Forecast Comparison

How to read this. Every set below is scored on the same matched cases: a case counts only where the gauge total and every compared forecast exist, so differences are not an artefact of one product covering more stations or more hours than another. Three questions, three panels — bias (is the mean right?), CRPS (is the whole distribution right?), and Brier skill (is it right about threshold exceedance?). A forecast can pass one and fail another, which is the point of showing all three.