Sort of a statistical puzzle.
The Weather Bureau does a poor job of predicting long all-day rain or snow events, which were common this winter and spring. They miss by many hours, sometimes miss entirely.
But they always do an excellent job with convective 'popup' events. Last night they predicted that the Spokane area would be affected by scattered T-storms from 2 to 5 AM. At the time of the prediction, nothing was showing on radar. At 2:11 the first little burst of rain; several more little bursts and a few lightning strokes; all done and moved out at 4:35. Just like they said!
Intuitively it seems like prediction should work the opposite way. All-day events should be easier to get right, and convection should be tricky.