Why there's no single schedule, how rain resets dirty panels, and the dry seasons in San Diego, Sacramento, Phoenix and Las Vegas. From NWS, NREL and Sandia.
By the SolarCleaningRated team · Facts checked against the sources listed below on November 1, 2026. General information, not professional advice.
How often to clean depends on how fast dirt builds up where you live and how often rain washes it off. REC, a panel maker, recommends cleaning when dirt can be seen on the glass. Maxeon's manual recommends periodic cleaning, saying it has resulted in improved performance especially in regions with low annual precipitation. The best-practice guide for solar operation and maintenance from the National Renewable Energy Laboratory (NREL) says soiling can often be traced to a source, such as dust from a construction site, and suggests cleaning after the end of pollen season. Rain is the other half of the equation.
Sandia National Laboratories' solar modeling guide describes the Kimber soiling model, which is used to estimate dirt losses. It assumes dust builds up at a steady daily rate until a rain heavy enough to clean the panels resets them, followed by a grace period in which no new soiling occurs. The model's authors estimated soiling rates of 0.1 to 0.3 percent a day at locations in the southwestern US. The model needs a soiling rate as an input, and Sandia says it is typically extracted from measured PV performance data.
How heavy is heavy enough? Sandia's worked example uses a cleaning threshold of 3 millimeters of rain and a 14-day grace period. The open-source pvlib software's version of the model defaults to 6 millimeters, about a quarter of an inch, in a day, with the same 14-day grace period and a maximum loss of 30 percent. Either way, a light sprinkle doesn't reset the panels in the model, and a long dry season lets losses add up.
Some simple arithmetic of our own shows why the dry season matters. At the 0.1 to 0.3 percent a day the model's authors estimated for the Southwest, 60 days of build-up would come to roughly 6 to 18 percent, and 120 days to 12 to 36 percent, which pvlib would cap at its 30 percent default maximum. These are illustrations, not model runs: in the model, the 14-day grace period after a cleaning rain shortens the build-up. Real losses depend on your own site, which is why your monitoring data beats any rule of thumb.
One more detail from pvlib: it allows for manual wash dates, and after a manual wash there is no grace period, so soiling starts building again straight away. In pvlib's version of the model, then, a hand cleaning in the middle of the dry season buys less time than a soaking rain does.
NREL researchers mapped soiling data from 83 sites in the United States and found that most soiling occurs in the southwestern US, with Southern California counties experiencing the greatest losses because of high particulate matter concentrations and long dry periods. San Diego County and the Inland Empire are both in Southern California, and Phoenix and Las Vegas are in the desert Southwest, so for most of our readers the local dry season is the number to know.
The National Weather Service's 2004 climate study of San Diego says 85 percent of the rainfall occurs from November through March, only 10 percent of the seasonal total normally falls from May through October, and only 2 percent in June through August. It gives an official 1971 to 2000 average of 10.77 inches a year. NREL's mapping, above, points to Southern California counties as a group for the greatest soiling losses.
The National Weather Service's 2005 climate study of Sacramento says more than half the year's precipitation falls in the winter rainy season, November through February, and describes summers as virtually cloudless, with warm, dry days. Even in winter, it says, measurable rain falls on only about ten days a month. Its 1971 to 2000 normals show 0.18 inches in June, 0.05 inches in July, 0.05 inches in August and 0.37 inches in September, out of 19.87 inches a year.
The Arizona State Climate Office's summary for Phoenix gives 1991 to 2020 normal precipitation of 7.22 inches a year, with 0.22 inches in April, 0.13 inches in May and 0.02 inches in June, and an average of 36 days a year with at least 0.01 inches of precipitation.
The National Weather Service office in Las Vegas puts the 1991 to 2020 normal annual precipitation at 4.18 inches, with 0.04 inches in June, 0.32 inches in September and 0.30 inches in November. Compare that with the quarter-inch daily threshold in the pvlib model above.
When you do clean, do it safely and the way your panel maker allows; our step-by-step cleaning guide covers that, and our guide to whether panels need cleaning has the measured losses.
Not much. The Kimber soiling model treats panels as cleaned only by rain above a threshold; Sandia's example uses a threshold of 3 millimeters, and the pvlib software defaults to 6 millimeters, about a quarter inch. Lighter showers don't reset the panels in the model.
It depends on your dust and rainfall. NREL's mapping found the greatest soiling losses in Southern California counties, and Phoenix and Las Vegas get little rain. REC's rule of thumb is to clean when dirt is visible.
Toward the end of your dry season, after pollen season as NREL suggests, or after dust from nearby construction. Clean in the cool part of the day, as REC advises.
41 companies in 5 areas, scored by the same published criteria.