loading the real-data bundle…
Scorecard · Taylor · ROC The METviewer/METplotpy report views, recomputed live in the browser
Card 12 · Report views Real · de-identified
VIEW A Scorecard — is the model significantly better? hover a cell for values · symbols are paired-bootstrap significance

METviewer's signature view: each cell compares the model to the reference on their common init cycles only (event equalization), resampling cycles in pairs so the CI of the difference respects the pairing. / = the 95% CI of the difference excludes zero; doubled symbols = the 99% CI does too; · = not distinguishable.

VIEW B Taylor diagram σ & r from SL1L2 moments

Every quantity comes straight from the SL1L2 line: σf = √(FFBAR−FBAR²), σo = √(OOBAR−OBAR²), r = Pearson from the cross moments. No pair-level data needed — the Taylor diagram is fully derivable from MET's partial sums.

VIEW C ROC — precip thresholds deterministic multi-threshold ROC

One (POFD, POD) point per precipitation threshold from the summed 2×2 counts, anchored at (0,0) and (1,1) — MET's CTC-based ROC. The archive has no probabilistic output, so this is the honest deterministic version (a PCT-based ROC needs probability forecasts).