Same one-hour storm, drawn at three resolutions. Bar height + brightness = real rain intensity at that slice (so light reads short + dim, heavy reads tall + bright, and you see the storm's shape). This is the data we actually have: the Pirate Weather proxy returns 1-minute readings (verified live). 15-minute is only the fallback if that proxy is down.
1-minute bars (60)Pirate Weather proxy · LIVE
Every minute its own bar. You can see the storm build, peak, and taper, and a light drizzle would read as a low, gently varying line, never an even row.
NOW15304560 min
5-minute bars (12)averaged from the 1-min data
Calmer, still clearly shows where it gets heavy and where it eases. A good default if 60 bars feels busy on a phone.
NOW15304560 min
15-minute bars (4)fallback only (proxy down)
Four blocky bars. The whole shape of the storm collapses into a few steps, this is what we do NOT want unless we're forced to.
NOW15304560 min
Bottom line: 1-minute and 5-minute are both on the table right now. My lean is 1-minute bars for the detail (it's our real data and it looks alive), with a clean fallback to fewer bars only when the proxy can't be reached. The fix for your "even rows" is making the built bars actually track intensity like these do.
Tall + bright = heavy rainShort + dim = light/drizzleFlat baseline stub = dry minute