks that have no locations on land resulted in an error. In these cases, route_path() now returns a tibble identical to that supplied (fitted or predicted locations) and issues a message on the console.

aniMotum 1.1-02

• update citations to include new aniMotum R package paper in Methods in Ecology and Evolution as the primary reference.

aniMotum 1.1-01

• fixes issue with map() where function could become paused one some calls.

aniMotum 1.1

• package name change to coincide with Methods in Ecology and Evolution manuscript: Jonsen et al. aniMotum, an R package for animal movement data: rapid quality control, behavioural estimation and simulation. Accepted 06/12/2022.

• adds format_data() to pre-process non-default data formats into that expected by fit_ssm
• adds greater flexibility for input data formats (via new arguments to fit_ssm), related to format_data
• fixes an issue with input data as an sf-tibble or sf-data.frame that caused an error when fitting SSM's via fit_ssm
• when installing package from source, C++ code is automatically compiled against the existing TMB package version. This should eliminate the warning message on installation and package load that some users may have experienced previously
• updated Overview vignette, describing input data structures in greater detail

aniMotum 1.0-5

• adds route_path() a wrapper function calling pathroutr to re-route fit_ssm estimated or simfit simulated tracks around land barriers
• adds a move persistence model via fit_ssm(model = "mp") to allow simultaneous estimation of locations and move persistence. This approach may be preferable to using fit_mpm() on a fit_ssm model object. fit_mpm() is retained for less error-prone (GPS) location data
• adds option for move persistence estimates to be normalised to 0,1 in grab(), either on tracks separately or as a group for a relative measure that spans 0 - 1.
• adds sim_filter() to calculate similarity between simulated and ssm-estimated tracks, and returns the most similar simulated tracks based on a user-specified quantile
• adds mapping function map() to replace fmap() for faster, more flexible estimated track maps & fixes to coastline and other mapping issues for tracks that cross -180,180
• adds rosm map tile layers to map() via ggspatial::annotation_map_tile for more detailed coastlines on large-scale maps
• adds faster crw model fitting via fit_ssm() by turning off travel rate standard error (s.se) estimation in ssm_control() as the default. SE estimation can be turned on via control = ssm_control(se = TRUE).
• adds a summary function for displaying information about SSM fits.
• replaces hcl.colors("Zissou1") palette for most plots - Zissou1 was fun but not colour blind-friendly; in most cases "Cividis" is now the default but users can specify any hcl.pals() palette, using the pal argument in many of the plot functions.
• fit object s3 classes fG_ssm, fG_mpm migrated to ssm_df, mpm_df
• generic plot method s3 classes migrated to plot.ssm_df, plot.mpm_df, plot.osar, plot.sim, plot.simfit
• adds new vignettes for more comprehensive documentation of package features

aniMotum 0.7-6

• adds fixes to example data for testing/examples so pkg builds on solaris and various linux boxes with older GDAL/PROJ libraries

aniMotum 0.7-5

• adds sim() to simulate animal tracks using the rw, crw or mpm process models. The rw and crw models can also be specified with state-switching between multiple behavioural states. Tracks can be simulated with or without Argos (LS or KF) errors, as time-regular or time-irregular
• adds simfit() to simulate animal tracks from fit_ssm fit objects
• adds generic plot() methods for sim and simfit objects
• adds fit_mpm can take a fit_ssm object directly as input, removing need for user to manipulate data prior to calling fit_mpm
• adds fit_mpm can fit to SSM-predicted (time regular) or SSM-fitted (time irregular) locations, via what argument
• adds fit_mpm can fit to lon,lat or x,y coordinates, via coords argument
• adds ssm_control() for centralized control over optimizer and optimization method choices, optimizer parameters, and foieGras model parameter bounds
• adds faster and more stable optimization
• adds time-series and acf() plots as option when visualising prediction residuals calculated from osar()
• removes hist plots as option when visualising prediction residuals calculated from osar()
• removes (with deprecation errors) verbose, optim, optMeth, and lpsi arguments to fit_ssm
• replaces wesanderson::wes_palette("Zissou1") with hcl.colors("Zissou1") and provides arguments to change this default palette

aniMotum 0.6-9

• adds wesanderson::wes_palette("Zissou1") as default palette for plots/maps
• generic plot method for fG_ssm objects can now plot individuals all on 1 page (pages = 1) or on separate pages (pages = 0)
• 2-d (track) plots now include confidence ellipses on estimated locations
• generic plot method for fG_mpm objects now available
• mapping function (fmap) can optionally take an fG_mpm object to colour locations by behavioural index (gamma_t)

aniMotum 0.6-7

• adds ability to fit move persistence models to temporally regular OR irregular location data
• adds ability to turn off predicted locations by setting time.step=NA causing locations to be estimated only at observation times
• adds ability to fit to Argos - GPS or Argos - GLS locations simultaneously when combined in a single input data.frame
• adds ability to specify an alternate emf (Error Multiplication Factor) data.frame for Least-Squares and/or GPS locations
• adds human-readable/understandable errors and warnings with (occasional) advice
• turns off estimation of psi parameter (ellipse semi-minor axis re-scaling factor) when rw process model is fit to Argos KF/KS data
• removes reliance on dplyr::do, which is superseded as of dplyr 1.0.0
• replaces dplyr progressbar with parameter trace when fitting SSM to single or multiple data sets in default verbose mode 1
• replaces argosfilter::sdafilter in favour of trip::sda (which is a faster, vectorized version of the former) to prefilter outlier locations

aniMotum 0.4.0

• adds move persistence models for behavioural inference along animal tracks, via fit_mpm()
• adds ability to fit SSM's to processed light-level geolocation or GPS data by adding lonerr, laterr variables to input data
• adds One-Step-Ahead prediction residuals for evaluating SSM fits, via osar() and generic plot() method for osar output
• adds diagnostic plot functions that handle multi-individual fits, via generic plot() method for fG_ssm fit objects
• adds sf-enabled mapping function, via fmap()
• fixes CRAN check errors from last version, where proj4string syntax +init:epsg= was not supported in non-PROJ4 emulation mode on some linux platforms

aniMotum 0.2.2

• patch for compatibility with latest sf version

aniMotum 0.2.1

• patch to improve C++ portability

aniMotum 0.2.0

• first release, updates will follow regularly and be documented here