shaded boxes for confidence boundaries:
95%Framed box where the difference is significant with 95% confidence, light colour box elsewhere.50%/95%Framed box where the difference is significant with 95% confidence, light colour box for 50% confidence, grey elsewhere.95%/99.7%Framed box where the difference is significant with 99.7% confidence, light colour box for 95% confidence, grey elsewhere.(legend) ||
significance triangles ||
bars
||
sample sizeColour map showing the fraction of the actual number of scores to the theoretical maximum, derived from the total list of dates in the scorecard definition.
n.hemNHem Extratropics (lat 20.0 to 90.0, lon -180.0 to 180.0)
s.hemSHem Extratropics (lat -90.0 to -20.0, lon -180.0 to 180.0)
tropicsTropics (lat -20.0 to 20.0, lon -180.0 to 180.0)
europeEurope (lat 35.0 to 75.0, lon -12.5 to 42.5)
n.amerN America (lat 25.0 to 60.0, lon -120.0 to -75.0)
e.asiaE Asia (lat 25.0 to 60.0, lon 102.5 to 150.0)
austnzAustralia N Zealand (lat -45.0 to -12.5, lon 120.0 to 175.0)
n.atlN Atlantic (lat 25.0 to 65.0, lon -70.0 to -10.0)
n.pacN Pacific (lat 25.0 to 60.0, lon 145.0 to -130.0)
arcticArctic (lat 60.0 to 90.0, lon -180.0 to 180.0)
antarcticAntarctic (lat -90.0 to -60.0, lon -180.0 to 180.0)
ccafAnomaly correlation/seepsStable Equitable Error in Probability Space
rmsefRoot mean square error
sdafStandard deviation of forecast anomaly
ccafAnomaly correlation/seepsStable Equitable Error in Probability Space
rmsefRoot mean square error
sdafStandard deviation of forecast anomaly
ccafAnomaly correlation/seepsStable Equitable Error in Probability Space
rmsefRoot mean square error
sdafStandard deviation of forecast anomaly
ccafAnomaly correlation/seepsStable Equitable Error in Probability Space
rmsefRoot mean square error
sdafStandard deviation of forecast anomaly
ccafAnomaly correlation/seepsStable Equitable Error in Probability Space
rmsefRoot mean square error
sdafStandard deviation of forecast anomaly
ccafAnomaly correlation/seepsStable Equitable Error in Probability Space
rmsefRoot mean square error
sdafStandard deviation of forecast anomaly
ccafAnomaly correlation/seepsStable Equitable Error in Probability Space
rmsefRoot mean square error
sdafStandard deviation of forecast anomaly
ccafAnomaly correlation/seepsStable Equitable Error in Probability Space
rmsefRoot mean square error
sdafStandard deviation of forecast anomaly
ccafAnomaly correlation/seepsStable Equitable Error in Probability Space
rmsefRoot mean square error
sdafStandard deviation of forecast anomaly
ccafAnomaly correlation/seepsStable Equitable Error in Probability Space
rmsefRoot mean square error
sdafStandard deviation of forecast anomaly
ccafAnomaly correlation/seepsStable Equitable Error in Probability Space
ccaf=Anomaly correlation,rmsef=Root mean square error,sdaf=Standard deviation of forecast anomaly,seeps=Stable Equitable Error in Probability Space
Each error plot is converted into a sequence of symbols (e.g.
▼▽░█░△▲)
where each symbol indicates for given time step whether the experiment is significantly better or worse than the control on different confidence levels.
Different colours are used for symbols comparing control and experiment activity or spread: purple â–¼ indicates the experiment is more active or has larger spread than the control and green â–² the opposite.
Symbol legend: for a given forecast step... â–² experiment better than control statistically significant with 99.7% confidence â–³ experiment better than control statistically significant with 95% confidence â–‘ experiment better than control statistically significant with 50% confidence â–ˆ not really any difference between control and experiment â–‘ experiment worse than control statistically significant with 50% confidence â–¾ experiment worse than control statistically significant with 95% confidence â–¼ experiment worse than control statistically significant with 99.7% confidence )
red = the experiment (esuite) is worse than the control. blue = the experiment (esuite) is better than control.
purple = the experiment is more active than the control. green = the experiment is less active than control.
Colour hue and saturation are proportional to normalised difference value.
A colour-framed cell indicate the differences are statistically significant on 95% confidence level.
However, where the differences are not significant on at least 0% level they are masked by a light grey box.
For more details:
Hover the mouse cursor over a cell and a small information window pops up with the difference values.
Click on a cell and a graph of normalized differences will show.
ecm7950@ac6-101 2026-05-06 08:14:48 by quaver/3.7.1
Command-line arguments: --title AIFS2.0 vs AIFS1.1 from 50r1 (both vs 50r1) --experiments vstream_root=prepml_j0b1,class=rd,refclass=od,refexpver=0080,stream=oper,type=fc,expver=j0b1,date=2026010100/to/2026033112/by/12 --controls vstream_root=prepml_j4av,class=rd,refclass=od,refexpver=0080,stream=oper,type=fc,expver=j4av --scope_file scope_aifs_exp.json --domains n.hem/s.hem/tropics/europe/n.amer/e.asia/austnz/n.atl/n.pac/arctic/antarctic