In nickmckay/GeoChronR: Tools to Analyze and Visualize Time-Uncertain Data

knitr::opts_chunk$set(fig.width = 8,fig.height = 6) 
knitr::opts_knit$set(progress = FALSE,verbose = FALSE)

1. Introduction to geoChronR
2. Age-uncertain correlation
3. Age-uncertain regression and calibration-in-time
4. Age-uncertain spectral analysis
5. Age-uncertain PCA analysis

Ensemble Regression and Calibration-in-time

Here, we replicate the analysis of Boldt et al. (2015), performing age-uncertain calibration-in-time on a chlorophyll reflectance record from northern Alaska, using geoChronR.

The challenge of age-uncertain calibration-in-time is that age uncertainty affects both the calibration model (the relation between the proxy data and instrumental data) and the reconstruction (the timing of events in the reconstruction). geoChronR simplifies handling these issues.

Let's start by loading the packages we'll need.

library(lipdR) #to read and write LiPD files
library(geoChronR) #of course
library(readr) #to load in the instrumental data we need
library(ggplot2) #for plotting

Load the LiPD file

OK, we'll begin by loading in the Kurupa Lake record from Boldt et al., 2015. The system.file(...) part of this pulls the example file from the package directory. You'd like just enter the path as a string for typical use.

K <- lipdR::readLipd("http://lipdverse.org/geoChronR-examples/Kurupa.Boldt.2015.lpd")

Check out the contents

sp <- plotSummary(K,paleo.data.var = "RABD",summary.font.size = 6)
print(sp)

Create an age model with Bacon

K <- runBacon(K,
              lab.id.var = 'labID', 
              age.14c.var = 'age14C', 
              age.14c.uncertainty.var = 'age14CUncertainty',
              age.var = 'age', 
              age.uncertainty.var = 'ageUncertainty', 
              depth.var = 'depth', 
              reservoir.age.14c.var = NULL, 
              reservoir.age.14c.uncertainty.var = NULL, 
              rejected.ages.var = NULL,
              bacon.acc.mean = 10,
              bacon.thick = 7,
              ask = FALSE,
              bacon.dir = "~/Cores",
              suggest = FALSE,
              close.connection = FALSE)

And plot the ensemble output

plotChron(K,age.var = "ageEnsemble",dist.scale = 0.2)

Map the age ensemble to the paleodata table

This is to get ensemble age estimates for each depth in the paleoData measurement table

K <- mapAgeEnsembleToPaleoData(K,age.var = "ageEnsemble")

Select the paleodata age ensemble, and RABD data that we'd like to regress and calibrate

kae <-  selectData(K,"ageEnsemble")
rabd <- selectData(K,"RABD")

Now load in the instrumental data we want to correlate and regress agains

kurupa.instrumental <- readr::read_csv("http://lipdverse.org/geoChronR-examples/KurupaInstrumental.csv")

Check age/time units before proceeding

kae$units

yep, we need to convert the units from BP to AD

kae <- convertBP2AD(kae)

Create a "variable list" for the instrumental data

kyear <- list()
kyear$values <- kurupa.instrumental[,1]
kyear$variableName <- "year"
kyear$units <- "AD"

kinst <- list()
kinst$values <- kurupa.instrumental[,2]
kinst$variableName <- "Temperature"
kinst$units <- "deg (C)"

Calculate an ensmeble correlation between the RABD and local summer temperature data

corout <- corEns(kae,rabd,kyear,kinst,bin.step=2,percentiles = c(.05,.5,.95 ))

And plot the output

Note that here we use the "Effective-N" significance option as we mimic the Boldt et al. (2015) paper.

plotCorEns(corout,significance.option = "eff-n")

Mixed results. But encouraging enough to move forward.

Perform ensemble regression

OK, you've convinced yourself that you want to use RABD to model temperature back through time. We can do this simply (perhaps naively) with regession, and lets do it with age uncertainty, both in the building of the model, and the reconstructing

regout <- regressEns(time.x = kae,
                    values.x = rabd,
                    time.y =kyear,
                    values.y =kinst,
                    bin.step=3,
                    gaussianize = FALSE,
                    recon.bin.vec = seq(-4010,2010,by=20))

And plot the output

regPlots <- plotRegressEns(regout,alp = 0.01,font.size = 8)

This result is consistent with that produced by Boldt et al., (2015), and was much simpler to produce with geoChronR.

In the next vignette learn about spectral analysis in geoChronR.

```

nickmckay/GeoChronR documentation built on April 9, 2024, 5:26 a.m.