The best solar day of a British year is not the hottest one. It is a bright, cold, clear day in spring, when irradiance is high and the panels are cool. That is not a curiosity. It follows directly from how a photovoltaic cell behaves.
The coefficient is on the datasheet
Every module datasheet lists a temperature coefficient of Pmax, expressed as a percentage of rated power lost per degree Celsius above 25°C. Current crystalline silicon modules generally sit between 0.29 and 0.45%/°C. A Canadian Solar KuMax module, to take one published example, specifies 0.36%/°C.
The 25°C in that definition is cell temperature, not air temperature, and that distinction is the whole article.
How hot the cells actually get
Datasheets also give a nominal operating cell temperature, sometimes written NOCT or NMOT. It is the cell temperature reached at 800 W/m² of irradiance, 20°C ambient air and a light breeze. Typical modern values are 42 to 45°C.
Take 42°C at 800 W/m². The cells are sitting 22°C above the surrounding air. Scale that rise to the 1,000 W/m² of bright midday sun and it becomes about 27°C above air temperature. Roof mounting, where air cannot move freely behind the panels, pushes it further.
So the arithmetic for any given day is short. Take the air temperature, add about 27, subtract 25, multiply by the coefficient.
| Air temperature | Estimated cell temperature | Below nameplate |
|---|---|---|
| 0°C | 27°C | 0.7% |
| 5°C | 32°C | 2.5% |
| 10°C | 37°C | 4.3% |
| 15°C | 42°C | 6.1% |
| 20°C | 47°C | 7.9% |
| 25°C | 52°C | 9.7% |
| 30°C | 57°C | 11.5% |
| 35°C | 62°C | 13.3% |
Cell temperature estimated as air temperature plus 27°C, derived from a 42°C nominal operating cell temperature at 800 W/m² scaled to 1,000 W/m². Both the coefficient and the nominal operating temperature vary by module; read your own datasheet rather than these.
A 30°C afternoon costs around 11 to 12% against the nameplate. That is a real number and it is why inverter output plateaus below what owners expect during a hot spell.
Why the annual cost is much smaller
PVGIS reports temperature and low irradiance together as a single annual term, and for the roof-mounted reference array used across these guides it comes to 7.0%. That is less than the loss at a single hot afternoon hour, for a straightforward reason: British roofs spend very few hours above 25°C ambient and a great many hours in weak light.
The seasonal shape shows where the energy actually is.
| Month | Output (kWh) | Irradiation (kWh/m²) |
|---|---|---|
| January | 144 | 42.9 |
| February | 201 | 60.5 |
| March | 331 | 101.4 |
| April | 416 | 131.5 |
| May | 456 | 146.7 |
| June | 450 | 146.9 |
| July | 461 | 152.5 |
| August | 410 | 133.3 |
| September | 352 | 112.5 |
| October | 240 | 74.6 |
| November | 167 | 50.5 |
| December | 134 | 40 |
May, June and July are within about 2% of each other despite very different temperatures. The heat penalty in high summer offsets part of the extra daylight.
May, June and July land close together. June comes in slightly below both of its neighbours in this dataset. Whichever month tops the table in a given year, the point stands: past a certain amount of daylight, extra heat is cancelling part of the gain.
What changes it, and what does not
- Mounting and airflow. Panels held clear of the roof surface run cooler than panels sitting tight to it, and in-roof mounting, where modules replace tiles, runs hottest of all. This is already reflected in the 7.0% annual figure, which is modelled for building-mounted rather than free-standing.
- Module choice. The gap between a 0.29%/°C module and a 0.45%/°C one is over a third of the temperature loss. It is one of the few datasheet numbers worth comparing directly when two quotes differ.
- Cooling the panels with water. Not worth doing. The gain lasts as long as the water does, it is a few per cent of output during the few hours a year it would apply, and it puts someone on a roof with a hose.
If an array is underperforming its model across a whole year rather than on hot afternoons, temperature is not the explanation and the loss budget is the place to start looking.
Annual and monthly modelling: PVGIS 5.2 (European Commission Joint Research Centre), PVGIS-SARAH2 radiation database, 2005–2020 average. 4 kWp crystalline silicon, roof-mounted, 14% system loss, Birmingham, UK. Retrieved 2026-09-12. Module temperature characteristics from published manufacturer datasheets. Checked 2026. Source
Drag the air temperature. A panel is rated at a cell temperature of 25°C, which is not an air temperature and is almost never the temperature the cells are actually at.
Air
25°C
Cells
53°C
Of nameplate
90.1%
9.9% below nameplate, which is 28°C above the rating multiplied by 0.36%/°C.
Cell temperature is estimated as air temperature plus 27.5°C, derived from a nominal operating cell temperature of 42°C at 800 W/m² scaled to full sun at 1000 W/m².
Across a whole British year the combined temperature and low-irradiance loss is only 7% of output, because hot hours are rare and dim hours are not. The penalty above is an instantaneous one.
Coefficient and nominal operating cell temperature from the Canadian Solar KuMax CS3U-MS datasheet. Published coefficients across the market run roughly 0.29 to 0.45%/°C; read your own.