What is soiling loss in solar PV? A working definition
Soiling loss is the drop in a module's power output caused by anything sitting on the glass between the sun and the cells. Dust, pollen, bird droppings, ash, salt spray near the coast, agricultural residue if you're next to a field. The panel still works. It just isn't seeing full sun anymore, and the inverter reports less than the irradiance on site would predict.
That gap, between what the array should produce given the weather and what it produces, is how soiling loss gets measured at the portfolio level. Pull the expected yield from an irradiance sensor or satellite model, compare it to metered output, and the shortfall that isn't explained by temperature derate, clipping, or known outages gets filed under soiling. Most O&M contracts carry a soiling loss assumption somewhere between 1% and 4% annually depending on climate, and it shows up as a line item in the monthly performance report whether anyone's looked at a panel or not.
Why the number alone doesn't tell you where to send a crew
Here's the problem with soiling loss as a portfolio-level percentage: it's an average. A 2% soiling loss across a 10 MW site could mean every table is lightly dusted and evenly underperforming, or it could mean three rows near the access road are caked and the rest of the site is fine. The production number can't tell those apart, and a wash crew dispatched on the average will clean rows that didn't need it and miss the ones dragging the number down.
Soiling also isn't flat across a site even when it looks flat from the ground. Rows closer to a gravel access road or a construction laydown area pick up more dust than rows set back. Low-tilt arrays and the bottom edge of any tilted row collect more than the top, because rain and wash cycles run off and redeposit grit along the lower edge. Bird activity concentrates under roosting structures or near open water. None of that shows up in a site-wide average. It shows up row by row, and usually only when someone looks.
Soiling loss versus a hard fault
Soiling and electrical faults can produce the same dent in production, and neither one is visible from the SCADA screen as "this is dirt" versus "this is a fault." Dispatch the wrong crew on that guess and you've burned a truck roll for nothing. Thermal imaging is what actually separates the two: a soiled module runs a few degrees warmer than its clean neighbors across the whole panel face, fairly uniform, while a cell-level fault or string outage shows up as a sharp hot spot or a dead section with no heat signature at all. Send a wash crew after a diode failure and the production dip doesn't move. Send a repair tech after a dusty row and you've wasted a service call on a table that just needed a hose.
That's the gap between knowing your site has soiling loss and knowing which tables to clean first. A thermal flyover paired with RGB imagery catches both signatures in the same pass and ranks the findings, so the crew shows up with a list instead of a camera and a guess. One annual flyover replaces walking every row with a handheld thermal camera, which is still how a lot of techs find out a table's underperforming, usually after someone's already noticed the production dip and asked why.
Soiling loss is a real number on your production report, but it's an average hiding a map. Knowing the average tells you the problem exists. Finding which rows carry the loss is a different job, and it's the one that decides where the truck goes next.
If your production reports show a soiling number you can't yet pin to specific rows, that's exactly the gap Panel Soiling's annual flyover is built to close.