Bushfire smoke is an unusual kind of risk for a solar farm. Unlike a hailstorm or a grid outage, it does not respect property boundaries. A fire burning in a forest 150 kilometres away can dim the sunlight reaching a photovoltaic array without any flame coming near the site. New research from UNSW Sydney has put numbers to that non-localised exposure-and in doing so, has made a case that forecasting, rather than hardware, may be the most direct lever available to protect output.
The research indicates that siting can also be regarded as a technique of long-term mitigation. Before selecting site locations, developers can evaluate previous fire hotspots, common wind direction and probable smoke trajectories to make informed decisions about smoke exposure, putting this factor on the same level with irradiance and grid access in terms of feasibility study variables. Existing gel assets can use a similar approach for predictive maintenance.
The physical connection that enables these numbers is not one single process. Smoke aerosols and tiny particles in the atmosphere reflect or absorb part of the solar radiation that is before the module plate. Particles fall on panels, causing a layer of dirt to form, thus uneven accumulation can lead to localized overheating of solar cells that will use their efficiency over time. The elevated temperatures of solar panels during smoke events also play a role in this process, since temperatures have gone up 10°C to 20°C from their climatological values during Black Summer resulting in additional 0.15% thermal degradation.
The researchers' central operational recommendation follows directly from that profile: combine bushfire information, wind patterns and smoke-trajectory modelling to predict which solar farms are likely to be in the path of a plume, in the same way weather forecasts already flag temperature and wind conditions. Advance warning does not eliminate the smoke, but rather it opens up avenues for targeted measures: conducting inspections, organizing an extra cleaning, sealing the affected zone by means of remote monitoring, or temporarily angling the panels in such a way as to reduce heat exposure and to help particles move off the surfaces. Each of these measures has its downside, which means, for example, that turning away the panel from the sun means to sacrifice a part of the energy production generated. With advance warning the trade-off can be made in advance and purposefully instead of reactively after the energy production is already compromised.
The research indicates that siting can also be regarded as a technique of long-term mitigation. Before selecting site locations, developers can evaluate previous fire hotspots, common wind direction and probable smoke trajectories to make informed decisions about smoke exposure, putting this factor on the same level with irradiance and grid access in terms of feasibility study variables. Existing gel assets can use a similar approach for predictive maintenance.
One point of caution needs to be taken into consideration: the losses are projections based on models rather than on actual measurements for specific farms, and the researchers were unable to eliminate the impact of smoke from the conscious decision to cut emissions. The authors of the study provide a map which shows the direction and magnitude of the risk instead of the formula for calculating yield losses. However, the direction is in line with the previous experience. Smoke that originated from the wildfires in California in 2020 reached Europe, while the smoke from the 2025 wildfires in Spain traveled to the UK and northwestern Europe. The methodology of modeling the process used by the authors can be applied to regions in Canada and the USA that are frequently affected by fires.
For an industry accustomed to treating yield risk as a function of irradiance, temperature and soiling cycles, the UNSW work adds a fourth, largely unpriced variable: the probability that a distant fire will put a plume over the array. Better forecasting does not eliminate that variable. It does, however, convert it from a surprise into something that can be planned around-and that is the difference between a yield loss and an unmanaged one.
