Bacon: Main age-depth modelling function

In rbacon: Age-Depth Modelling using Bayesian Statistics

Main age-depth modelling function

Description

This is the main age-depth modelling function of the rbacon package.

Usage

Bacon(
  core = "MSB2K",
  thick = 5,
  coredir = "",
  prob = 0.95,
  d.min = NA,
  d.max = NA,
  add.bottom = TRUE,
  d.by = 1,
  seed = NA,
  depths.file = FALSE,
  depths = c(),
  depth.unit = "cm",
  age.unit = "yr",
  unit = depth.unit,
  acc.shape = 1.5,
  acc.mean = 20,
  mem.strength = 10,
  mem.mean = 0.5,
  boundary = NA,
  hiatus.depths = NA,
  hiatus.max = 10000,
  add = c(),
  after = 1e-04/thick,
  cc = 1,
  cc1 = "IntCal20",
  cc2 = "Marine20",
  cc3 = "SHCal20",
  cc4 = "ConstCal",
  cc.dir = c(),
  postbomb = 0,
  delta.R = 0,
  delta.STD = 0,
  t.a = 3,
  t.b = 4,
  normal = FALSE,
  suggest = TRUE,
  accept.suggestions = FALSE,
  reswarn = c(10, 200),
  remember = TRUE,
  ask = TRUE,
  run = TRUE,
  defaults = "defaultBacon_settings.txt",
  sep = ",",
  dec = ".",
  runname = "",
  slump = c(),
  remove = FALSE,
  BCAD = FALSE,
  ssize = 4000,
  th0 = c(),
  burnin = min(500, ssize),
  youngest.age = c(),
  oldest.age = c(),
  MinAge = c(),
  MaxAge = c(),
  cutoff = 0.01,
  plot.pdf = TRUE,
  dark = 1,
  date.res = 100,
  age.res = 200,
  yr.res = age.res,
  close.connections = TRUE,
  save.info = TRUE,
  older.than = c(),
  younger.than = c(),
  save.elbowages = FALSE,
  verbose = TRUE,
  ...
)

Arguments

core	Name of the core, given using quotes. Defaults to one of the cores provided with rbacon, core="MSB2K". An alternative core provided with this package is RLGH3 (Jones et al., 1989). To run your own core, produce a .csv file with the dates as outlined in the manual, add a folder with the core's name to the default directory for cores (see coredir), and save the .csv file there. For example, the file's location and name could be Bacon_runs/MyCore/MyCore.csv. Then run Bacon as follows: Bacon("MyCore")
thick	Bacon will divide the core into sections of equal thickness specified by thick (default thick=5).
coredir	Folder where the core's files core are and/or will be located. This will be a folder with the core's name, within either the folder coredir='Bacon_runs/', or the folder Cores/ if it already exists within R's working directory, or a custom-built folder. For example, use coredir="." to place the core's folder within the current working directory, or coredir="F:" if you want to put the core's folder and files on a USB drive loaded under F:. Thinner (and thus more) sections will result in smoother age-models, but too many sections can cause ‘run-away’ models.
prob	Confidence interval to report. This should lie between 0 and 1, default 0.95 (95 %).
d.min	Minimum depth of age-depth model (use this to extrapolate to depths higher than the top dated depth).
d.max	Maximum depth of age-depth model (use this to extrapolate to depths below the bottom dated depth).
add.bottom	Add a model section at the bottom of the core, in order to ensure the bottommost date is taken into account. Default add.bottom=TRUE. This is a new option and can cause age-models to differ from previous version. Please re-run the model if in doubt.
d.by	Depth intervals at which ages are calculated. Defaults to d.by=1. Please ensure that the value of d.by is smaller than that of 'thick', otherwise plots might turn out wrong.
seed	Seed used for C++ executions. If it is not assigned (seed=NA; default) then the seed is set by system.
depths.file	By default, Bacon will calculate the ages for the depths d.min to d.max in steps of d.by. If depths.file=TRUE, Bacon will read a file containing the depths for which you require ages. This file, containing the depths in a single column without a header, should be stored within coredir, and its name should start with the core's name and end with ‘_depths.txt’. Then specify depths.file=TRUE (default FALSE). See also depths.
depths	By default, Bacon will calculate the ages for the depths d.min to d.max in steps of d.by. Alternative depths can be provided as, e.g., depths=seq(0, 100, length=500) or as a file, e.g., depths=read.table("CoreDepths.txt". See also depths.file.
depth.unit	Units of the depths. Defaults to depth.unit="cm".
age.unit	Units of the ages. Defaults to age.unit="yr".
unit	Deprecated and replaced by depth.unit.
acc.shape	The prior for the accumulation rate consists of a gamma distribution with two parameters. Its shape is set by acc.shape (default acc.shape=1.5; higher values result in more peaked shapes).
acc.mean	The accumulation rate prior consists of a gamma distribution with two parameters. Its mean is set by acc.mean (default acc.mean=20 yr/cm (or whatever age or depth units are chosen), which can be changed to, e.g., 5, 10 or 50 for different kinds of deposits). Multiple values can be given in case of hiatuses or boundaries, e.g., Bacon(hiatus.depths=23, acc.mean=c(5,20))
mem.strength	The prior for the memory (dependence of accumulation rate between neighbouring depths) is a beta distribution, which looks much like the gamma distribution. but its values are always between 0 (no assumed memory) and 1 (100% memory). Its default settings of mem.strength=10 (higher values result in more peaked shapes) allow for a large range of posterior memory values. Please note that the default memory prior has been updated from rbacon version 2.5.1 on, to repair a bug.
mem.mean	The prior for the memory is a beta distribution, which looks much like the gamma distribution but its values are always between 0 (no assumed memory) and 1 (100% memory). Its default settings of mem.mean=0.5 allow for a large range of posterior memory values. Please note that the default memory prior has been updated from rbacon version 2.5.1. on, to repair a bug.
boundary	The assumed depths of any boundary, which divides sections of different accumulation rate regimes (e.g., as indicated by major change in the stratigraphy). No hiatus is assumed between these sections, and memory is reset crossing the boundary. Different accumulation priors can be set for the sections above and below the boundary, e.g., acc.mean=c(5, 20). See also hiatus.depths, mem.mean, acc.mean and acc.shape. Setting many boundaries might not work, and having more than one boundary per model section (see 'thick') might not work either.
hiatus.depths	The assumed depths for any hiatus should be provided as, e.g., hiatus.depths=20 for one at 20cm depth, and hiatus.depths=c(20,40) for two hiatuses at 20 and 40 cm depth.
hiatus.max	The prior for the maximum length of the hiatus. Hiatus length is a uniform distribution, with equal probabilities between 0 and hiatus.max yr (or whatever other age.unit is chosen).
add	Add a value to the maximum hiatus length if a boundary is chosen. Defaults to 100 yr (or whatever other age unit is chosen). Can be adapted if Bacon complains that the parameters are out of support.
after	Sets a short section above and below hiatus.depths within which to calculate ages. For internal calculations - do not change.
cc	Calibration curve for C-14 dates: cc=1 for IntCal20 (northern hemisphere terrestrial), cc=2 for Marine20 (marine), cc=3 for SHCal20 (southern hemisphere terrestrial). For dates that are already on the cal BP scale use cc=0.
cc1	For northern hemisphere terrestrial 14C dates (IntCal20).
cc2	For marine 14C dates (Marine20).
cc3	For southern hemisphere 14C dates (SHCal20).
cc4	Provide the name of an alternative curve (3 columns: cal BP, 14C age, error, separated by white spaces and saved as a plain-text file). It is important here to first produce a tailor-made folder for your and the default calibration curves to live in. See cc.dir. Defaults to cc4="mixed.14C".
cc.dir	Directory where the calibration curves for C14 dates cc are located. By default uses the location of the rintcal package which provides the calibration curves. If you want to use custom-made calibration curves, first set up a new folder using the function new.ccdir() in the rintcal package, e.g., new.ccdir="MyCurves", then place the custom curve in that folder using mix.ccurves(, cc.dir="MyCurves", save=TRUE).
postbomb	Use a postbomb curve for negative (i.e. postbomb) 14C ages. 0 = none, 1 = NH1, 2 = NH2, 3 = NH3, 4 = SH1-2, 5 = SH3
delta.R	Mean of core-wide age offsets (e.g., regional marine offsets).
delta.STD	Error of core-wide age offsets (e.g., regional marine offsets).
t.a	The dates are treated using the t distribution (Christen and Perez 2009) by default (normal=FALSE). This t-distribution has two parameters, t.a and t.b, set at 3 and 4 by default (see Christen and Perez, 2010). If you want to assign narrower error distributions (more closely resembling the normal distribution), set t.a and t.b at for example 33 and 34 respectively (e.g., for specific dates in your .csv file). For symmetry reasons, t.a must always be equal to t.b-1.
t.b	The dates are treated using t distribution by default (normal=FALSE). The t-distribution has two parameters, t.a and t.b, set at 3 and 4 by default (see Christen and Perez, 2009). If you want to assign narrower error distributions (more closely resembling the normal distribution), set t.a and t.b at for example 33 and 34 respectively (e.g., for specific dates in your .csv file). For symmetry reasons, t.a must always be equal to t.b-1.
normal	By default, Bacon uses the t-distribution to treat the dates. Use normal=TRUE to use the normal/Gaussian distribution. This will generally give higher weight to the dates.
suggest	If initial analysis of the data indicates abnormally slow or fast accumulation rates, Bacon will suggest to change the prior.
accept.suggestions	Automatically accept the suggested values. Use with care. Default accept.suggestions=FALSE. Also, if the length of the core would cause too few or too many sections with the default settings, Bacon will suggest an alternative section thickness thick. Accept these suggested alternative settings by typing "y" (or "yes please" if you prefer to be polite), or leave as is by typing "n" (or anything else, really). To get rid of these suggestions, use suggest=FALSE.
reswarn	Bacon will warn you if the number of sections lies outside the safe range (default between 10 and 200 sections; reswarn=c(10,200)). Too few sections could lead to an ‘elbowy’ model while with too many sections the modelling process can get lost, resulting in age-models far away from the dated depths.
remember	Bacon will try to remember which settings you have applied to your cores (default remember=TRUE). If you run into inconsistencies or other problems, try running your core again with remember=FALSE, or, start cleanly by typing Bacon.cleanup().
ask	By default Bacon will ask you to confirm that you want to run the core with the provided settings. Disable this using ask=FALSE (e.g., for batch runs).
run	In order to load an existing Bacon run instead of producing a new one, you can use run=FALSE.
defaults	Name of the file containing settings for the core. For internal use only - do not change.
sep	Separator between the fields of the plain text file containing the dating information. Default sep=",".
dec	Character for decimal points. Default to dec=".".
runname	Text to add to the corename for specific runs, e.g., runname="MyCore_Test1".
slump	Upper and lower depths of any sections of assumed abrupt accumulation, that require excising before age-modelling (and adding after age-modelling). Requires pairs of depths, e.g., slump=c(10,15,60,67) for slumps at 67-60 and 15-10 cm core depth.
remove	Whether or not to remove depths within slumps. Defaults to remove=FALSE.
BCAD	The calendar scale of graphs and age output-files is in cal BP (calendar or calibrated years before the present, where the present is AD 1950) by default, but can be changed to BC/AD using BCAD=TRUE.
ssize	The amount of iterations to store at the end of the MCMC run. Default 4000; decrease for faster (but less reliable) runs or increase for cores where the MCMC mixing (panel at upper-left corner of age-model graph) appears problematic.
th0	Starting years for the MCMC iterations. These are randomly chosen by default.
burnin	Amount of initial, likely sub-optimal MCMC iterations that will be removed.
youngest.age	Minimum age limit for Bacon runs, default at current year in cal BP. To set plot limits, use age.min instead.
oldest.age	Maximum age limit for Bacon runs, default at 1,000,000 cal BP. To set plot limits, use age.max instead.
MinAge	Deprecated - use youngest.age instead.
MaxAge	Deprecated - use oldest.age instead.
cutoff	Avoid plotting very low probabilities of date distributions (default cutoff=0.001).
plot.pdf	Produce a pdf file of the age-depth plot. Defaults to plot.pdf=TRUE after a Bacon run.
dark	Darkness of the greyscale age-depth model. The darkest grey value is dark=1 by default. Lower values will result in lighter grey but values >1 are not allowed.
date.res	Date distributions are plotted using date.res=100 segments by default.
age.res	Resolution or amount of greyscale pixels to cover the age scale of the age-model plot. Default yr.res=200.
yr.res	Deprecated - use age.res instead
close.connections	Internal option to close connections after a run. Default close.connections=TRUE.
save.info	By default, a variable called ‘info’ with relevant information about the run (e.g., core name, priors, settings, ages, output) is saved into the working directory. Note that this will overwrite any existing variable with the same name - as an alternative, one could run, e.g., myvar <- Bacon(), followed by supplying the variable myvar in any subsequent commands.
older.than	an option to enable dates at the limit of C-14 dating. If there are older.than dates, they tell us that the core should be older than a certain age at that depth. For example, if the 7th and 8th dates in the core's .csv file are older-than dates, use as older.than=c(7,8). The MCMC run could be problematic if the older-than ages do not fit with the other information.
younger.than	an option to provide younger-than ages, for example a historical pollen marker. If there are younger-than dates, they tell us that the core should be younger than a certain age at that depth. For example, if the 7th and 8th dates in the core's .csv file are younger.than dates, use as younger.than=c(7,8). The MCMC run could be problematic if the younger.than ages do not fit with the other information.
save.elbowages	If you want to have a file with the MCMC-derived ages for all the age-depth model's elbows, set save.elbowages=TRUE and a file with the ages will be saved in the core's folder, starting with the core name, followed by its number of sections, d.min, and section thickness, and ending in "_elbowages.txt".
verbose	Provide feedback on what is happening (default verbose=TRUE).
...	options for the age-depth graph. See agedepth and calib.plot

Details

Bacon is an approach to age-depth modelling that uses Bayesian statistics in order to reconstruct Bayesian accumulation histories for deposits, through combining radiocarbon and other dates with prior information ('Blaauw' and 'Christen', 2011).

Bacon divides a core into many thin vertical sections (by default of thick=5 cm thickness), and through millions of Markov Chain Monte Carlo (MCMC) iterations estimates the accumulation rate (in years/cm; so more correctly, sedimentation times) for each of these sections. Combined with an estimated starting date for the first section, these accumulation rates then form the age-depth model. The accumulation rates are constrained by prior information on the accumulation rate (acc.mean, acc.shape) and its variability between neighbouring depths, or "memory" (mem.mean, mem.strength). Hiatuses can be introduced as well, also constrained by prior information (hiatus.max).

Although Bacon works with any kind of absolute dates (e.g., OSL, tephra or other dates on a calendar scale), it is often used to age-model 14C-dated sequences. Radiocarbon dates should be calibrated using either IntCal20 (for terrestrial northern hemisphere material; Reimer et al., 2020), Marine20 (for marine dates; Hughen et al., 2020), SHCal20 (for southern hemisphere dates; Hogg et al., 2020) or any other calibration curve (see below), while modern 14C dates are calibrated using one of the post-bomb calibration curves (NH1, NH2 or NH3 for the northern hemisphere, SH1-2 or SH3 for the southern hemisphere; Hua et al., 2022). See http://calib.org/CALIBomb/ if you are unsure which postbomb curve you need. If Bacon finds postbomb dates (negative 14C ages) and you haven't specified a postbomb curve, you will be prompted. Provide postbomb curves as, e.g., postbomb=1 for the NH1 postbomb curve (2 for NH2, 3 for NH3, 4 for SH1-2, 5 for SH3). For calendar dates, i.e. dates that are already on the calendar scale and thus should not be calibrated, setcc=0.

Since version 3.1.0, rbacon can also handle younger-than and older-than ages, with the model aiming to either go 'above' or 'below' such dates as requested. If the resulting combination of parameters becomes problematic (e.g., no initial combination of parameters can be found that obeys the priors or is in chronological order), then the output will often be wrong. If so, using the function set.initvals could help.

By default, the initial MCMC values of the Bacon age-depth model (upper ages and accumulation rate for each model section) are estimated randomly. Since version 3.1.0, these starting values can also be provided in a file with extension _bacon.init, placed within the core's folder. This file will need to have two rows, each for one of the two initial sets of parameters required (the t-walk requires two starting estimates for all MCMC parameters). If such a file is found (and correctly formatted), Bacon will use the values within this file as starting points for the MCMC run. See function set.initvals for more information.

From version 2.5.1 on (i.e., since February 2021), the default memory prior has changed to mem.mean=0.5 and mem.strength=10. Previously used c++ code contained a bug which caused the prior information for the memory not to be taken into account correctly. Now that this bug has been repaired, the default memory prior has been updated such that it should work for most types of cores, and should result in similar output to previous versions of Bacon. There is no need to re-do previous runs. However, it is considered good practice to test the impact of different settings on a site's age-depth model (e.g., thick, acc.mean, acc.shape, mem.mean, acc.strength).

Value

An age-depth model graph, its age estimates, and a summary.

Author(s)

Maarten Blaauw, J. Andres Christen

References

Blaauw, M. and Christen, J.A., 2011. Flexible paleoclimate age-depth models using an autoregressive gamma process. Bayesian Anal. 6(3): 457-474.

Christen, J.A., Perez E., S., 2010. A new robust statistical model for radiocarbon data. Radiocarbon 51: 1047-1059.

Reimer et al., 2020. The IntCal20 Northern Hemisphere radiocarbon age calibration curve (0–55 cal kBP). Radiocarbon 62(4): 725-757. \Sexpr[results=rd]{tools:::Rd_expr_doi("10.1017/RDC.2020.41")}

Hogg et al. 2020 SHCal20 Southern Hemisphere calibration, 0-55,000 years cal BP. Radiocarbon 62(4): 759-778. \Sexpr[results=rd]{tools:::Rd_expr_doi("10.1017/RDC.2020.59")}

Hughen et al. 2020 Marine20-the marine radiocarbon age calibration curve (0-55,000 cal BP). Radiocarbon 62(4): 779-820. \Sexpr[results=rd]{tools:::Rd_expr_doi("10.1017/RDC.2020.68")}

Hua et al. 2022 Atmospheric radiocarbon for the period 1950-2019. Radiocarbon 64(4), 723-745, \Sexpr[results=rd]{tools:::Rd_expr_doi("10.1017/RDC.2021.95")}

Jones, V.J., Stevenson, A.C., Battarbee, R.W., 1989. Acidification of lakes in Galloway, south west Scotland - a diatom and pollen study of the post-glacial history of the Round Loch of Glenhead. Journal of Ecology 77: 1-23.

Examples

  Bacon(ask=FALSE, coredir=tempfile())
  Bacon(cc=2, delta.R=80, delta.STD=40, coredir=tempfile())

rbacon documentation built on Sept. 30, 2024, 9:40 a.m.