VALUE Validation Indices and Performance Measures

The following list presents different classes of validation indices. For each meteorological variable, the specific values used to describe the indices (e.g. thresholds) are chosen based on expert knowledge (e.g., impact researchers, STARDEX report, etc.). This list will be updated on a regular basis.

 

mpe: mean percentage error

mse: mean squared error

  

Index

Performance Measure

 

Distributionwise Validation

Pairwise Validation


Marginal Distribution

Mean

bias/mpe

 

Variance

mpe

 

Skewness

bias

 


Temporal Dependence

time series

 

mse

acf lag 1,2,3

just indices

 

quantiles/return values

bias

 

threshold exceedance

 

Brier score/mse of 0/1 series

number of threshold exceedances

bias

 

amount above threshold

bias

 

shape parameter of GEV

bias

 

time of maximum/minimum of annual cycle

bias

 

amplitude of annual cycle

mpe

 

spell length dist quants

bias

 

transition probabilities

just indices

 

proportion of variance in low frequency band

just indices

 

sign of the low pass filtered series

 

Brier score/mse of 0/1 series


Spatial Dependence

distribution of daily relative areas of threshold exceedances

see above

mse

range of smoothed variogram

mpe

 

madogram

mpe

 


Multivariate Dependence

corr(detrended variables)

just indices

 

variable conditional on (no) exceedance

see above

 

joint exceedances

see above

 


Spatial Climatological Summaries

 all indices

mean absolute error, mean bias

 

absolute errors of spatial means

 

mpe of spatial means

 

means of Brier scores

 

Taylor diagram