REVEALSinR

Introduction

The REVEALS model, introduced by Shinya Sugita in 2007,  aims to translate pollen data from large sites into regional vegetation composition. The model applies pollen productivity estimates (PPEs) to account for the productivity bias and a ‘dispersal and deposition factor’ K to account for the dispersal bias in pollen data. K is calculated with a dispersal model.  REVEALS does not account for the homogeneity bias; we thus have to assume that the vegetation of the pollen source area has been homogeneous. 

The REVEALSinR function is an implementation of the REVEALS model (Theuerkauf et al. 2016). By default, it uses a Lagrangian stochastic dispersal model (Kuparinen et al. 2007) to calculate the K factor for each taxon. Further options are available. To arrive at error estimates, model runs are repeated at least 1000 times with random error added in pollen data and PPEs during each model run. To account for pollen deposition in lakes, a lake model is included. For details see the documentation of the DISQOVER package.

Application

Using the REVEALS model with REVEALSinR requires a single function call. Model runs require two files, the pollen data file and the parameter file with taxon specific parameters. By default, both files should be provided as comma separated text files (csv). Alternatively, spread sheet files (‘xls’ or ‘xlsx’) may be used. For file formatting, see the example files from our publication.

Requiered parameters:

  • pollen: pollen counts
  • params: fall speed of pollen (in m s-1) and relative pollen productivity (PPE or RPP) for each taxon
  • tBasin: basin type, either ‘lake’ or ‘peatland’
  • dBasin:  basin diameter in meter

Optional parameters:

  • dwm (default = lsm_unstable): Distance weighting model. By default the Lagrangian Stochastic model (Kuparinen et al. 2007) adjusted for unstable atmospheric conditions is used. Further options are available.
  • repeats (default = 1e3): Model runs per time slice
  • regionCutoff (default = 1e5): Diameter of the reconstruction region in meters
  • ppefun (default = rnorm_reveals): Function to randomise PPEs
  • pollenfun (default = rmultinom_reveals): Function to randomise pollen counts
  • writeresults (default = TRUE): Bboolean, indicats wether to write results to csv-files or to return them on the command-line
  • verbose (default = TRUE): Boolean, indicating whether to print progress messages

Output:

  • for each sample, mean, median, standard deviatian, 10% and 90% quantile of the repeated model runs
  • results of each single model run

 

Example function calls

To explore REVEALSinR you may use example data from our Lake Tiefer See study. The example includes the pollen data file and the parameter file in three versions (English/US csv format, German csv format and Excel format) and example R code to run REVEALSinR with these three file versions.

 
## Application on a core section from Lake Tiefer See:
data("Tiefer_See")
cover <- REVEALSinR(pollenTS, paramsTS, dwm='lsm unstable', tBasin='lake', dBasin=600)

## Application on a single sample:
data("Tiefer_See")
abc<-REVEALS(t(pollenTS[1,,drop=FALSE]), PPEs=paramsTS[,c(1,3:4)],rep=10,write=FALSE,verbose=FALSE,depo=do.call(rbind,lapply(paramsTS[,2], DispersalFactorK, dwm='lsm unstable', tBasin='peatland', dBasin=1e5, regionCutoff=1e5)))
matplot(sapply(abc,function(x)x),type="l")

## Inspect the sample data
data("Tiefer_See")
pollenTS
paramsTS

## Load pollen data from an Excel file (e.g.'Pollen_Tiefer_See_1880-2010.xlsx')
datafile <- 'Pollen_Tiefer_See_1880-2010.xlsx' # change working directory if needed
pollen <- do.call(read_excel,args=c(list(datafile,sheet = 1))) #function reads first worksheet

## Load parameters from an Excel file (e.g.'pf_PPE_MV2015.xlsx')
datafile <- 'pf_PPE_MV2015.xlsx' # change working directory if needed
params <- do.call(read_excel,args=c(list(datafile,sheet = 1))) #function reads first worksheet