At 2 Sicilian solar farms on pale limestone soil, bifacial trackers were measured string by string, and the rear face of every panel delivered 5.3 percent more power than standard models assumed
Bone white limestone bakes under a southern Sicilian sky, and a floor that bright is supposed to hand free electricity to the back of every panel in the field.
The rear glass catches that bounce and converts it.
Most project models assign the bonus a rough round number and move on.
To put that in context, a 10 MW plant generating roughly 17,000 megawatt-hours a year picks up approximately 900 megawatt-hours annually from the rear face alone.
Two operating farms measured what actually comes back off the ground, string by string.
So why do rear-face estimates keep missing, and what does the soil underneath have to do with it?
Why every model starts with an educated guess on the back face
A bifacial panel is a two sided collector. The front face catches direct sunlight. The rear face catches whatever bounces back from the ground, neighboring panels and a cloudy sky, a quality the industry calls ground albedo, and albedo is the hard part to model because it depends on soil color, grass cover, gravel, dust and the geometry of every row around it.
The standard approach estimates albedo as a single flat number for the whole site, multiplies it by a bifaciality factor printed on the module datasheet, and calls it done. That shortcut made sense when bifacial panels were a niche product and a one percent modeling error was inside the noise. But bifacial technology has now crossed 90 percent of the global module market, meaning nearly every new utility scale plant relies on the same rough estimate to close its financing.
Ground albedo is not uniform across a site. It varies row by row, hour by hour and season by season, and a model that flattens that variation into one figure will be wrong in ways that compound across a twenty-year contract.
What the Sicilian farms actually look like on the ground
The study examined two multi-MW solar plants in southern Sicily, equipped with monocrystalline silicon bifacial modules on single axis tracking systems aligned north to south. That configuration is now the global industry standard for utility scale solar: trackers sweep the panel face through the day, following the sun’s arc from east to west, while the north-south axis maximizes the afternoon sweep.
Sicily is a near-textbook site for testing this hardware. Sitting at roughly 37 degrees north, the sun angles are steep for much of the year. The pale limestone and clay soils push ground albedo far higher than the green-grass default that most European models assume, and those soils dry to near-white in summer, exactly when panels are producing hardest.
Working at the stringbox level means the team was not averaging across an entire field. Instead, they were reading the output of small groups of panels at specific row positions, which is precisely where rear-face variation shows up most clearly.
The 5.3 percent that standard models leave on the table
The researchers built an optimized energy model at the stringbox level, using data filtering, clear-sky condition selection and numerical estimation of bifaciality factors, calibrated on measurements taken during the first operational months. Their result: the rear-side contribution produced additional energy gains of approximately 5.3 percent.
To put that in context, a 10 MW plant generating roughly 17,000 megawatt-hours a year picks up approximately 900 megawatt-hours annually from the rear face alone. At a wholesale price of around 50 dollars per megawatt-hour, that is close to 45,000 dollars a year flowing through a gap that many project models set to zero or round to two percent.
The Sicily result arrived from measured hardware, not a modeled scenario, and that distinction carries real weight in a financing conversation. Solar PV plants worldwide have systematically underperformed, with underperformance rates ranging from 7 to 13 percent and triggering multimillion-dollar performance contract disputes. A model that undershoots bifacial gain compounds that gap from day one.
Where the default breaks down and who it hurts most
The flat-albedo shortcut is not wrong in every setting. On sites with dark soil, dense vegetation or frequent cloud cover, the rear face contributes little and the rough estimate lands close enough. It breaks down on bright, arid surfaces, the very conditions across much of the American Southwest, the Middle East, northern Chile and the Mediterranean basin, which together account for a large share of the world’s utility scale pipeline.
Trackers make things worse. A fixed-tilt panel always presents the same geometry to the ground. A tracker changes its angle continuously, so the shadow pattern beneath each row shifts all day and rear irradiance at any given string changes with it. Collapsing that into a single albedo constant was a reasonable shortcut when software could not resolve the geometry. Today it is simply a choice to leave money in the ground.
The investors and lenders who sign twenty-year power purchase agreements feel the gap most acutely, because the shortfall compounds over the life of the contract. For a look at how automated construction is closing other cost gaps on solar sites, see how robotics platforms are cutting build costs, a pressure that makes accurate yield modeling even more important when margins are already tight.
What comes next for bifacial accounting
The Sicily team’s method is not exotic. Working at stringbox resolution simply requires a finer data pipeline than most developers commit to during early-stage modeling, and many operating plants already have the sensors in place. The barrier is not hardware but the willingness to replace a convenient assumption with a measured one before financing closes rather than after the first annual report lands short.
The authors calibrated the model only during the first operational months. Longer-term drift in soil albedo as vegetation establishes or dust accumulates differently row by row remains an open question, and the rapid growth of bifacial market share calls for systematic reduction in uncertainty, especially on single axis tracker systems where rear-side irradiation behaves in significantly more complex ways.
The honest conclusion is that 5.3 percent is not guaranteed at every bifacial tracked site; it is what two specific Sicilian farms delivered when someone looked carefully enough. For context on what happens when solar output meets peak demand, the story of grid operators under peak load is the other half of the same picture. A plant that models its output accurately is one a grid operator can plan around, and that reliability matters as much as the yield itself.
Hugo is an engineer with strong technical expertise. Multilingual from an early age, his writing combines technical clarity with a strong interest in science and energy.
