Run Particle Swarm Optimizations of the HBV Model using hydroPSO
and TUWmodel by importing configurations from R files. Supports
specifiying multiple configuration files to run optimizations in batches.
hbv_pso_run(configpath, recursive = FALSE, suffix = NULL, gof.name = NULL, from_validation = NULL, to_validation = NULL, ...)
| configpath | character vector of full path and/or file names. Any file
given is sourced and an |
|---|---|
| recursive | logical, should config files be searched in subdirectories?
If true, |
| suffix | String, will be appended to the optimization run identifier |
| gof.name | String, Used as part of the optimization run identifier. Shorthand for setting plotting$gof.name. If gof.name is not set directly, the value from plotting$gof.name is used. |
| from_validation | Date or date string in default format, start of validation run with best parameter set found during optimization |
| to_validation | Date or date string in default format, end of validation period |
| ... | Parameters passed to hbv_pso, will overwrite those set in in the configuration files. |
A list of the following items:
summary data
frame with id, gof, gof_valid, from, to, from_valid and to_valid for each
optimization run
results list with hydro_pso output from all
optimization runs, named by id
In addition, the output files from each optimization run are written inside a directory which is located in the same directory as the configuration file, and named by the identifier of the respective optimization run.
Runs Particle Swarm Optimizations of HBV for given parameter sets defined in simple R scripts. Timeseries input data (prec, airt, ep, obs, area, elevation_zones) can be automatically imported from csv or HBV-light format.
Each optimization run is given an identification string concatenated from
the name of the input data directory (i.e. the location of the
configuration file), the name of the gof variable and an optional suffix.
This identifier is used as directory name for the output files, in plot
titles and file names, and in the result list returned by hbv_pso_run.
Parameters for the different optimization runs
are defined by simple variable assignment, where the variable name
corresponds to the parameter as used in hbv_pso(). They are defined in
simple R files passed to hbv_pso_run (see configpath). In order for the
configuration files to be picked up when passing directory paths to
hbv_pso_run, the configuration file names must start with "run".
Parameters can also be set by passing them as additional arguments to
hbv_pso_run, which will overwrite corresponding values defined in
configuration file(s).
Be aware that when using this option, any lists passed are not merged but
completely replace the values set in the configfile.
In addition to parameters described in hbv_pso(), the following optional
parameters unique to hbv_pso_run can also be set in configuration files:
suffix, gof.name,from_validation, to_validation (see above).
For each optimization run, the following
variables have to be defined and contain valid values as defined in
hbv_pso(): prec, airt, ep, obs, area, elev_zones. The input data
can be assinged in the following ways (in descending order of precedence):
assignment inside the configfile: Add code which assigns the input
text files inside input data directories:
The files must be named as the input time series (with either
.csv, .txt or no suffix, e.g. prec.csv) and be located in the same
directory as the configfile. They must produce a valid zoo object with
index when read by read.zoo with default arguments
HBV-light input files: Analogue to the text files above, input data can also be imported from files as used by HBV-light ("ptq.text" and "ep.txt")
If a given input data set is not defined in the configuration file,
hbv_pso_run will try to load it from a corresponding file in the input data
directory - first from a csv file, and then from the corrseponding HBV-light
file (e.g. PTQ.txt for prec, airt or obs).