Skip to contents

This is the main function to execute the machine learning training and prediction. Note it will be deployed by each node you requested if the qsub feature is enabled below.

Usage

downscale_main(
  settings,
  analysis,
  covariates.dir,
  time,
  variable,
  outdir,
  base.map.dir,
  method = "randomForest",
  cores = parallel::detectCores()
)

Arguments

settings

character: physical path that points to the pecan settings XML file.

analysis

numeric: data frame (rows: ensemble member; columns: site*state_variables) of updated ensemble analysis results from the `sda_enkf` function.

covariates.dir

character: path to the exported covariates GeoTIFF file.

time

character: the time tag used to differentiate the outputs from others.

variable

character: name of state variable. It should match up with the column names of the analysis data frame.

outdir

character: the output directory where the downscaled maps will be stored.

base.map.dir

character: path to the GeoTIFF file within which the extents and CRS will be used to generate the ensemble maps.

method

character: machine learning method, default is randomForest (currently support randomForest and xgboost).

cores

numeric: how many CPus to be used in the calculation, the default is the total CPU number you have.

Value

paths to the ensemble downscaled maps.

Author

Dongchen Zhang