sadverif {sad}

dual-tree verification

Description

verify the scale, anisotropy and direction of a number of forecasts

Usage

sadverif(x, dec = TRUE, xmin = 0.1, log = TRUE, a = 1, nbr = 33,
  rsm = 0, Nx = NULL, Ny = NULL, J = NULL, boundaries = "pad",
  return_specs = FALSE)

## S3 method for class 'sadverif'
plot(x, ...)

## S3 method for class 'sadverif'
summary(object, ...)

Arguments

x

a list of equally sized matrices, the first element is assumed to be the observation

dec

logical, do you want to use the decimated transform

xmin

values smaller than xmin are set to zero

log

logical, do you want to log-transfrom the data? (recommended for precipitation)

a

relative weight of directional errors compared to scale errors in semdd

nbr

number of breaks for the scale histograms, has no effect if dec=TRUE

rsm

number of pixels which are linearly smoothed at the edge

Nx

size to which the data is extended in x-direction

Ny

size to which the data is extended in y-direction

J

largest scale considered

boundaries

how to handle the boundary conditions, either "pad", "mirror" or "periodic"

return_specs

if TRUE, the spatial mean spectra are returned as well

...

further arguments, currently ignored.

object

object of class sadverif

Details

each element of x is transformed via dtcwt from the 'dualtrees' package. Scores and centres based on the mean spectra are calculated. If dec=FALSE, scale histograms and the corresponding score hemd are calcualted as well.

Value

an object of class sadverif, containing the following elements

settings

a dataframe containing the parameters that were originally passed to dtverif

centres

a matrix cotaining the anisotropy rho, angle phi and central scale z derived from the mean spectra. Rain area and sum are included as well.

detscores

a matrix containing the differences in centre components, the direction/anisotropy score dxy, the emd between direction-averaged spectra (semd) and the emd between the directional spectra (semdd). If dec=FALSE, the emd between the scale histograms, hemd, is included as well.

time

the time the calculation took in seconds

if there is more than one forecast, the ensemble scores SpEn and (if available), hemd are computed as well, treating all forecasts as members of the ensemble to be verified.

References

Selesnick, I.W., R.G. Baraniuk, and N.C. Kingsbury (2005) <doi:10.1109/MSP.2005.1550194> Buschow et al. (2019) <doi:10.5194/gmd-12-3401-2019> Buschow and Friederichs (2020) <doi:10.5194/ascmo-6-13-2020>

Examples

oldpar <- par(no.readonly=TRUE)
on.exit(par(oldpar))
data(rrain)
ra <- as.sadforecast( list( rrain[1,1,,], rrain[1,2,,], rrain[2,1,,], rrain[3,1,,] ) )
plot(ra)
verif <- sadverif( ra, log=FALSE, xmin=0 )
summary(verif)
par( mfrow=c(2,2) )
plot( verif )