How it works, step by step
- Desk studyWe list the dust sources around the site from maps and satellite images, and gather rainfall, wind, humidity and dew records. Wind direction and dew decide where dust lands and whether it sticks (IEA PVPS 2022).
- Site visitAn engineer walks the site and its surroundings, confirms each source and its distance and direction from the planned blocks, and chooses where the instruments go.
- Measurement set-upReference panels or a soiling station, set at the planned tilt, in the blocks nearest and furthest from the main sources. Optical soiling sensors add positions where they suit.
- Dust and water samplesDust samples from the reference glass for laboratory analysis, and a test of every water source the plant could use: borehole, municipal or trucked.
- Measurement periodAt least one dry spell of 14 days or more, the minimum to fit a soiling rate (IEA PVPS 2022). IEA PVPS says a direct pre-installation measurement needs at least a year; shorter records can calibrate a soiling model (IEA PVPS 2025).
- Design and O&M inputsSoiling rates by season, the dust and its sources, the water result and a costed cleaning plan, with the method and uncertainty behind each figure.
Why measure soiling before you build?
Soiling changes from site to site. Across the Atacama Desert, uncleaned modules lost 39% of a year’s energy at Arica on the coast, about 18% at Iquique and 3% or less at high-altitude and southern sites such as Calama, Copiapó and La Serena (Cordero et al. 2018). NREL (2018) reports soiling of 0.36% a day in an area of heavy farming, against 0.01% a day in desert with no farming, construction or industry.
- Arica, on the coast39%
- Iquiqueabout 18%
- High-altitude and southern sites3% or less
Same desert, very different losses (Cordero et al. 2018).
Yet soiling measurements are often missing for a project’s site, and where they exist the records are usually short (IEA PVPS 2025). Models built from weather data can fill the gap, but their accuracy depends on the site and on how they treat rain, which IEA PVPS calls a major source of uncertainty. Site measurements calibrate them. In Southern Africa the gap is wide: we found no peer-reviewed PV soiling rates for Upington, Kathu, Prieska, Kimberley, Namibia or Botswana.
The decisions that soiling affects most are made early. IEA PVPS (2025) says soiling data belongs in the yield assessment, and that mitigation starts before operation: in site assessment, component selection and system design. Once the modules, rows and water supply are built, they are hard to change.
What we measure
| What | How | What it decides |
|---|---|---|
| Soiling rate, by season | Reference panels or a soiling station at the planned tilt; optical soiling sensors where they suit | The soiling loss in the yield model, and the cleaning interval |
| The dust itself | Samples from the reference glass, analysed in a laboratory | Whether water removes it, whether it cements, and which product will |
| Local sources | Site walk, maps, satellite images and wind records | Which blocks soil fastest, and where sensors and cleaning effort go |
| Cleaning water | Hardness, TDS and pH of every source the plant could use | Treatment, storage, and which modules’ limits it meets |
A dust sample taken from the glass can be analysed in a laboratory to show where the soiling comes from (NREL 2018). At the Green Energy Park in Benguerir, Morocco, researchers analysed the local dust by XRF, FTIR and SEM and found calcite, dolomite, phosphorus and quartz, from industrial activity and construction nearby (Elamim et al. 2023).
Local sources we map
The source decides whether dust will cement, stain or wash off, and the wind decides which blocks catch it. Soiling can vary across a site with the prevailing winds and the spread of dust sources: see soiling monitoring. We map:
- Mines, haul roads and dirt roads. Mineral and road dust. IEA PVPS (2022) suggests lining nearby dirt roads with vegetation to deflect dust gusts. See desert and mining dust.
- Farmland. Ploughing raises dust; NREL (2018) suggests timing cleans to follow it.
- Cement, lime and quarry works. Alkaline dust that cures into films water can’t remove. See cement dust.
- Smelters, steelworks and rail lines. Iron and other metal oxides that stain the glass. See rust and metal oxides.
- The coast. Sea salt is one of the natural aerosols that soil PV (IEA PVPS 2025).
- Pans and dry riverbeds. Between 2005 and 2008, 328 daytime dust plumes more than 10 km long were traced to 101 sources, mostly ephemeral inland lakes, coastal pans and dry river valleys in Namibia, Botswana and South Africa (Vickery et al. 2013).
- Birds and water. Roosts, perches and dams bring droppings. See bird droppings.
Design and O&M inputs
| Decision | What the study gives |
|---|---|
| Soiling loss in the energy yield model | A measured soiling rate by season and a planned cleaning interval, in place of an assumed figure |
| Module, glass and coating | The water result against each shortlisted maker’s cleaning limits, and site evidence on coatings: a hydrophobic coating promoted dust soiling in a Northern Cape field study (du Plessis 2017; du Plessis et al. 2020) |
| Tilt, tracker and row layout | Soiling falls as tilt rises (IEA PVPS 2022), and Trina asks for more frequent cleaning of modules laid flat than of modules at 10° or more. In Pretoria’s dry winter, a single-axis tracker within 0.5 m of the ground lost 1.2% of performance ratio a week, and a dual-axis tracker beside it 0.4% (CSIR data in IEA PVPS 2022) |
| Cleaning method and equipment | Many robots need the plant designed around them (NREL 2018; Ilse et al. 2019), and tractor rigs need access roads and row spacing. See equipment selection |
| Water storage and treatment | Volume per wash from the method’s water use, about 1 L per m² for a wet clean in a US study cited by NREL (2018), and the treatment each source needs. See water testing |
| Cleaning budget for the O&M contract | The right interval depends on your tariff, soiling rate and what a clean costs, and it changes with the season. The calculator works it out for your plant. |
| Operational monitoring | Where the permanent soiling sensors go. See soiling monitoring |
Who uses the results
- Developers: site choice, and the soiling line in the yield estimate.
- EPCs: module choice, layout, access for cleaning machines and the water supply.
- Owners’ engineers and lenders’ technical advisers: a soiling assumption with a stated method and uncertainty behind it.
- Asset owners: the cleaning budget in the O&M contract. See asset management support.
- O&M contractors: a method, calendar and water plan from the first day of operation.
We run studies for planned plants across Southern Africa, in South Africa, Namibia, Botswana, Zimbabwe, Mozambique and beyond. For an operating plant, the equivalent is a site soiling assessment, or a full plant study on very large sites. The measurement method is set out in how to measure PV soiling.
CSP sites too. For a planned CSP plant the same logic applies to mirrors. Around a ferromanganese smelter at Emalahleni, a study of whether a heliostat field could work nearby found test reflectors lost 32.6% of their reflectance per 14-day dry-season period on average. The best location soiled 13.1% less than the worst workable one, and the authors recommend optimising mitigation, including the location relative to the source, for the worst season (Swart et al. 2023). We measure reflectance on test mirrors: see CSP mirror reflectivity.
