evaluation: Model statistical evaluation

In Schuch666/hackWRF: Functions to Work with WRF, NetCDF and Model Evaluation

Model statistical evaluation

Description

Statistical evaluation from 2 data.frames. The input data.frames (model and observation) must contain a date column (with POSIXlt data) and also contains the same station (or pollutant) name. The function test and combine the time pairs and perform some basic tests. If a data.frame is provided to table argument a crbind is performed with a new row.

Usage

evaluation(
  mo,
  ob,
  station = "ALL",
  fair = NULL,
  table = NULL,
  wd = FALSE,
  cutoff = NA,
  cutoff_NME = NA,
  no_tz = FALSE,
  nobs = 8,
  clean = FALSE,
  formate = FALSE,
  ndig = 2,
  summaryze = FALSE,
  use_n = FALSE,
  NAME = "AVERAGE",
  verbose = TRUE,
  ...
)

Arguments

mo	model data.frame
ob	observed data.frame
station	name of the station or ALL, see notes
fair	model data.frame (or list of names) to perform a fair comparison, see notes
table	a data.frame with output from evaluate or stats
wd	default is FALSE, see notes
cutoff	minimum (optionally the maximum) valid value for observation
cutoff_NME	minimum (optionally the maximum) valid value for observation for NME
no_tz	ignore tz from input
nobs	minimum number of valid observations, default is 8
clean	remove rows with zero observations
formate	format the output for 2 digit (default)
ndig	number of digits for formate
summaryze	add a last line with the the average values and format the table
use_n	only for summaryze = TRUE, use n as weight to calculate the average
NAME	row.name for summaryze option
verbose	display additional information
...	arguments to be passing to stats and plot

Note

for wind direction some the ME and MB are calculated using Mughal et al. (2017)

station == 'ALL' make the function put all observations and model together, for this option a additional data.frame (or character containging the station names) can be used to perform a fair comparison, considering only stations (ie columns) in the fair data.frame and ob data.frame (or the name list).

Special thanks to Kiarash and Libo to help to test the wind direction option.

References

Mughal MO, Lynch M, Yu F, McGann B, Jeanneret F, Sutton J (2017) Wind modeling, validation and sensitivity study using Weather Research and Forecasting model in complex terrain. Environ Model Softw 90:107–125. https://doi.org/10.1016/j.envsoft.2017. 01.009

Examples

model <- readRDS(paste0(system.file("extdata",package="hackWRF"),"/model.Rds"))
obs   <- readRDS(paste0(system.file("extdata",package="hackWRF"),"/obs.Rds"))

# if first a test with no observed data
# the function return an empty row
table <- evaluation(mo = model, ob = obs, station = "VVIbes")
print(table)

# now a test with a few observed values
table <- evaluation(mo = model, ob = obs, station = "Americana", table = table)
print(table)

# new tests with no data will be discated
table <- evaluation(mo = model, ob = obs, station = "VVIbes", table = table)
print(table)

# if the station are not in the input data frame a message is displayed
# and the function return an empty row
table <- evaluation(mo = model, ob = obs, station = "Ibirapuera", table = table)
print(table)

# if the first evaluation has no data, the last call can remove the line
table <- evaluation(mo = model, ob = obs, station = "Americana", table = table, clean = TRUE)
print(table)

Schuch666/hackWRF documentation built on June 9, 2025, 1:46 p.m.