isoWater (version 1.2.1)

mixSource: Water Source as a Mixture

Description

Given isotopic compositions of two or more potential sources, generate a posterior sample of source mixtures conditioned on one or more sample values.

Usage

mixSource(obs, sources, slope, prior=rep(1,nrow(sources)), 
  shp = 1, edist = "unif", eprior = NULL, ngens = 1e5, ncores = 1)

Value

Returns an object of class “mixSource”, a list containing:

summary

matrix. Summary table of JAGS MCMC results, including parameter posterior distributions and convergence statistics.

results

data.frame. Posterior samples of model parameters.

mixture_d2H

Hydrogen isotopic composition of unevaporated source mixture.

mixture_d18O

Oxygen isotopic composition of unevaporated source mixture.

sX_fraction

Fractional contribution of each source.

S

Evaporation line slope.

E

Evaporation index, equal to the difference between the sample and unevaporated source mixture oxygen isotope values.

Arguments

obs

iso object containing isotope values for one or more samples.

sources

iso object containing isotope values for two or more sources which may have contributed to the observed samples.

slope

numeric. Vector of length two specifying prior parameters for the evaporation line slope (mean, standard deviation).

prior

numeric. Vector of length equal to the number of sources, giving prior estimates of relative contributions of different sources.

shp

numeric. Shape parameter constant used in specifying prior estimates of source contributions (see Details).

edist

character. One of "unif" or "gamma", specifying whether the evaporation prior is modeled using a uniform or gamma distribution.

eprior

numeric. Vector of length 2 giving prior parameter estimates for the oxygen isotope evaporation effect. For edist = "unif" these are maximum and minimum values and the default values are c(0, 15). For edist = "gamma" these are shape and rate parameters, and the defaults are c(1, 1).

ngens

integer. Number of posterior samples to obtain (per chain).

ncores

integer. Number of cores to use for parallel processing.

Details

The Dirichlet distribution is used to represent the fractional contribution of each source. The prior estimate is a Dirchlet where the shape parameter for source i is given by prior[i] / min(prior) * shp.

If ncores = 1, three chains will be run on a single core. If ncores > 1, ncores chains will be run in parallel on ncores cores.

Examples

#Prep sources
O = runif(3, -15, -2)
H = O * 8 + 10 + rnorm(3, 0, 6)
sources = iso(H, O, 1, 0.2, 0.17)

#Sample data
obs = iso(-60, -6, 0.5, 0.1, 0)

#Evaporation slope
slope = c(5, 0.3)

#Run and report...likely not converged!
ws = mixSource(obs, sources, slope, ngens = 1e3)
ws$summary

#A traceplot
plot(ws$results$mixture_d18O[1:1000], type = "l")
lines(ws$results$mixture_d18O[1001:2000], col = 2)
lines(ws$results$mixture_d18O[2001:3000], col = 3)