Soiling stations vs software soiling monitoring (2026)

A soiling station is dedicated hardware. The reference-cell type, such as Fracsun’s station (sold as NX Clario since Nextpower acquired Fracsun in November 2025), compares a daily-washed cell against a naturally soiling one. The optical type, such as Kipp and Zonen’s DustIQ, measures scattered light from dust on glass, needs no water, but requires local dust calibration for stated accuracy. Atonometrics’ RDE300 series uses a clean reference cell against a soiled reference module. All three now sit inside larger hardware groups after the November 2025 consolidation, when Nextpower bought Fracsun and DustIQ owner OTT HydroMet bought Atonometrics in the same week.

Software soiling monitoring uses no added hardware. A physics model corrects each inverter’s output for weather, temperature, and curtailment; the residual, slowly-varying loss that rain resets is soiling. Because the estimate exists for every inverter, it produces a soiling map of the plant rather than a reading at one mast.

IEA-PVPS Task 13, the industry reference on soiling, states that soiling does not occur homogeneously across a plant, that single-point short-circuit measurements can underestimate the real power impact, and that accounting for heterogeneity requires integrating multiple soiling monitors, an underestimated cost factor. Its 2025 fact sheet is more direct: soiling is highly heterogeneous at module and plant level and requires multi-sensor networks for accurate cleaning decisions, while sensors should ideally be maintenance free because many sites are unmanned. NREL adds that uncertainty grows with distance from the measurement point and nearby sites can soil differently. The same body estimates soiling now costs the industry 4 to 7 percent of global PV energy, a multi-billion-euro annual loss.

That is the case for software estimation in one paragraph: it is the only approach whose spatial resolution matches the documented spatial variability of the problem, and it carries none of the per-unit maintenance burden the hardware guidance itself flags.

A station measures; software infers. For resource assessment before construction, for bankability reports an independent engineer must sign, and for validating a cleaning contractor’s performance against a contractual soiling threshold, a measured instrument record carries weight a model does not. Reference stations with automated washing, like Fracsun’s, also remove most of the manual-cleaning labor that made older two-panel stations unpopular. If a lender requires a station, install the station.

The practical 2026 answer for large plants is both: one or two stations as calibration anchors, software for the per-inverter map that decides which blocks get cleaned this week. For distributed C&I portfolios, where nobody will maintain a station per rooftop, software estimation is realistically the only option with full coverage.