Smoke season dims the panels. The data is in.

What smoke actually does to a panel

Smoke is aerosol, which is just fine particles suspended in the air, and those particles scatter and absorb sunlight before it ever reaches the glass. The biggest loss is to direct normal irradiance (the sunlight arriving in a straight line from the sun's disk), which is what a panel converts most efficiently. Some of that light is not destroyed but scattered into diffuse light, the flat glow of a hazy sky, and a panel still harvests part of it. The total that lands on the array, what engineers call plane-of-array irradiance, still falls.

The second mechanism is slower and dirtier. Ash and fine soot settle on the glass, a process the industry calls soiling, and a coated panel loses output even after the sky clears. In western Washington the wet season usually rinses panels clean on its own, which is part of why the problem reads as a late-summer event here rather than a year-round one. East of the crest, in the drier Columbia Basin, a dusty or ashy panel can stay that way until someone hoses it off or the rain finally comes.

How much output goes missing

Here is where this desk has to be careful. Measurements from heavy-smoke days in the western United States have shown solar output falling on the order of ten to thirty percent on the worst days, with thin high smoke costing a few percent and a thick, low plume costing much more. Those figures come from observations in California and the broader West in recent wildfire summers, and Washington-specific public numbers are thinner, so treat any single percentage you see with suspicion.

Timing is the sharp part. August and early September, the stretch when smoke is most likely to settle in, is also the stretch when the sun sits highest and a panel earns most of its year. A smoky week in January would barely register. A smoky week in late August lands on the fattest part of the production curve, which is why a modest percentage loss can feel larger on the statement than the raw number suggests.

Is the forecasting catching up

Short answer, partly. The National Oceanic and Atmospheric Administration runs a weather model called HRRR-Smoke (a high-resolution forecast that tries to predict where plumes will drift and how thick they will be), and it has grown steadily better at the one-to-two-day horizon that matters for scheduling. Washington's Department of Ecology also runs an air-monitoring network whose readings, built for public health, double as a rough proxy for how much the sun is being blocked on a given day.

For planning, the people who move the most power care less about any single rooftop. Bonneville Power Administration and the state's utilities balance smoke-dimmed solar against the Columbia and Snake River dams, which do not care about haze, so the lights stay on. What forecasters still cannot do well is tell a solar owner a week out whether the second half of August will be clear or orange, because that depends on fires and winds that have not happened yet.

Worth watching this month

1. Washington's Department of Ecology air-quality dashboard, which updates through the fall fire season and is the easiest public read on how smoky a given day actually is.

2. NOAA's HRRR-Smoke forecast maps, worth a glance the night before if you are deciding whether to run a big load off your own panels.

3. Bonneville Power Administration's public generation and load figures, which show how little a smoky week moves a hydro-heavy system (routine, but reassuring).

4. Any Washington Utilities and Transportation Commission filings on net metering or community solar, which could change what a smoke-dimmed month is worth to you (usually slow, occasionally consequential).

5. The first real wet front of the fall, the unofficial end of smoke season west of the Cascades, after which rooftop output usually recovers on its own.