EBASE (version 1.1.0)

ebase_years: Estuarine Bayesian Single-station Estimation method for ecosystem metabolism for long time series

Description

Estuarine Bayesian Single-station Estimation method for ecosystem metabolism for long time series

Usage

ebase_years(
  dat,
  Z,
  interval,
  ndays = 1,
  aprior = c(4, 2),
  rprior = c(300, 150),
  bprior = c(0.251, 0.125),
  bmax = 0.502,
  nogas = FALSE,
  doave = TRUE,
  maxinterp = 43200/interval,
  n.iter = 10000,
  update.chains = TRUE,
  n.burnin = n.iter * 0.5,
  n.chains = 3,
  n.thin = 10,
  model_file = NULL,
  ncores = NULL,
  quiet = TRUE,
  maxtry = 5
)

Value

Output identical to that returned by ebase, where the results for each year are appended to the data frame as the function progresses through the years. Note that the grp column that specifies the optimization period defined by ndays is unique to each year, e.g., values will be repeated across years.

Arguments

dat

input data frame

Z

numeric as single value for water column depth (m) or vector equal in length to number of rows in dat

interval

timestep interval in seconds

ndays

numeric for number of days in dat for optimizing the metabolic equation, see details

aprior

numeric vector of length two indicating the mean and standard deviation for the prior distribution of the a parameter, see details

rprior

numeric vector of length two indicating the mean and standard deviation for the prior distribution of the R parameter, see details

bprior

numeric vector of length two indicating the mean and standard deviation for the prior distribution of the b parameter, see details

bmax

numeric value for the upper limit on the prior distribution for bprior, set as twice the default value of the mean

nogas

logical indicating if gas exchange is not included in the metabolic model, see details

doave

logical indicating if the average dissolved oxygen concentration is used as the starting value for the estimation (default), otherwise the first observation will be used if FALSE, see details

maxinterp

numeric value for minimum number of continuous observations that must not be interpolated within a group defined by ndays to assign as NA in output, see details

n.iter

number of MCMC iterations, passed to jags

update.chains

logical to run metab_update if chains do not converge

n.burnin

number of MCMC chains to delete, passed to jags

n.chains

number of MCMC chains to run, passed to jags

n.thin

number of nth iterations to save for each chain, passed to jags

model_file

NULL to use the model file included with the package or a path to a model text file can be used

ncores

numeric for number of cores to use for parallel processing, use NULL to suppress

quiet

logical to suppress progress messages to the console

maxtry

numeric for maximum number of times to retry the model if it fails

Details

ebase is run for each year in the supplied data. This facilitates running ebase on long time series by running the model sequentially on each year of data, with progress messages printed to the console if quiet = FALSE. The model run for each year will restart if it fails, up to maxtry times, after which processing continues with the next year. The model is run in parallel using the number of cores used set by ncores. If ncores = NULL, sequential processing is used. All other arguments are passed to ebase.

Similar results can be obtained by running ebase on the entire data set, but this function is useful for long time series where the model may fail for some years, e.g., when weather data may be missing for some years.

Examples

# get one day of data
dat <- exdat[as.Date(exdat$DateTimeStamp, tz = 'America/Jamaica') == as.Date('2012-06-01'), ]

# run ebase, use more chains and iterations for a better fit, update.chains as T
ebase_years(dat, Z = 1.85, interval = 900, n.chains = 2, n.iter = 50, 
 update.chains = FALSE)