What to look for: a dark band at the same time of day on most dates is recurring shading — something is blocking that part of the sky at that hour. If the band drifts up or down across the year, that's the normal seasonal swing of a fixed obstruction (tree, building, ridge), not new damage; a band that stays flat all year is more likely a structural cause close to the array, or an instrumentation issue, than horizon shading. The colour scale saturates outside the 5th–95th percentile, so a few very bright red cells on their own are usually cloud-transient noise (the pyranometer catching a cloud edge the array doesn't see) — it's the band pattern that matters, not isolated cells.

What to look for: line up a dip in the PR line above with the grid directly below it. Gaps in the grid under a PR dip mean the dip is a data problem, not a real performance drop — don't escalate it as an O&M issue. The cell colour tells you which side dropped: Power missing means the inverter/power telemetry went quiet while irradiance kept reporting, so check the power ingest path; Irradiance missing means the pyranometer went quiet while the plant kept producing, so check that sensor instead. A whole blank column points to an ingest outage; scattered single cells point to intermittent sensor dropout.

What to look for: this is the one grid where a red DOWN in daylight cell means the device really was off while the sun was up — a genuine availability loss worth flagging to O&M, unlike the grid above. Grey No state reported cells are a gap in the state channel itself, not a claim the device was down, so don't count those against it. Amber Unknown state code cells usually mean an unmapped value came through — worth a quick check that nothing changed on the device firmware/API side. Use the state-source selector to compare devices at the same site: one device DOWN while its neighbours keep operating points to a device-specific fault; all of them going down together points to a site-wide event (grid outage, curtailment) instead.