Skip to contents

Calibrate models

Source: R/cali_ensemble.R
cali_ensemble.Rd

Use one of three methods to calibrate specified models.

Usage

cali_ensemble(
  config_file,
  num = NULL,
  param_file = NULL,
  cmethod = "LHC",
  qualfun = qual_fun,
  parallel = FALSE,
  job_name,
  model = c("FLake", "GLM", "GOTM", "Simstrat", "MyLake"),
  folder = ".",
  spin_up = NULL,
  out_f = "cali",
  ncores = NULL,
  tmp_dir = NULL,
  ...
)

Arguments

config_file

file path; to LakeEnsemblr yaml master config file

num

integer; the number of random parameter sets to generate. If param file is provided num = number of parameters in that file.

param_file

file path; to previously created parameter file set. If NULL creates a new parameter set. Defaults to NULL

cmethod

character; Method for calibration. Can be "LHC", "LHC_old, "MCMC" or "modFit". Defaults to "LHC". LHC and LHC_old only differ in the way the parallelization is set up, whereas the new and default LHC version is more efficient and LHC_old is only kept for possible backwards compatibility.

qualfun

function; function that calculates measure of fit from observed and simulated variables, takes the two arguments Observed and Simulated

parallel

Boolean; should the model calibration be parallelized

job_name

character; optional name to use as an RStudio job and as output variable name. It has to be a syntactically valid name. Check out this webpage for more info on jobs: https://blog.rstudio.com/2019/03/14/rstudio-1-2-jobs/

model

vector; model to export driving data. Options include c("GOTM", "GLM", "Simstrat", "FLake", "MyLake")

folder

file path; to folder which contains the model folders generated by export_config()

spin_up

numeric; Number of days to disregard as spin-up for analysis.

out_f

character; name of the folder to store results into

ncores

numeric; number of cores to be used. If NULL, will default to `parallel::detectCores() - 1`.

tmp_dir

location where the temporary files for LHC calibration in parallel are stored

...

additional arguments passed to FME::modFit or FME::modMCMC. Only used when method is modFit or MCMC

Details

Parallelisation is done using the `parallel` package and `parLapply()`. The number of cores used is set to the value specified in `ncores`.

Examples

if (FALSE) { # \dontrun{

config_file <- 'LakeEnsemblR.yaml'

# LHC method
cali_ensemble(config_file = config_file, num = 200, cmethod = "LHC",
             model = c("FLake", "GLM", "GOTM", "Simstrat", "MyLake"))

# MCMC method
resMCMC <- cali_ensemble(config_file = config_file, num = 200, cmethod = "MCMC",
                         model = c("FLake", "GLM", "GOTM", "Simstrat", "MyLake"))

# modFit method using the Nelder-Mead algorithm
resMmodFit <- cali_ensemble(config_file = config_file, num = 200, cmethod = "modFit",
                            model = c("FLake", "GLM", "GOTM", "Simstrat", "MyLake"),
                            method = "Nelder-Mead")

# LHC method using multiple cores
cali_ensemble(config_file = config_file, num = 200, cmethod = "LHC",
             model = c("FLake", "GLM", "GOTM", "Simstrat", "MyLake"),
             parallel = TRUE)

# LHC method deployed as a job
cali_ensemble(config_file = config_file, num = 200, cmethod = "LHC",
             model = c("FLake", "GLM", "GOTM", "Simstrat", "MyLake"),
             job_name = "test")
test # View output from job

} # }