A panel is rated at 25 degrees. It works at 53.
The wattage on a solar panel's datasheet is measured at Standard Test Conditions: a cell temperature of 25 degrees Celsius. A working panel on a real roof runs far hotter, because it sits in the sun absorbing the very irradiance it converts. Across the 40 markets in Solar Analytica's yield baseline, the modelled annual operating cell temperature averages 53.1 degrees, and reaches 65.8 degrees in the hottest sites (Bamako in the Sahel and Jodhpur in Rajasthan). Every degree above 25 shaves output: mainstream mono-silicon (PERC-class) datasheets state a power temperature coefficient of roughly a third of a percent per degree, and heterojunction (HJT) cells roughly a quarter of a percent, which is the main reason HJT commands a premium in hot markets.
Priced across a full year of weather, that gap between the rating plate and the roof is the static heat tax, and it is far larger than most buyers are ever shown. The average market in the baseline loses 9.8% of nameplate output to cell temperature with PERC-class modules, and 6.7% with HJT. The hottest markets lose more than an entire seventh:
| Market | Cell temp °C | Heat loss, PERC | Heat loss, HJT | Days above 35°C/yr |
|---|---|---|---|---|
| Bamako (Sahel), Mali | 65.8 | 14.3% | 9.8% | 165 |
| Jodhpur (Rajasthan), India | 65.8 | 14.3% | 9.8% | 180 |
| São Paulo, Brazil | 56.8 | 11.1% | 7.6% | 2 |
| Seville, Spain | 56.6 | 11.1% | 7.6% | 68 |
| Perth, Australia | 53.3 | 9.9% | 6.8% | 2 |
| Berlin, Germany | 45.4 | 7.1% | 4.9% | 1 |
| Helsinki, Finland | 39.2 | 5.0% | 3.4% | 0 |
| Reykjavik, Iceland | 37.7 | 4.4% | 3.0% | 0 |
Solar Analytica yield model: modelled annual operating cell temperature and the resulting temperature-coefficient loss against nameplate, from NASA MERRA-2 (1994-2024) temperatures and NASA CERES plane-of-array irradiance. Days above 35°C are 30-year averages, rounded. Set averages across all 40 markets: 53.1°C, 9.8% (PERC), 6.7% (HJT).
The record: two decades of measured warming
The static tax is priced against today's climate. The second question is how that climate has moved. NASA's GISTEMP v4 global surface record, held in our substrate and verified against the published NASA table, puts the global anomaly at +0.53°C in 2004 and +1.28°C in 2024, the warmest year in the instrumental record, against the 1951-1980 baseline. Averaged in five-year windows to strip single-year noise, the global anomaly rose from +0.61°C (2004-2008) to +1.08°C (2021-2025).
Global is not local, and the difference matters for solar. Fitting a linear trend to each market's own 30-year MERRA-2 temperature series (1994-2024), the 40 baseline markets warm at +0.39°C per decade on average, but the spread is wide: Berlin warms at +0.95°C per decade, Seoul at +0.80, São Paulo at +0.76, while four markets in the set (Cape Town, Bamako, Reykjavik and Jodhpur) show a flat-to-negative local trend within this specific 30-year window. A local 30-year fit is an observation, not a guarantee of the next 25 years, but it is a far better planning input than a global average applied everywhere.
The warming penalty, market by market
Projecting each market's own trend forward across a 25-year system life and pricing it through the module's temperature coefficient gives the warming penalty: the additional heat loss a system faces in its final year, beyond the static tax it was born with.
| Market | Local trend °C/decade | Added loss by year 25, PERC | Added loss, HJT |
|---|---|---|---|
| Berlin, Germany | +0.95 | 0.83% | 0.57% |
| Seoul, South Korea | +0.80 | 0.70% | 0.48% |
| São Paulo, Brazil | +0.76 | 0.67% | 0.46% |
| Albuquerque, USA | +0.74 | 0.65% | 0.44% |
| Madrid, Spain | +0.72 | 0.63% | 0.43% |
| Perth, Australia | +0.38 | 0.33% | 0.23% |
| Jodhpur (Rajasthan), India | -0.17 | -0.15% | -0.10% |
| Cape Town, South Africa | -0.38 | -0.33% | -0.23% |
Solar Analytica yield model: each market's 1994-2024 MERRA-2 linear temperature trend, projected 25 years and converted to output loss through the module temperature coefficient. Set averages: +0.34% (PERC), +0.23% (HJT). Negative values reflect a negative local trend in the observation window, not a forecast of cooling.
Two things stand out. First, the size: the average warming penalty is about a third of a percent by year 25, roughly one thirtieth of the static heat tax the same systems already pay today. Second, the geography: the largest penalties land in temperate mid-latitude cities, not in the deserts. The hottest markets pay the largest static tax and, in this observation window, some of the smallest warming drifts.
The question the buyer is actually asking
"Will warming hurt my solar system?" is really two questions with opposite answers. How much does heat cost me? A lot: around a tenth of nameplate in the average market, more than a seventh in the hottest, every year, from day one. How much more will warming add? Very little per system lifetime: a third of a percent on average by year 25, under one percent even in the fastest-warming market in the set.
Priced in money through our retail-tariff and yield join, the contrast sharpens. In Berlin, the fastest-warming market in the baseline, the warming penalty at year 25 costs about 4 US dollars per kilowatt of panels per year at 2025 retail prices. In Perth it is about 1 dollar 20. Choosing an HJT module over a PERC-class module in the same two markets is worth about 11 and 13 dollars per kilowatt per year respectively, three to ten times the warming drift, because it attacks the full static tax rather than the marginal creep. Annual module degradation, typically warranted around 0.4% per year, also outweighs the warming drift roughly tenfold over a system's life.
The practical order of operations for a buyer follows directly. First, reject any production estimate quoted at the nameplate rating: insist on a yield figure modelled at operating cell temperature for your location, because the 9.8% average gap is the largest single correction in this report. Second, in hot markets, treat the temperature coefficient as a specification to shop on, because it is the one heat variable you control. Third, include the warming drift in a 25-year cash-flow model because it is real and measurable, but treat it as a decimal place, not a reason to hesitate: over two decades in which the global anomaly rose three quarters of a degree, solar became the cheapest new electricity in most of the world anyway.
Method and limits
The figures here come from Solar Analytica's location yield baseline: 40 global markets, each carrying a 30-year NASA MERRA-2 temperature series (1994-2024), NASA CERES satellite irradiance, a cell-temperature model of the plane-of-array environment, and datasheet-class temperature coefficients for PERC and HJT module families. The warming penalty is that market's linear 30-year trend, projected 25 years and passed through the coefficient. Three limits matter. A local linear fit over one 30-year window is an observation with real variance, and the four negative-trend markets illustrate exactly that; they should be read as flat, not as cooling forecasts. The model holds irradiance constant while varying temperature; brightening and dimming of the resource itself is a separate variable with its own 20-year record, covered in part 02 of this series. And soiling, smoke and extreme events sit outside this model entirely; they are parts 05, 06 and 08. Every input series is public and named, so the numbers can be recomputed.