int.scale.verify {verification} | R Documentation |
For a spatial forecast, evaluates the forecast skill as a function of precipitation rate intensity and spatial scale of the error.
int.scale.verify(frcs, obs, thres = quantile(frcs, p = seq(0,0.9,0.1)), ... )
frcs |
Forecast matrix. Must be of $2^n$ dimensions. |
obs |
Observation matrix. Must be of $2^n$ dimensions. |
thres |
A vector of thresholds to be considered. By default, the percentiles 0, 90 are used. |
... |
Optional arguments may be passed to the image plot |
SSul |
Skill score as matrix. The rownames are the thresholds, the colnames are $n$ where $2^n$ is the spatial scale of the skill score decomposition. |
MSE |
A matrix with the mean squared error of the forecast |
l.frcs |
Number of rows in forecast. Used in plotting routine. |
thres |
Thresholds used in model |
This function creates an image plot of the intensity plot of the skill scores as a function of spatial scale and threshold. The top row is equivalent to the bias of the forecast.
Barabara Casati <barbara.casati (at) ec.gc.ca>
B.Casati, D.B. Stephenson, G. Ross. A new intensity-scale approach for the verification of spatial precipitation forecasts. Meteorological Application (RMS), in press.
http://www.met.rdg.ac.uk/~swr00bc/
## simulated example n<- 5 set.seed(10) forecast1 <- matrix( log(rlnorm(n = (2^n *2^n) )) , nrow = 2^n) obs1 <- matrix(log( rlnorm(n = (2^n *2^n) )) , nrow = 2^n) int.scale.verify(forecast1, obs1, main = "Test Case") ## real example. Data source referenced below. data(analysis.dat) data(forecast.dat) require(waveslim) require(fields) A<- int.scale.verify(forecast.dat, analysis.dat, thres = c(0, 2^seq(-5,6)), main = "NIMROD example" ) plot(A)