| Title: | Fit Continuous-Time State-Space and Latent Variable Models for Quality Control of Argos Satellite (and Other) Telemetry Data and for Estimating Changes in Animal Movement |
|---|---|
| Description: | Fits continuous-time random walk, correlated random walk and move persistence state-space models for location estimation and behavioural inference from animal tracking data ('Argos', processed light-level 'geolocation', 'GPS'). Template Model Builder ('TMB') is used for fast random-effects estimation. The 'Argos' data can be: (older) least squares-based locations; (newer) Kalman filter-based locations with error ellipse information; or a mixture of both. The models estimate two sets of location states corresponding to: 1) each observation, which are (usually) irregularly timed; and 2) user-specified time intervals (regular or irregular). A track re-routing function is provided to adjust location estimates for known movement barriers. Track simulation functions are provided. Latent variable models are also provided to estimate move persistence from track data not requiring state-space model filtering. |
| Authors: | Ian Jonsen [aut, cre, cph], W. James Grecian [aut, ctb], Toby Patterson [aut, ctb] |
| Maintainer: | Ian Jonsen <[email protected]> |
| License: | MIT + file LICENSE |
| Version: | 1.2-07 |
| Built: | 2024-11-07 14:40:54 UTC |
| Source: | https://github.com/ianjonsen/aniMotum |
fit Continuous-Time Random Walk and Correlated Random Walk state-space models to filter Argos Least Squares or Kalman Filter location data
Maintainer: Ian Jonsen [email protected] ORCID
Contributor: Toby Patterson ORCID
Contributor: W. James Grecian ORCID
Jonsen ID, Grecian WJ, Phillips L, et al. (2023) aniMotum, an R package for animal movement data: rapid quality control, behavioural estimation and simulation. Methods in Ecology and Evolution 14:806-816.
Jonsen ID, Patterson TA, Costa DP, et al. (2020) A continuous-time state-space model for rapid quality-control of Argos locations from animal-borne tags. Movement Ecology 8:31.
Jonsen ID, McMahon CR, Patterson TA, et al. (2019) Movement responses to environment: fast inference of variation among southern elephant seals with a mixed effects model. Ecology. 100(1):e02566.
Useful Links:
Report bugs/issues at https://github.com/ianjonsen/aniMotum/issues/
set aesthetics as a named list for: 1) mapping i) estimated locations; ii) estimated confidence ellipses; iii) estimated track lines; iv) observed locations; v) land masses; vi) water bodies; and 2) colour palettes for i) behavioural indices or ii) individual animal tracks
aes_lst( est = TRUE, conf = TRUE, line = FALSE, mp = TRUE, obs = FALSE, shape = c(19, NA, NA, 17, NA, NA), size = c(1.25, NA, 0.2, 0.8, NA, NA), col = c("dodgerblue", NA, "grey50", "orange", NA, NA), fill = c("dodgerblue", "dodgerblue", NA, "orange", "grey60", "grey85"), alpha = c(1, 0.4, 1, 1, 1, NA), mp_pal = hcl.colors(100, palette = "Plasma", rev = FALSE), id_pal = "Harmonic", date_pal = hcl.colors(100, palette = "Viridis", rev = FALSE) )aes_lst( est = TRUE, conf = TRUE, line = FALSE, mp = TRUE, obs = FALSE, shape = c(19, NA, NA, 17, NA, NA), size = c(1.25, NA, 0.2, 0.8, NA, NA), col = c("dodgerblue", NA, "grey50", "orange", NA, NA), fill = c("dodgerblue", "dodgerblue", NA, "orange", "grey60", "grey85"), alpha = c(1, 0.4, 1, 1, 1, NA), mp_pal = hcl.colors(100, palette = "Plasma", rev = FALSE), id_pal = "Harmonic", date_pal = hcl.colors(100, palette = "Viridis", rev = FALSE) )
est |
logical; turns on estimated locations (default = TRUE) |
conf |
logical; turns on estimated confidence ellipses. Default varies depending on whether behavioural index is being mapped &/or if single versus multiple tracks are being mapped. |
line |
logical; turns on estimated track line(s) (default varies) |
mp |
logical; turns on move persistence index (default = TRUE, if present in model fit) |
obs |
logical; turns on observed locations (default = FALSE) |
shape |
gpplot2 shape value (integer: 0, 25) for estimated & observed locations |
size |
ggplot2 size value for estimated locations & track lines, and observed locations |
col |
colour for estimated locations and track lines, and observed locations |
fill |
fill colour for estimated locations & confidence ellipses, observed locations, land polygons, and water |
alpha |
transparency for specified fills/colours |
mp_pal |
continuous colour palette for move persistence values |
id_pal |
discrete colour palette for track id's |
date_pal |
continuous colour palette for displaying date along track |
a named list, with 9 elements, of map components and aesthetics
elements 1-5 - map components: est, conf, line, mp, obs
element 6: a data.frame named df containing ggplot2 aesthetics: shape,
size, col, fill, and alpha. df has 6 rows, corresponding to the map
features: estimated locations, confidence ellipses, track lines, observed
locations, land polygons, and water
element 7: mp_pal
element 8: id_pal
element 9: date_pal
# generate custom aes list aes <- aes_lst(conf = FALSE, mp_pal = hcl.colors(n=100, palette = "RdBu")) # modify aesthetics aes$df$size[1] <- 1.5 aes$df$fill[6] <- grey(0.9)# generate custom aes list aes <- aes_lst(conf = FALSE, mp_pal = hcl.colors(n=100, palette = "RdBu")) # modify aesthetics aes$df$size[1] <- 1.5 aes$df$fill[6] <- grey(0.9)
Example elephant seal Argos tracking data. Data were sourced from the Integrated Marine Observing System (IMOS) Sourced from the Australian Integrated Marine Observing System (IMOS) deployments at Davis Research Station, Antarctica and are publicly available (http:// imos.aodn.org.au). IMOS is supported by the Australian Government through the National Collaborative Research Infrastructure Strategy and the Super Science Initiative.
.RData
emf
emf( gps = 0.1, emf.x = c(1, 1.54, 3.72, 13.51, 23.9, 44.22), emf.y = c(1, 1.29, 2.55, 14.99, 22, 32.53) )emf( gps = 0.1, emf.x = c(1, 1.54, 3.72, 13.51, 23.9, 44.22), emf.y = c(1, 1.29, 2.55, 14.99, 22, 32.53) )
gps |
error multiplication factor(s) for GPS locations, can be a scalar (x = y) or vector of length 2 (x != y) |
emf.x |
error multiplication factors for Argos longitude classes 3, 2, 1, 0, A, B (Z assumed equal to B) |
emf.y |
error multiplication factors for Argos latitude classes 3, 2, 1, 0, A, B (Z assumed equal to B) |
Error Multiplication Factors for Argos (and GPS) locations. Default assumption is that GPS locations are 10x more accurate than Argos lc 3 in both x and y directions.
User-specified Error Multiplication Factors (emf). emf's must be provided as a data.frame with the following columns:
emf.x emf values for the x direction
emf.y emf values for y direction
lc location class designations
The location class designations can be the standard Argos lc values: 3, 2, 1, 0, A, B, Z or other values.
The number of classes specified is flexible though may not be amenable to a large number of classes.
Whatever class designations are chosen must also appear in the input data lc column.
A GPS location class ("G") is provided by default and assumes that GPS locations are 10 x more precise than Argos lc 3 locations.
mpm)fit a random walk with time-varying move persistence to temporally regular or irregular location data
fit_mpm( x, what = "predicted", model = c("jmpm", "mpm"), coords = 3:4, control = mpm_control(), inner.control = NULL, optim = NULL, optMeth = NULL, verbose = NULL )fit_mpm( x, what = "predicted", model = c("jmpm", "mpm"), coords = 3:4, control = mpm_control(), inner.control = NULL, optim = NULL, optMeth = NULL, verbose = NULL )
x |
a |
what |
if a |
model |
mpm model to fit; either |
coords |
column numbers of the location coordinates (default = 3:4) |
control |
list of control settings for the outer optimizer (see mpm_control for details) |
inner.control |
list of control parameters for the inner optimization |
optim |
is deprecated, use ssm_control(optim = "optim") instead, see ssm_control for details |
optMeth |
is deprecated, use ssm_control(method = "L-BFGS-B") instead, see ssm_control for details |
verbose |
is deprecated, use ssm_control(verbose = 1) instead, see ssm_control for details |
a list with components
fitted a dataframe of fitted locations
par model parameter summary
data input data.frame
tmb the TMB object
opt the object returned by the optimizer
Jonsen ID, McMahon CR, Patterson TA, et al. (2019) Movement responses to environment: fast inference of variation among southern elephant seals with a mixed effects model. Ecology. 100(1):e02566
## fit jmpm to two southern elephant seal tracks xs <- fit_ssm(sese2, spdf=FALSE, model = "rw", time.step=72, control = ssm_control(verbose = 0)) fmpm <- fit_mpm(xs, model = "jmpm")## fit jmpm to two southern elephant seal tracks xs <- fit_ssm(sese2, spdf=FALSE, model = "rw", time.step=72, control = ssm_control(verbose = 0)) fmpm <- fit_mpm(xs, model = "jmpm")
fits: i) a simple random walk (rw) ii) a correlated random walk
(crw - a random walk on velocity), or iii) a time-varying move persistence
model (mp), all in continuous-time, to filter Argos LS, and/or KF/KS
location data, GPS data, and/or generic locations with associated standard
errors (e.g., processed light-level geolocation data, or high-resolution
acoustic telemetry data). Location data of different types can combined in a
single data frame (see details). Predicts locations at user-specified time
intervals (regular or irregular).
fit_ssm( x, vmax = 5, ang = c(15, 25), distlim = c(2500, 5000), spdf = TRUE, min.dt = 0, pf = FALSE, model = "crw", time.step = NA, emf = NULL, map = NULL, parameters = NULL, fit.to.subset = TRUE, control = ssm_control(), inner.control = NULL, ... )fit_ssm( x, vmax = 5, ang = c(15, 25), distlim = c(2500, 5000), spdf = TRUE, min.dt = 0, pf = FALSE, model = "crw", time.step = NA, emf = NULL, map = NULL, parameters = NULL, fit.to.subset = TRUE, control = ssm_control(), inner.control = NULL, ... )
x |
a |
vmax |
max travel rate (m/s) to identify implausible locations |
ang |
angles (deg) of implausible location "spikes" |
distlim |
lengths (m) of implausible location "spikes" |
spdf |
(logical) turn pre-filtering on (default; TRUE) or off |
min.dt |
minimum allowable time difference between observations;
|
pf |
just pre-filter the data, do not fit the SSM (default is FALSE) |
model |
fit a simple random walk ( |
time.step |
options: 1) the regular time interval, in hours, to predict to; 2) a vector of prediction times, possibly not regular, must be specified as a data.frame with id and POSIXt dates; 3) NA - turns off prediction and locations are only estimated at observation times. |
emf |
optionally supplied data.frame of error multiplication factors for Argos location quality classes. Default behaviour is to use the factors supplied by emf |
map |
a named list of parameters as factors that are to be fixed during
estimation, e.g., |
parameters |
a list of initial values for all model parameters and unobserved states, default is to let sfilter specify these. Only play with this if you know what you are doing |
fit.to.subset |
fit the SSM to the data subset determined by prefilter (default is TRUE) |
control |
list of control settings for the outer optimizer (see ssm_control for details) |
inner.control |
list of control settings for the inner optimizer (see TMB::MakeADFun for additional details) |
... |
variable name arguments passed to format_data, see format_data for details |
x is a data.frame, tibble, or sf-tibble with 5, 7 or 8
columns (the default format), depending on the tracking data type. Argos
Least-Squares and GPS data should have 5 columns in the following order:
id, date, lc, lon, lat. Where date can be a POSIX object or text
string in YYYY-MM-DD HH:MM:SS format. If a text string is supplied then the
time zone is assumed to be UTC. lc (location class) can include the
following values: 3, 2, 1, 0, A, B, Z, G, or GL. The latter two are for GPS
locations and 'Generic Locations', respectively. Class Z values are assumed
to have the same error variances as class B. By default, class G (GPS)
locations are assumed to have error variances 10x smaller than Argos class 3
variances, but unlike Argos error variances the GPS variances are the same for
longitude and latitude.
The format_data function can be used as a data pre-processing
step or called automatically within fit_ssm to restructure data that is
not in one of the above default formats. The minimum essential variables:
id, date, lc, lon, lat must exist in the input data but they can
have different names and exist in a different column order. See
format_data for details.
See emf for details on how to modify these assumptions.
Argos Kalman Filter (or Kalman Smoother) data should have 8 columns,
including the above 5 plus smaj, smin, eor that contain Argos error
ellipse variables (in m for smaj, smin and deg for eor).
Generic locations can be modelled provided each longitude and latitude
(or X and Y) coordinate has a corresponding standard error. These data should
have 7 columns, including the above 5 plus two extra columns, typically
named x.sd, y.sd that provide the standard errors for the longitude,
latitude (or X, Y) coordinates. Longitude and latitude standard errors should
be in degrees, whereas X and Y standard errors should be in m. In either case,
all lc values should be set to GL (Generic Location), the helper function
format_data will add the lc variable to the input data automatically.
Multiple location data types can be combined in a single data frame (see the Overview vignette for examples).
When data are provided as an sf-tibble, the user-specified projection is
respected, although projected units are always transformed to km to improve
SSM convergence efficiency. Otherwise, longlat data are re-projected
internally to a global Mercator grid and provided as the default output.
A simple tibble, without a geom, of lon,lat and x,y location estimates
can be obtained by using grab with the argument as_sf = FALSE.
a list with components
call the matched call
predicted an sf tbl of predicted location states
fitted an sf tbl of fitted locations
par model parameter summary
data an augmented sf tbl of the formatted input data
inits a list of initial values
pm the process model fit, either "rw" or "crw"
ts time time.step in h used
opt the object returned by the optimizer
tmb the TMB object
rep TMB sdreport
aic the calculated Akaike Information Criterion
time the processing time for sfilter
Jonsen ID, Patterson TA, Costa DP, et al. (2020) A continuous-time state-space model for rapid quality-control of Argos locations from animal-borne tags. Movement Ecology 8:31
Jonsen ID, McMahon CR, Patterson TA, et al. (2019) Movement responses to environment: fast inference of variation among southern elephant seals with a mixed effects model. Ecology. 100(1):e02566
## fit crw model to Argos LS data fit <- fit_ssm(ellie, vmax = 4, model = "crw", time.step = 24, control = ssm_control(verbose = 0)) ## time series plots of fitted values and observations plot(fit, what = "fitted", type = 1, ask = FALSE) ## 2-D tracks plots of predicted values and observations plot(fit, what = "predicted", type = 2, ask = FALSE)## fit crw model to Argos LS data fit <- fit_ssm(ellie, vmax = 4, model = "crw", time.step = 24, control = ssm_control(verbose = 0)) ## time series plots of fitted values and observations plot(fit, what = "fitted", type = 1, ask = FALSE) ## 2-D tracks plots of predicted values and observations plot(fit, what = "predicted", type = 2, ask = FALSE)
map aniMotum fitted or predicted locations, with or without Argos observations, optionally apply a different projection
fmap( x, y = NULL, what = c("fitted", "predicted", "rerouted"), conf = TRUE, obs = FALSE, obs.shp = 17, by.id = TRUE, by.date = TRUE, crs = NULL, ext.rng = c(0.05, 0.05), size = 0.25, col.ssm = "dodgerblue", col.obs = "black", lines = FALSE, landfill = grey(0.6), map_type = "default", pal = "Cividis", rev = FALSE, last_loc = NULL, ... )fmap( x, y = NULL, what = c("fitted", "predicted", "rerouted"), conf = TRUE, obs = FALSE, obs.shp = 17, by.id = TRUE, by.date = TRUE, crs = NULL, ext.rng = c(0.05, 0.05), size = 0.25, col.ssm = "dodgerblue", col.obs = "black", lines = FALSE, landfill = grey(0.6), map_type = "default", pal = "Cividis", rev = FALSE, last_loc = NULL, ... )
x |
a |
y |
optionally, a |
what |
specify which location estimates to map: fitted, predicted or rerouted |
conf |
include confidence regions around estimated location (logical; default = TRUE, unless y is an mpm fit object then conf is FALSE) |
obs |
include Argos observations on map (logical; default = FALSE) |
obs.shp |
point shape for observations (default = 17) |
by.id |
when mapping multiple tracks, should locations be coloured by id (logical; default = TRUE if nrow(x) > 1 else FALSE) |
by.date |
when mapping single tracks, should locations be coloured by date (logical; default = TRUE if nrow(x) == 1 else FALSE) |
crs |
|
ext.rng |
factors to extend the plot range in x and y dimensions (can exceed 1) |
size |
size of estimated location points (size = NA will draw no points). Optionally, a vector of length 2 with size of observed locations given by 2nd value (ignored if obs = FALSE) |
col.ssm |
colour of ssm-fitted or ssm-predicted locations (ignored if by.id = TRUE) |
col.obs |
colour of observed locations (ignored if obs = FALSE) |
lines |
logical indicating if lines are added to connect estimated locations (default = FALSE) |
landfill |
colour to use for land (default = grey(0.6)) |
map_type |
background map type (default = NULL, which uses rnaturalearth
to add landmasses); if packages |
pal |
hcl.colors palette to use (default: "Cividis"; type
|
rev |
reverse colour palette (logical) |
last_loc |
colour to render last location of each track (default = NULL) |
... |
additional arguments passed to ggspatial::annotation_map_tile |
aniMotum formatformat data by mapping supplied variable names to those expected by
fit_ssm(), and ensuring variables are put into the expected order. Can be
run manually by user as a data pre-processing step prior to calling fit_ssm()
or can be called automatically by fit_ssm(). In the latter case, any custom
variable names must be declared as arguments to fit_ssm(); see examples, below.
format_data( x, id = "id", date = "date", lc = "lc", coord = c("lon", "lat"), epar = c("smaj", "smin", "eor"), sderr = c("x.sd", "y.sd"), tz = "UTC" )format_data( x, id = "id", date = "date", lc = "lc", coord = c("lon", "lat"), epar = c("smaj", "smin", "eor"), sderr = c("x.sd", "y.sd"), tz = "UTC" )
x |
input data |
id |
the name (as a quoted character string) of id variable: a unique identifier for individual (animal) track data sets. |
date |
the name (as a quoted character string)of the date/time variable: date and time (as YYYY-MM-DD HH:MM:SS) of each observation. |
lc |
the name (as a quoted character string) of the location quality class variable: Argos location quality class (values in the set: 3,2,1,0,"A","B","Z"). Can also include "G" for GPS data and/or "GL" for light-level geolocation (GLS) and other data types. |
coord |
the names (as quoted character strings) of the location coordinate
variables: defaults are c("lon","lat"), but could also be c("x","y") for planar
coordinates; or if input data is an |
epar |
the names (as quoted character strings) of the Argos error ellipse parameters: defaults are "smaj" (ellipse semi-major axis), "smin" (ellipse semi-minor axis), and "eor" (ellipse orientation). Ignored if these variables are missing from the input data. |
sderr |
the names (as quoted character strings) of provided standard
errors for |
tz |
the timezone the applies to the data/time variable if they are not
in |
a data.frame or sf-tibble of input data in expected aniMotum format.
Additional columns required by fit_ssm(), if missing, will be added to the
formatted tibble: smaj, smin, eor, x.sd, and y.sd.
## as a data pre-processing step data(sese2_n) head(sese2_n, 5) d <- format_data(sese2_n, date = "time", coord = c("longitude","latitude"), tz = "America/Halifax") fit <- fit_ssm(d, model = "crw", time.step = 24) ## called automatically within fit_ssm() fit <- fit_ssm(sese2_n, date = "time", coord = c("longitude", "latitude"), tz = "America/Halifax", model = "crw", time.step = 24)## as a data pre-processing step data(sese2_n) head(sese2_n, 5) d <- format_data(sese2_n, date = "time", coord = c("longitude","latitude"), tz = "America/Halifax") fit <- fit_ssm(d, model = "crw", time.step = 24) ## called automatically within fit_ssm() fit <- fit_ssm(sese2_n, date = "time", coord = c("longitude", "latitude"), tz = "America/Halifax", model = "crw", time.step = 24)
tibble's by name from a aniMotum model objectgrab() lets you obtain fitted, predicted, rerouted or
data tibble's from a compound tibble created when fitting to multiple
individual data sets. The specified tibble's are appended to a single output
tibble.
grab(x, what = "fitted", as_sf = FALSE, normalise = FALSE, group = FALSE)grab(x, what = "fitted", as_sf = FALSE, normalise = FALSE, group = FALSE)
x |
a |
what |
the tibble to be grabbed; either |
as_sf |
logical; if FALSE (default) then return a |
normalise |
logical; if output includes a move persistence estimate,
should |
group |
logical; should |
if multiple ssm_df model objects are present in x, as_sf = TRUE,
and at least 1 estimated track has a coordinate reference system (crs) with
longitude centered on 180 (e.g. a track straddling -180,180) then all tracks
will be re-projected to that crs.
a tibble with all individual tibble's appended
## generate an ssm fit object xs <- fit_ssm(ellie, spdf=FALSE, model = "rw", time.step=24, control = ssm_control(verbose = 0)) ## grab predicted values as an un-projected tibble preds <- grab(xs, what = "predicted")## generate an ssm fit object xs <- fit_ssm(ellie, spdf=FALSE, model = "rw", time.step=24, control = ssm_control(verbose = 0)) ## grab predicted values as an un-projected tibble preds <- grab(xs, what = "predicted")
join() joins ssm-predicted locations and mpm-estimated behavioural index into a single tibble. If the ssm-predicted tibble is a projected sf object then the output of join will also be an sf object (default). This can be avoided by using as_sf = FALSE.
join( ssm, mpm, what.ssm = "predicted", as_sf = FALSE, normalise = FALSE, group = FALSE )join( ssm, mpm, what.ssm = "predicted", as_sf = FALSE, normalise = FALSE, group = FALSE )
ssm |
an |
mpm |
an |
what.ssm |
specifies whether ssm |
as_sf |
logical; if FALSE then return a tibble with un-projected lonlat coordinates, otherwise return an sf tibble |
normalise |
logical; if output includes a move persistence estimate, should g (the move persistence index) be normalised to have minimum = 0 and maximum = 1 (default = FALSE). |
group |
logical; should g be normalised among individuals as a group, a 'relative g', or separately to highlight regions of lowest and highest move persistence along a track (default = FALSE). |
a single tbl with all individuals
## load example aniMotum fit objects (to save time) ## generate a ssm fit object xs <- fit_ssm(ellie, spdf=FALSE, model = "rw", time.step=24, control = ssm_control(verbose = 0)) xm <- fit_mpm(xs, what = "p", model = "mpm") ## join predicted values as an un-projected tibble xsm <- join(xs, xm) xsm## load example aniMotum fit objects (to save time) ## generate a ssm fit object xs <- fit_ssm(ellie, spdf=FALSE, model = "rw", time.step=24, control = ssm_control(verbose = 0)) xm <- fit_mpm(xs, what = "p", model = "mpm") ## join predicted values as an un-projected tibble xsm <- join(xs, xm) xsm
map aniMotum-estimated locations and behavioural indices with coastline and projection options
map( x, y = NULL, what = c("fitted", "predicted", "rerouted"), aes = aes_lst(), by.id = TRUE, by.date = FALSE, crs = NULL, ext.rng = c(0.05, 0.05), buffer = 10000, map_type = "default", normalise = TRUE, group = FALSE, silent = FALSE, ... )map( x, y = NULL, what = c("fitted", "predicted", "rerouted"), aes = aes_lst(), by.id = TRUE, by.date = FALSE, crs = NULL, ext.rng = c(0.05, 0.05), buffer = 10000, map_type = "default", normalise = TRUE, group = FALSE, silent = FALSE, ... )
x |
a |
y |
optionally, a |
what |
specify which location estimates to map: fitted, predicted or rerouted |
aes |
a list of map controls and aesthetics (shape, size, col, fill, alpha)
for each map feature (estimated locations, confidence ellipses, track lines,
observed locations, land masses, water bodies). Constructed by |
by.id |
when mapping multiple tracks, should locations be coloured by
id (logical; default = TRUE if |
by.date |
when mapping single tracks, should locations be coloured by date (logical; default = FALSE; ignored if behavioural index provided) |
crs |
|
ext.rng |
proportion (can exceed 1) to extend the plot range in x and y dimensions |
buffer |
distance (in km) to buffer locations for subsetting land
polygons (default = 10000). If map extents are expanded by many factors then
the buffer distance may need to be increased, otherwise this should not be
used. Ignored if |
map_type |
background map type ("default" uses rnaturalearth::ne_countries
to add land polygons). If the |
normalise |
logical; if output includes a move persistence estimate, should g (the move persistence index) be normalised to have minimum = 0 and maximum = 1 (default = TRUE). |
group |
logical; should g be normalised among individuals as a group, a 'relative g', or separately to highlight regions of lowest and highest move persistence along a track (default = FALSE). |
silent |
logical; generate maps silently (default = FALSE). |
... |
additional arguments passed to ggspatial::annotation_map_tile |
a map as a ggplot2 object
# create an ssm fit object fit <- fit_ssm(ellie, model = "rw", time.step = 24, control = ssm_control(verbose = 0)) # render default map map(fit, what = "p")# create an ssm fit object fit <- fit_ssm(ellie, model = "rw", time.step = 24, control = ssm_control(verbose = 0)) # render default map map(fit, what = "p")
fit_mpm.mpm_control selects the numerical minimizer, method, associated
control parameters, and parameter bounds used by fit_mpm.
mpm_control( optim = c("nlminb", "optim"), method = c("L-BFGS-B", "BFGS", "Nelder-Mead", "CG", "SANN", "Brent"), lower = NULL, upper = NULL, verbose = 1, ... )mpm_control( optim = c("nlminb", "optim"), method = c("L-BFGS-B", "BFGS", "Nelder-Mead", "CG", "SANN", "Brent"), lower = NULL, upper = NULL, verbose = 1, ... )
optim |
the numerical optimizer used in the fit |
method |
if optim = "optim" then the optimization method to be used
can be one of "BFGS", "L-BFGS-B", "Nelder-Mead", "CG", "SANN", or "Brent"
see |
lower |
a list named parameter lower bounds, if NULL then built in
defaults are used when |
upper |
a list of named parameter upper bounds, if NULL then built in
defaults are used when |
verbose |
integer; report progress during minimization: 0 = silent; 1 = optimizer trace; 2 = parameter trace (default)) |
... |
control parameters for the chosen optimizer |
The optimizer used to minimize the objective function is
selected by the optim argument. Additional control
parameters specific to the chosen optimizer are specified via the
dots argument. See nlminb and optim
for available options. Adapted from S. Wotherspoon
https://github.com/SWotherspoon/RWalc/blob/master/R/RWalc.R
Returns a list with components
optim |
the name of the numerical optimizer as a string, "nlminb" or "optim" |
method |
optimization method to be used |
lower |
named list of lower parameter bounds |
upper |
named list of upper parameter bounds |
verbose |
level of tracing information to be reported |
control |
list of control parameters for the optimizer |
aniMotum
ssm fitcalculate one-step-ahead (prediction) residuals from a aniMotum
ssm fit
osar(x, method = "fullGaussian", ...)osar(x, method = "fullGaussian", ...)
x |
a |
method |
method to calculate prediction residuals
(default is |
... |
other arguments to TMB::oneStepPredict |
One-step-ahead residuals are useful for assessing goodness-of-fit
in latent variable models. This is a wrapper function for TMB::oneStepPredict
(beta version). osar tries the fullGaussian (fastest) method first and
falls back to the oneStepGaussianOffMode (slower) method for any failures.
Subsequent failures are dropped from the output and a warning message is given.
Note, OSA residuals can take a considerable time to calculate if there are
many individual fits and/or deployments are long. The method is automatically
parallelised across 2 x the number of individual fits, up to the number of
processor cores available.
Thygesen, U. H., C. M. Albertsen, C. W. Berg, K. Kristensen, and A. Neilsen. 2017. Validation of ecological state space models using the Laplace approximation. Environmental and Ecological Statistics 24:317–339.
# generate a ssm fit object (call is for speed only) xs <- fit_ssm(ellie, spdf=FALSE, model = "rw", time.step=24, control = ssm_control(verbose = 0)) res <- osar(xs)# generate a ssm fit object (call is for speed only) xs <- fit_ssm(ellie, spdf=FALSE, model = "rw", time.step=24, control = ssm_control(verbose = 0)) res <- osar(xs)
visualize fits from an mpm object
## S3 method for class 'mpm_df' plot( x, y = NULL, se = FALSE, pages = 0, ncol = 1, ask = TRUE, pal = "Plasma", rev = FALSE, ... )## S3 method for class 'mpm_df' plot( x, y = NULL, se = FALSE, pages = 0, ncol = 1, ask = TRUE, pal = "Plasma", rev = FALSE, ... )
x |
a |
y |
optional |
se |
logical (default = FALSE); should points be scaled by |
pages |
plots of all individuals on a single page (pages = 1; default) or each individual on a separate page (pages = 0) |
ncol |
number of columns to use for faceting. Default is ncol = 1 but this may be increased for multi-individual objects. Ignored if pages = 0 |
ask |
logical; if TRUE (default) user is asked for input before each plot is rendered. set to FALSE to return ggplot objects |
pal |
grDevices::hcl.colors palette to use (default: "Plasma"; see grDevices::hcl.pals for options) |
rev |
reverse colour palette (logical) |
... |
additional arguments to be ignored |
a ggplot object with either: 1-d time series of gamma_t estimates
(if y not provided), with estimation uncertainty ribbons (95 % CI's);
or 2-d track plots (if y provided) coloured by gamma_t, with smaller points
having greater uncertainty (size is proportional to SE^-2, if se = TRUE).
Plots can be rendered all on a single page (pages = 1) or on separate pages.
# generate a ssm fit object (call is for speed only) xs <- fit_ssm(sese2, spdf=FALSE, model = "rw", time.step=72, control = ssm_control(verbose = 0)) # fit mpm to ssm fits xm <- fit_mpm(xs, model = "jmpm") # plot 1-D mp timeseries on 1 page plot(xm, pages = 1)# generate a ssm fit object (call is for speed only) xs <- fit_ssm(sese2, spdf=FALSE, model = "rw", time.step=72, control = ssm_control(verbose = 0)) # fit mpm to ssm fits xm <- fit_mpm(xs, model = "jmpm") # plot 1-D mp timeseries on 1 page plot(xm, pages = 1)
plot One-Step-Ahead (prediction) residuals from a
aniMotum osar object
## S3 method for class 'osar' plot( x, type = c("ts", "qqnorm", "acf"), pages = 1, ncol = 1, ask = TRUE, pal = "Zissou1", ... )## S3 method for class 'osar' plot( x, type = c("ts", "qqnorm", "acf"), pages = 1, ncol = 1, ask = TRUE, pal = "Zissou1", ... )
x |
a |
type |
type of residual plot to generate: time-series (ts; default), qqnorm (qq), or acf |
pages |
plots of all individuals on a single page (pages = 1; default) or each individual on a separate page (pages = 0) |
ncol |
number of columns to use for faceting. Default is ncol = 2 but this may be increased for multi-individual fit objects |
ask |
logical; if TRUE (default) user is asked for input before each plot is rendered. set to FALSE to return ggplot objects |
pal |
grDevices::hcl.colors colour palette to use (default = "Zissou1";
see |
... |
additional arguments to be ignored |
## generate a fG_ssm fit object (call is for speed only) xs <- fit_ssm(ellie, spdf=FALSE, model = "rw", time.step=24, control = ssm_control(se = FALSE, verbose = 0)) res <- osar(xs) plot(res, type = "qq")## generate a fG_ssm fit object (call is for speed only) xs <- fit_ssm(ellie, spdf=FALSE, model = "rw", time.step=24, control = ssm_control(se = FALSE, verbose = 0)) res <- osar(xs) plot(res, type = "qq")
visualize simulated tracks from a sim data.frame
## S3 method for class 'sim' plot(x, type = 2, error = FALSE, pal = "Plasma", rev = FALSE, ...)## S3 method for class 'sim' plot(x, type = 2, error = FALSE, pal = "Plasma", rev = FALSE, ...)
x |
a |
type |
either 1, a 1-D time-series of speed (if model is |
error |
logical, plot locations with error (TRUE) or without. Ignored in 1-D time-series plots |
pal |
grDevices::hcl.colors palette to use (default: "Plasma");
see |
rev |
reverse direction of colour palette; logical (default = FALSE) |
... |
additional arguments to be ignored |
Plots of simulated tracks. Can be rendered all on a single page (pages = 1) or on separate pages (pages = 0).
tr <- sim(N=200, model = "mp") plot(tr)tr <- sim(N=200, model = "mp") plot(tr)
visualize tracks simulated from a aniMotum model fit
## S3 method for class 'sim_fit' plot(x, type = NULL, zoom = TRUE, ncol = 1, hires = FALSE, ortho = TRUE, ...)## S3 method for class 'sim_fit' plot(x, type = NULL, zoom = TRUE, ncol = 1, hires = FALSE, ortho = TRUE, ...)
x |
a |
type |
deprecated. All tracks are rendered as points. |
zoom |
logical; should map extent be defined by track extent (TRUE; default) or should global map be drawn (FALSE). |
ncol |
number of columns to arrange multiple plots |
hires |
logical; use high-resolution coastline data. Attempts to use
high-res coastline data via rnaturalearth::ne_countries with |
ortho |
logical; use an orthographic projection centered on the track starting location(s) (TRUE; default). An orthographic projection may be optimal for high latitude tracks and/or tracks that traverse long distances. If FALSE then a global Mercator projection is used. |
... |
additional arguments to be ignored |
Plots of simulated tracks.
fit <- fit_ssm(ellie, model = "crw", time.step = 24) trs <- sim_fit(fit, what = "p", reps = 2) plot(trs)fit <- fit_ssm(ellie, model = "crw", time.step = 24) trs <- sim_fit(fit, what = "p", reps = 2) plot(trs)
visualize tracks simulated from a aniMotum model fit
## S3 method for class 'sim_post' plot( x, type = c("lines", "points", "both"), zoom = TRUE, ncol = 1, hires = TRUE, ortho = TRUE, alpha = 0.5, ... )## S3 method for class 'sim_post' plot( x, type = c("lines", "points", "both"), zoom = TRUE, ncol = 1, hires = TRUE, ortho = TRUE, alpha = 0.5, ... )
x |
a |
type |
plots tracks as "line", "points" or "both" (default). |
zoom |
logical; should map extent be defined by track extent (TRUE; default) or should global map be drawn (FALSE). |
ncol |
number of columns to arrange multiple plots |
hires |
logical; use high-resolution coastline data. Attempts to use
high-res coastline data via rnaturalearth::ne_countries with |
ortho |
logical; use an orthographic projection centered on the track starting location(s) (TRUE; default). An orthographic projection may be optimal for high latitude tracks and/or tracks that traverse long distances. If FALSE then a global Mercator projection is used. |
alpha |
opacity of simulated track points/lines. Lower opacity can ease visualization when multiple simulated overlap one another. |
... |
additional arguments to be ignored |
Plots of posterior simulated tracks.
fit <- fit_ssm(ellie, model = "crw", time.step = 24) psim <- sim_post(fit, what = "p", reps = 10) plot(psim, type = "lines")fit <- fit_ssm(ellie, model = "crw", time.step = 24) psim <- sim_post(fit, what = "p", reps = 10) plot(psim, type = "lines")
visualize fits from an ssm object
## S3 method for class 'ssm_df' plot( x, what = c("fitted", "predicted", "rerouted"), type = 1, outlier = TRUE, alpha = 0.3, pages = 0, ncol = 1, ask = TRUE, pal = "default", normalise = TRUE, group = FALSE, ... )## S3 method for class 'ssm_df' plot( x, what = c("fitted", "predicted", "rerouted"), type = 1, outlier = TRUE, alpha = 0.3, pages = 0, ncol = 1, ask = TRUE, pal = "default", normalise = TRUE, group = FALSE, ... )
x |
a |
what |
specify which location estimates to display on time-series plots: fitted, predicted, or rerouted |
type |
of plot to generate: 1-d time series for lon and lat separately
(type = 1, default); 2-d track plot (type = 2); 1-d time series of move
persistence estimates (type = 3; if fitted model was |
outlier |
include outlier locations dropped by prefilter (outlier = TRUE, default) |
alpha |
opacity of standard errors. Lower opacity can ease visualization when multiple ellipses overlap one another |
pages |
each individual is plotted on a separate page by default (pages = 0), multiple individuals can be combined on a single page; pages = 1 |
ncol |
number of columns to arrange plots when combining individuals on a single page (ignored if pages = 0) |
ask |
logical; if TRUE (default) user is asked for input before each plot is rendered. set to FALSE to return ggplot objects |
pal |
grDevices::hcl.colors palette to use (see |
normalise |
logical; if plotting move persistence estimates from an |
group |
logical; should |
... |
additional arguments to be ignored |
a ggplot object with either: (type = 1) 1-d time series of fits to data, separated into x and y components (units = km) with prediction uncertainty ribbons (2 x SE); or (type = 2) 2-d fits to data (units = km)
## generate a ssm fit object (call is for speed only) xs <- fit_ssm(sese2, spdf=FALSE, model = "rw", time.step=72, control = ssm_control(verbose = 0)) # plot fitted locations as 1-D timeseries on 1 page plot(xs, what = "f", pages = 1)## generate a ssm fit object (call is for speed only) xs <- fit_ssm(sese2, spdf=FALSE, model = "rw", time.step=72, control = ssm_control(verbose = 0)) # plot fitted locations as 1-D timeseries on 1 page plot(xs, what = "f", pages = 1)
aniMotum fit object summary informationprint aniMotum fit object summary information
## S3 method for class 'ssm' print(x, ...)## S3 method for class 'ssm' print(x, ...)
x |
a aniMotum ssm fit object |
... |
unused. For compatibility with the generic method. |
## see summary fit output ## generate a ssm fit object (call is for speed only) xs <- fit_ssm(ellie, spdf=FALSE, model = "rw", time.step=24, control = ssm_control(se = FALSE, verbose = 0)) xs$ssm[[1]]## see summary fit output ## generate a ssm fit object (call is for speed only) xs <- fit_ssm(ellie, spdf=FALSE, model = "rw", time.step=24, control = ssm_control(se = FALSE, verbose = 0)) xs$ssm[[1]]
pathroutr packageA wrapper function that uses the pathroutr package
https://jmlondon.github.io/pathroutr/
to re-route movement paths that cross a land barrier. The current
implementation will take either the output from a fit_ssm model or the
simulations generated by sim_fit.
route_path( x, what = c("fitted", "predicted"), map_scale = 50, dist = 50000, append = TRUE, ... )route_path( x, what = c("fitted", "predicted"), map_scale = 50, dist = 50000, append = TRUE, ... )
x |
either a |
what |
if using a |
map_scale |
scale of rnaturalearth map to use for land mass: one of 110,
50 (default), or 10. Note that map_scale = 10 is only available if you have
the |
dist |
buffer distance (m) to add around track locations. The convex hull of these buffered locations defines the size of land polygon used to aid re-routing of points on land. Larger buffers can result in longer computation times. See London (2020) for further details. The default buffer distance is a constant 50000 m. |
append |
should re-routed locations be appended to the |
... |
additional arguments passed to pathroutr::prt_visgraph |
route_path uses rnaturalearth::ne_countries at the medium (50)
scale, by default, to generate a land barrier. For efficient computation,
route_path clips the polygons to the buffered bounds (set by dist (in m)
of the movement track(s).
When the input is a ssm object route_path can append the
re-routed path locations to the ssm (ssm fit) object. This is useful
when move persistence is to be estimated from the re-routed locations via
fit_mpm, or tracks are to be visualised with map. route_path can also
return a standalone tibble of the re-routed path with the same number of
locations as either the original fitted or predicted locations.
When the re-routed path is appended to the ssm object, the path can be
extracted using the grab function, e.g. grab(fit, what = "rerouted").
When the input is a sim_fit object then route_path returns the same
object but with the locations within each simulation re-routed.
We recommend that users working on complex rerouting problems and/or
requiring higher resolution land barrier data work with the pathroutr
package directly by first exctracting aniMotum-estimated locations with
grab. Higher resolution land barrier data (polygon shapefiles) must be
obtained independently.
Josh M. London. (2020) pathroutr: An R Package for (Re-)Routing Paths Around Barriers (Version v0.2.1) https://zenodo.org/record/5522909#.YnPxEy_b1qs
# if 'pathroutr' is installed then ok to use route_path() if(requireNamespace("pathroutr", quietly = TRUE)) { fit <- fit_ssm(ellie, vmax = 4, model = "crw", time.step = 24) fit <- route_path(fit, what = "predicted") grab(fit, what = "rerouted") }# if 'pathroutr' is installed then ok to use route_path() if(requireNamespace("pathroutr", quietly = TRUE)) { fit <- fit_ssm(ellie, vmax = 4, model = "crw", time.step = 24) fit <- route_path(fit, what = "predicted") grab(fit, what = "rerouted") }
Example elephant seal Argos tracking data. Data were sourced from the Integrated Marine Observing System (IMOS) Sourced from the Australian Integrated Marine Observing System (IMOS) in collaboration with the French IPEV and SNO-MEMO project deployments at Iles Kerguelen and are publicly available (http:// imos.aodn.org.au). IMOS is supported by the Australian Government through the National Collaborative Research Infrastructure Strategy and the Super Science Initiative.
.RData
Example elephant seal Argos tracking data, highly sub-sampled. These example data are included purely to speed up examples where a fit object is required. Generating a fit object is preferred as storing an example fit risks GDAL errors on platforms with older GDAL libraries. Sourced from the Australian Integrated Marine Observing System (IMOS) in collaboration with the French IPEV and SNO-MEMO project deployments at Iles Kerguelen and are publicly available (http:// imos.aodn.org.au). IMOS is supported by the Australian Government through the National Collaborative Research Infrastructure Strategy and the Super Science Initiative.
.RData
Example elephant seal Argos tracking data, highly sub-sampled with default variable order scrambled and renamed. These example data are included purely to speed up examples where a fit object is required. Generating a fit object is preferred as storing an example fit risks GDAL errors on platforms with older GDAL libraries. Sourced from the Australian Integrated Marine Observing System (IMOS) in collaboration with the French IPEV and SNO-MEMO project deployments at Iles Kerguelen and are publicly available (http:// imos.aodn.org.au). IMOS is supported by the Australian Government through the National Collaborative Research Infrastructure Strategy and the Super Science Initiative.
.RData
simulate from the rw, crw, or mp process models
to generate a set of x,y (or lon,lat) coordinates with or without error
from supplied input parameters.
sim( N = 100, start = list(c(0, 0), as.POSIXct(format(Sys.time(), tz = "UTC", usetz = TRUE))), model = c("rw", "crw", "mp"), vmax = 4, sigma = c(4, 4), rho_p = 0, D = 0.05, sigma_g = 1.25, error = c("ls", "kf"), tau = c(1.5, 0.75), rho_o = 0, tdist = c("reg", "gamma"), ts = 6, tpar = 1.2, alpha = c(0.9, 0.8) )sim( N = 100, start = list(c(0, 0), as.POSIXct(format(Sys.time(), tz = "UTC", usetz = TRUE))), model = c("rw", "crw", "mp"), vmax = 4, sigma = c(4, 4), rho_p = 0, D = 0.05, sigma_g = 1.25, error = c("ls", "kf"), tau = c(1.5, 0.75), rho_o = 0, tdist = c("reg", "gamma"), ts = 6, tpar = 1.2, alpha = c(0.9, 0.8) )
N |
number of time steps to simulate |
start |
coordinates and datetime of start location for simulated track |
model |
simulate from the |
vmax |
maximum travel rate (m/s) of simulated animal |
sigma |
a vector of process error sd's for the |
rho_p |
correlation parameter for |
D |
diffusion coefficient for |
sigma_g |
random walk sd for time-varying move persistence parameter
(ignored if |
error |
indicates whether measurement error should mimic Argos
Least-Squares ( |
tau |
vector of LS measurement error sd's (ignored if |
rho_o |
correlation parameter for LS covariance matrix
(ignored if |
tdist |
distribution for simulating location times ( |
ts |
time interval in h |
tpar |
rate parameter for the gamma distributed times, shape is take to be
|
alpha |
transition probabilities switching model versions of
|
a tibble is returned with columns that can include some or all of the following, depending on the arguments used
date time as POSIXct, tz = UTC (default)
lc Argos location class
lon longitude with error
lat latitude with error
x x in km from arbitrary origin without error
y y in km from arbitrary origin without error
x.err a random deviate drawn from Argos LS or KF error distribution
y.err a random deviate drawn from Argos LS or KF error distribution
smaj Argos error ellipse semi-major axis in m (if error = "kf")
smin Argos error ellipse semi-minor axis in m (if error = "kf")
eor Argos error ellipse orientation in degrees (if error = "kf")
u velocity in x direction (if model = "crw"), unit = km/h
v velocity in y direction (if model = "crw"), unit = km/h
b behavioural state (if model = "rw" or model = "crw" and multiple process variances given, see examples)
g movement persistence - the autocorrelation between successive movements on the interval 0,1 (if model = "mp")
tr <- sim(N = 200, model = "crw", D = 0.1, error = "kf", tdist = "reg", ts=12) plot(tr, error = TRUE) tr <- sim(N = 200, model = "mp", sigma_g = 1.2, error = "ls", tau = c(2, 1.5), ts=12, tdist = "gamma", tpar = 1.5) plot(tr, error = TRUE, pal = "Cividis")tr <- sim(N = 200, model = "crw", D = 0.1, error = "kf", tdist = "reg", ts=12) plot(tr, error = TRUE) tr <- sim(N = 200, model = "mp", sigma_g = 1.2, error = "ls", tau = c(2, 1.5), ts=12, tdist = "gamma", tpar = 1.5) plot(tr, error = TRUE, pal = "Cividis")
sim_fit simulationsThis function calculates the similarity between the simulations
generated by sim_fit and the SSM-estimated path from the ssm fit,
and returns a sim_fit object containing the most similar tracks based on
a user specified quantile. In this context, similarity is calculated
as the sum of normalised differences in net displacement (km) and overall
bearing (deg) between the SSM-estimated path and the simulated paths.
sim_filter(trs, keep = 0.25, flag = 2, var = NULL, FUN = "mean", ...)sim_filter(trs, keep = 0.25, flag = 2, var = NULL, FUN = "mean", ...)
trs |
a |
keep |
the quantile of flag values to retain |
flag |
the similarity flag method (see details). Ignored if var != NULL. |
var |
the name(s) of the appended variable(s) to use for similarity calculations. Default is NULL, in which case similarity is calculated based on distance and bearing - e.g., Hazen et al (2017). |
FUN |
one of the following functions in quotes: mean, median, var, sd, sum, min, or max. Ignored if var = NULL. |
... |
additional arguments to the specified FUN (e.g., na.rm = TRUE). Ignored if var = NULL. |
flag = 1 will use an index based on Hazen (2017)
flag = 2 (the default) will use a custom index
a sim_fit object containing the filtered paths
Hazen et al. (2017) WhaleWatch: a dynamic management tool for predicting blue whale density in the California Current J. Appl. Ecol. 54: 1415-1428
## fit crw model to Argos LS data fit <- fit_ssm(ellie, model = "crw", time.step = 72) set.seed(pi) ## generate 5 simulated paths from ssm fit trs <- sim_fit(fit, what = "predicted", reps = 5) ## filter simulations and keep paths in top 40% of flag values trs_f <- sim_filter(trs, keep = 0.4, flag = 2) ## compare unfiltered and filtered simulated paths plot(trs) | plot(trs_f)## fit crw model to Argos LS data fit <- fit_ssm(ellie, model = "crw", time.step = 72) set.seed(pi) ## generate 5 simulated paths from ssm fit trs <- sim_fit(fit, what = "predicted", reps = 5) ## filter simulations and keep paths in top 40% of flag values trs_f <- sim_filter(trs, keep = 0.4, flag = 2) ## compare unfiltered and filtered simulated paths plot(trs) | plot(trs_f)
ssm fitsimulate from the rw or crw process models to generate
either a set of x,y or lon,lat coordinates from a ssm fit with length
equal to the number of observations used in the SSM fit.
sim_fit( x, what = c("fitted", "predicted"), reps = 1, start = NULL, end = NULL, grad = NULL, beta = c(-300, -300), cpf = FALSE, sim_only = FALSE )sim_fit( x, what = c("fitted", "predicted"), reps = 1, start = NULL, end = NULL, grad = NULL, beta = c(-300, -300), cpf = FALSE, sim_only = FALSE )
x |
a |
what |
simulate fitted (typically irregular in time) or predicted (typically regular in time) locations |
reps |
number of replicate tracks to simulate from an |
start |
a 2-element vector for the simulated track start location (lon,lat or x,y) |
end |
a 2-element vector for the simulated track end location (lon,lat or x,y) |
grad |
a SpatRaster of x- and y-gradients as separate layers (see details) |
beta |
a 2-element vector of parameters defining the potential function
magnitude in x- and y-directions (ignored if |
cpf |
logical; should simulated tracks return to their start point (ie. a central-place forager) |
sim_only |
logical, do not include |
A potential function can be applied to the simulated paths to help
avoid locations on land (or in water), using the grad and beta
arguments. A coarse-resolution rasterStack of global x- and y-gradients of
distance to land are provided. Stronger beta parameters result in stronger
land (water) avoidance but may also introduce undesirable/unrealistic artefacts
(zig-zags) in the simulated paths. See Brillinger et al. (2012) and
vignette("momentuHMM", package = "momentuHMM") for more details on the
use of potential functions for simulating constrained animal movements.
WARNING: This application of potential functions to constrain simulated
paths is experimental, likely to change in future releases, and NOT guaranteed
to work enitrely as intended, especially if cpf = TRUE!
a fG_sim_fit object containing the paths simulated from a
ssm fit object
Brillinger DR, Preisler HK, Ager AA, Kie J (2012) The use of potential functions in modelling animal movement. In: Guttorp P., Brillinger D. (eds) Selected Works of David Brillinger. Selected Works in Probability and Statistics. Springer, New York. pp. 385-409.
fit <- fit_ssm(ellie, model = "crw", time.step = 24) trs <- sim_fit(fit, what = "predicted", reps = 3) plot(trs)fit <- fit_ssm(ellie, model = "crw", time.step = 24) trs <- sim_fit(fit, what = "predicted", reps = 3) plot(trs)
ssm fit.simulates track locations from the joint precision matrix of a
ssm model fit. Currently, the joint precision of the SSM movement
parameters is not included (ie. a full posterior simulation).
sim_post(x, what = "predicted", reps = 1, sim_only = FALSE)sim_post(x, what = "predicted", reps = 1, sim_only = FALSE)
x |
a |
what |
simulate fitted or predicted locations |
reps |
number of replicate tracks to simulate from the |
sim_only |
logical, do not include |
a fG_sim_post object containing the paths simulated from a
ssm fit object
fit <- fit_ssm(ellie, model = "crw", time.step = 24) psim <- sim_post(fit, "p", reps = 10) plot(psim, type = "lines")fit <- fit_ssm(ellie, model = "crw", time.step = 24) psim <- sim_post(fit, "p", reps = 10) plot(psim, type = "lines")
fit_ssm.ssm_control selects the numerical minimizer, method, associated
control parameters, and parameter bounds used by fit_ssm.
ssm_control( optim = c("nlminb", "optim"), method = c("L-BFGS-B", "BFGS", "Nelder-Mead", "CG", "SANN", "Brent"), lower = NULL, upper = NULL, verbose = 1, se = FALSE, ... )ssm_control( optim = c("nlminb", "optim"), method = c("L-BFGS-B", "BFGS", "Nelder-Mead", "CG", "SANN", "Brent"), lower = NULL, upper = NULL, verbose = 1, se = FALSE, ... )
optim |
the numerical optimizer used in the fit |
method |
if optim = "optim" then the optimization method to be used
can be one of "BFGS", "L-BFGS-B", "Nelder-Mead", "CG", "SANN", or "Brent"
see |
lower |
a list named parameter lower bounds, if NULL then built in
defaults are used when |
upper |
a list of named parameter upper bounds, if NULL then built in
defaults are used when |
verbose |
integer; report progress during minimization: 0 = silent; 1 = parameter trace (default); 2 = optimizer trace |
se |
logical; should standard errors for speed estimates be calculated (default = FALSE). Turning this on will slow down computation time but provide SE's for speed-along-track calculations |
... |
control parameters for the chosen optimizer |
The optimizer used to minimize the objective function is
selected by the optim argument. Additional control
parameters specific to the chosen optimizer are specified via the
dots argument. See nlminb and optim
for available options. Adapted from S. Wotherspoon
https://github.com/SWotherspoon/RWalc/blob/master/R/RWalc.R
Returns a list with components
optim |
the name of the numerical optimizer as a string, "nlminb" or "optim" |
method |
optimization method to be used |
lower |
named list of lower parameter bounds |
upper |
named list of upper parameter bounds |
verbose |
level of tracing information to be reported |
control |
list of control parameters for the optimizer |
fit <- fit_ssm(ellie, vmax = 4, model = "crw", time.step = 72, control = ssm_control( optim = "nlminb", eval.max = 2000) )fit <- fit_ssm(ellie, vmax = 4, model = "crw", time.step = 72, control = ssm_control( optim = "nlminb", eval.max = 2000) )
return a summary of an ssm_df fit object
## S3 method for class 'ssm_df' summary(object, ...)## S3 method for class 'ssm_df' summary(object, ...)
object |
an |
... |
additional arguments to be ignored |
Example Weddell seal Argos tracking data, deployed at Scott Base, Ross Island, Antarctica. This example data set is included for demonstration purposes. Sourced from the Australian Integrated Marine Observing System (IMOS) & the New Zealand National Institute of Water and Atmospheric Research (NIWA) deployments at Scott Base, Antarctica and are publicly available (http:// imos.aodn.org.au). IMOS is supported by the Australian Government through the National Collaborative Research Infrastructure Strategy and the Super Science Initiative.
.RData