The hardware and bandwidth for this mirror is donated by METANET, the Webhosting and Full Service-Cloud Provider.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]metanet.ch.
This vignette is an introduction to the admtools package.
Load the package using
This makes all functions in the package available.
Use
to get an overview of the available help pages of the package, and
to view a simple help page for the package.
Vignettes are a long form of package documentation that provide more detailed examples. To list the available vignettes, use
The admtools package defines three main classes:
adm,
sac
and multiadm
. The class
adm
represent a single age-depth model, from which
information can be extracted (e.g. completeness, number of hiatuses,
etc.) and that can be used to transform data between the stratigraphic
domain and time domain. The multiamd
class is a list of
adm
objects. multiadm
objects are used to
represent uncertainties of age-depth models. Objects of type
sac
are sediment accumulation curves that can contain
information on erosion, and can be turned into age-depth models.
In contrast to its name, the admtools package currently deals with time and height instead of age and depth. In this sense, the age-depth models are time-height models. Both time and height can be negative values. To handle ages, use time before the present. To handle depths, use height below a point of reference (e.g., the sediment surface).
This example explains the construction and application of
adm
objects. As example data we use outputs from CarboCAT
Lite, a model of carbonate platform growth (Burgess 2013, 2023). This data is automatically
loaded in the background by the package. To get some info of the data
use
This data is identical to scenario A from Hohmann et al. (2024) as published in Hohmann et al. (2023). See therein for details on the simulation, reproducibility, and a chronostratigraphic chart and a figure showing a transect through the carbonate platform.
The standard constructor for age-depth models is
tp_to_adm
(“tie point to age-depth model”). It returns an
objecto of class adm
. This object combines information of
stratigraphic heights and times and erosive interval. It allows to
transform data between the stratigraphic and the time domain, and
identify which data is destroyed due to hiatuses.
As example, I use the timing and stratigraphic positions of tie points taken from CarboCAT Lite to construct an age-depth model, and use the option to directly associate length and time units with it.
# see ?tp_to_adm for detailed documentation
my_adm = tp_to_adm(t = CarboCATLite_data$time_myr,
h = CarboCATLite_data$height_2_km_offshore_m,
L_unit = "m",
T_unit = "Myr")
This age-depth model represents the relationship between elapsed model time and accumulated sediment 2 km offshore in a synthetic carbonate platform.
Typing the name my_adm
in the console will only tell
that the generated variable is an age-depth model
To get a quick overview of the properties of my_adm
, use
summary
:
summary(my_adm)
#> age-depth model
#> Total duration: 2 Myr
#> Total thickness: 146.0621 m
#> Stratigraphic completeness: 32.65 %
#> 10 hiatus(es)
If you want to inspect the insides of the object, use
str
:
str(my_adm)
#> List of 5
#> $ t : num [1:2001] 0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 ...
#> $ h : num [1:2001] 0 0 0.609 1.095 1.722 ...
#> $ destr : logi [1:2000] TRUE FALSE FALSE FALSE FALSE FALSE ...
#> $ T_unit: chr "Myr"
#> $ L_unit: chr "m"
#> - attr(*, "class")= chr "adm"
You can manually manipulate the fields of the adm
object
by treating it like a list. I do not recommend doing so, as it might
result in unexpected downstream behavior. If you want to extract tie
points use get_L_tp
and get_T_tp
.
You can plot adm
objects via the standard
plot
function. Here, I use the option to highlight hiatuses
in red, and increase the linw width of the conservative ( =
non-destructive) intervals.
# see ?plot.adm for plotting options for adm objects
plot(my_adm,
col_destr = "red",
lwd_acc = 2)
T_axis_lab() # plot time axis label, see ?T_axis_lab for details
L_axis_lab() # plot height axis label, see ?L_axis_lab for details
You can also plot sedimentation rates in the time domain using
plot_sed_rate_t
“:
For more information on the extraction of sedimentation rates in the
time domain see the functions sed_rate_t
and
sed_rate_t_fun
. Sedimentation rates in the length domain
can be extracted using sed_rate_l
and
sed_rate_l_fun
and plotted via
plot_sed_rate_l
. In addition, condensation (time preserved
per stratigraphic increment) can be examined using the functions
condensation
, condensation_fun
and
plot_condensation
.
Use the functions get_total_duration
,
get_total_thickness
, get_completeness
, and
get_hiat_no
to extract information:
get_total_duration(my_adm) #total time covered by the age-depth model
#> [1] 2
get_total_thickness(my_adm) # total thickness of section represented by the adm
#> [1] 146.0621
get_completeness(my_adm) # stratigraphic completeness as proportion
#> [1] 0.3265
get_incompleteness(my_adm) # stratigraphic incompleteness (= 1- strat. incompleteness)
#> [1] 0.6735
get_hiat_no(my_adm) # number of hiatuses
#> [1] 10
For more detailed information, you can use
get_hiat_duration
to get a vector of hiatus
durationsget_hiat_list
to get a list of hiatus positions and
duration.For example, to plot a histogram of hiatus durations, use
hist(x = get_hiat_duration(my_adm),
freq = TRUE,
xlab = "Hiatus duration [Myr]",
main = "Hiatus duration 2 km offshore")
The function is_destructive
can be used to examine
whether points in time coincide with hiatuses:
The functions get_height
and get_time
are
the workhorses to transform data using age-depth models.
get_time
takes and adm
object and vector
of heights h
(stratigraphic positions), and returns a
vector of timesget_height
takes an adm
object and vector
of times t
and returns a vector of associated stratigraphic
positionsAs example, say we want to know the time of deposition of the following stratigraphic positions:
Conversely, to determine what parts of the section are deposited as a specific time, use
Here, the NA
indicates that the time 1.4 coincides with
erosion. If you want to know the stratigraphic position of the hiatus
that coincides with that time, use the option
destructive = FALSE
:
Alternatively, you can also use the wrappers
strat_to_time
and time_to_strat
for the
transformation. For expanded modeling features please use the
StratPal
package (Github | Webpage | CRAN). It
provides more biological context and utility functions to build modeling
pipelines that include ecological, taphonomic, stratigraphic, and
evolutionary effects.
The admtools
package can transform complex objects
between the time and stratigraphic domain. This is done using the
functions strat_to_time
and time_to_strat
.
As an example, we transform a chronogram (a phylogenetic tree where
branch length represents time). An example tree following the
birth-death model is stored with the package as the variable
timetree
. See ?timetree
for details on how
this tree was generated.
#install.packages("ape") Package for analyses of phylogenetics and evolution
# see ?ape::rlineage for help
#set.seed(1)
ape::plot.phylo(timetree) # see also ?ape::plot.phylo
axis(1)
mtext("Time [Myr]", side = 1, line = 2.5)
You can transform the tree using time_to_strat
:
Plotting the resulting tree along the stratigraphic column shows how the evolutionary relationships would be preserved 2 km offshore in the simulated carbonate platform:
admtools can transform lists from the time to the height
domain and vice versa given they have elements with names h
or t
. These lists can be interpreted as time/stratigraphic
series, where times and stratigraphic positions are associated with
measured values. Note that these are not ts
objects as used
by the stats
package, as they will be generally
heterodistant due to the irregular nature of the age-depth
transformation.
As example, we simulate trait evolution over 2 Myr using a Brownian motion, and transform the simulation into the stratigraphic domain.
t = seq(0, 2, by = 0.001) # times
BM = function(t){
#" Simulate Brownian motion at times t
li = list("t" = t,
"y" = cumsum(c(0, rnorm(n = length(t) - 1, mean = 0, sd = sqrt(diff(t))))))
class(li) = c("timelist", "list") # assign class `timelist` for easy plotting, see ?plot.timelist
return(li)
}
evo_list = BM(t)
plot(x = evo_list,
xlab = "Time [Myr]",
ylab = "Trait Value",
type = "l")
strat_list = time_to_strat(obj = evo_list,
x = my_adm)
plot(x = strat_list,
orientation = "lr",
type = "l",
xlab = "Stratigraphic Height [m]",
ylab = "Trait Value",
main = "Trait Evolution 2 km Offshore")
Note the jump in traits generated by the erosional interval in
my_adm
. Both time_to_strat
and
strat_to_time
return stratlist
and
timelist
objects when applied to lists. These are like
ordinary lists, but come with simplified plotting optionality, see
?plot.stratlist
and ?plot.timelist
for
details.
For expanded modeling features with biological context, please use
the StratPal
package (Github | Webpage | CRAN). It
provides light wrappers around admtools
and out of the box
modeling of trait evolution.
Numeric vectors can be transformed using time_to_strat
and strat_to_time
too. These are essentially high level
wrappers around get_height
and get_time
. See
functions time_to_strat.numeric
and
strat_to_time.numeric
for details.
For an overview of the structure of the admtools
package
and the classes used therein see
For details on plotting ADMs see
For information on estimating age-depth models from sedimentation rates, see
For information on estimating age-depth models from tracer contents of rocks and sediments, see
Burgess, Peter. “CarboCAT: A cellular automata model of heterogeneous carbonate strata.” Computers & geosciences 53 (2013): 129-140. DOI: 10.1016/j.cageo.2011.08.026
Burgess, Peter. “CarboCAT Lite v1.0.1”. Zenodo 2023. DOI: 10.5281/zenodo.8402578
Hohmann, Niklas; Koelewijn, Joël R.; Burgess, Peter; Jarochowska, Emilia. 2024. “Identification of the mode of evolution in incomplete carbonate successions.” BMC Ecology and Evolution 24, 113. https://doi.org/10.1186/s12862-024-02287-2.
Hohmann, Niklas, Koelewijn, Joël R.; Burgess, Peter; Jarochowska, Emilia. 2023. “Identification of the Mode of Evolution in Incomplete Carbonate Successions - Supporting Data.” Open Science Framework. https://doi.org/10.17605/OSF.IO/ZBPWA, published under the CC-BY 4.0 license.
These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.