Skip to contents

predict_subdaily_met

Source: R/tdm_predict_subdaily_met.R
predict_subdaily_met.Rd

This is the main function of the tdm family workflow. This function predicts subdaily meteorology from daily means using a linear regression modeling approach. It takes a dataset with daily resolution and temporally downscales it to hourly resolution using the statistics generated by gen.subdaily.models(). It references the predict.subdaily.function located in lm_ensemble_sims() which uses a linear regression based approach to downscale. We generate multiple ensembles of possible hourly values dictated from the models and betas generated in gen.subdaily.models. Each ensemble member is saved as a netCDF file in CF conventions and these files are ready to be used in the general PEcAn workflow.

Usage

predict_subdaily_met(
  outfolder,
  in.path,
  in.prefix,
  path.train,
  direction.filter = "forward",
  lm.models.base,
  yrs.predict = NULL,
  ens.labs = 1:3,
  resids = FALSE,
  adjust.pr = 1,
  force.sanity = TRUE,
  sanity.tries = 25,
  overwrite = FALSE,
  verbose = FALSE,
  seed = format(Sys.time(), "%m%d"),
  print.progress = FALSE,
  ...
)

Arguments

outfolder

- directory where output file will be stored

in.path

- base path to dataset you wish to temporally downscale; Note: in order for parallelization to work, the in.prefix will need to be appended as the final level of the file structure. For example, if prefix is GFDL.CM3.rcp45.r1i1p1, there should be a directory with that title in in.path.

in.prefix

- prefix of model dataset, i.e. if file is GFDL.CM3.rcp45.r1i1p1.2006 the prefix is 'GFDL.CM3.rcp45.r1i1p1'

path.train

- path to CF/PEcAn style training data where each year is in a separate file.

direction.filter

- Whether the model will be filtered backward or forwards in time. options = c("backward", "forwards") (default is forward; PalEON will go backward, anybody interested in the future will go forwards)

lm.models.base

- path to linear regression model folders generated using gen.subdaily.models

yrs.predict

- years for which you want to generate met. if NULL, all years in in.path will be done

ens.labs

- vector containing the labels (suffixes) for each ensemble member; this allows you to add to your ensemble rather than overwriting with a default naming scheme

resids

- logical stating whether to pass on residual data or not

adjust.pr

- adjustment factor fore precipitation when the extracted values seem off

force.sanity

- (logical) do we force the data to meet sanity checks?

sanity.tries

- how many time should we try to predict a reasonable value before giving up? We don't want to end up in an infinite loop

overwrite

logical: replace output file if it already exists?

verbose

logical: should ncdf4 functions print debugging information as they run?

seed

- manually set seed for results to be reproducible

print.progress

- print the progress bar?

...

further arguments, currently ignored

Details

Predict Subdaily Meteorology Predict Subdaily Meteorology based off of statistics created in gen.subdaily.models()

See also

Other tdm - Temporally Downscale Meteorology: gen.subdaily.models(), lm_ensemble_sims(), model.train(), nc.merge(), save.betas(), save.model(), subdaily_pred(), temporal.downscale.functions()

Author

Christy Rollinson, James Simkins

Examples

if (FALSE) { # \dontrun{
library(PEcAn.data.atmosphere)
outfolder = '~/Downscaled_GCM'
in.path = '~/raw_GCM'
in.prefix = 'GFDL'
lm.models.base = 'sf_scratch/US-WCr'
dat.train_file = 'Training_data/US-WCr_dat.train.nc'
start_date = '2010-01-01'
end_date = '2014-12-31'
cores.max = 12
n.ens = 3} # }