Comparison of scores of model cycle 38r2 (combination of research experiments and the current e-suite) and cycle 38r1 (the current operational model) verified by the respective analyses or radiosonde observations at 00UTC for combined periods from 1 January 2012 to 31 March 2012 and from 1 June 2012 to 23 April 2013.

    Anomaly CorrelationRMS Error
Europe

against analysis

Relative humidity300hPa
700hPa
Temperature100hPa
500hPa
850hPa
1000hPa
Wind200hPa
850hPa
Geopotential100hPa
500hPa
850hPa
1000hPa
against observationsTemperature100hPa
500hPa
850hPa
1000hPa
Wind200hPa
850hPa
Geopotential100hPa
500hPa
1000hPa
Northern hemisphere extra-tropicsagainst analysis10m wind over ocean 
Ocean wave period
Ocean wave height
Relative humidity300hPa
700hPa
Temperature100hPa
500hPa
850hPa
1000hPa
Wind200hPa
850hPa
Geopotential100hPa
500hPa
850hPa
1000hPa
against observationsTemperature100hPa
500hPa
850hPa
1000hPa
Wind200hPa
850hPa
Geopotential100hPa
500hPa
1000hPa
Southern hemisphere extra-tropicsagainst analysis10m wind over ocean 
Ocean wave period
Ocean wave height
Relative humidity300hPa
700hPa
Temperature100hPa
500hPa
850hPa
1000hPa
Wind200hPa
850hPa
Geopotential100hPa
500hPa
850hPa
1000hPa
against observationsTemperature100hPa
500hPa
850hPa
1000hPa
Wind200hPa
850hPa
Geopotential100hPa
500hPa
1000hPa
Tropicsagainst analysis10m wind over ocean 
Ocean wave period
Ocean wave height
Relative humidity300hPa
700hPa
Temperature100hPa
500hPa
850hPa
1000hPa
Wind200hPa
850hPa
against observationsTemperature100hPa
500hPa
850hPa
1000hPa
Wind200hPa
850hPa

Score card provides a quick visual overview over the performance of the experiment scores compared to control. It is a simplified summary of verify error plots of various domains, scores, parameters etc.

Each error plot is converted into a sequence of symbols (e.g. ) where each symbol indicates for given time step whether or not the experiment is significantly better or worse than the control.

Symbol legend: for a given forecast step... (d: score difference, s: confidence interval width)

Cy38r2 far better than Cy38r1 statistically significant (the confidence bar above zero by more than its height )(d/s>3)
Cy38r2 better than Cy38r1 statistically significant (d/s≥1)
Cy38r2 better than Cy38r1, yet not statistically significant (d/s≥0.5)
not really any difference between Cy38r2 and Cy38r1
Cy38r2 worse than Cy38r1, yet not statistically significant (d/s≤-0.5)
Cy38r2 worse than Cy38r1 statistically significant (d/s≤-1)
Cy38r2 far worse than Cy38r1 statistically significant (the confidence bar below zero by more than its height) (d/s<-3)

Statistical significance evaluated by paired t-test for 95% significance level.