Do solar panels lose power in the heat? 10 to 23 percent in a hot Phoenix hour, about 5 to 9 percent over a year
Yes. Cells lose 0.24 to 0.377 percent of rated power for every degree Celsius above 25 on four current datasheets. In a Phoenix July hour as hot as the hottest tenth of 1 p.m. hours (107.1 °F), that removes 10 to 18 percent of the rating for the hour on an open rack or a standoff roof, and about 5 to 9 percent across a full year of sunlight.
The coefficient is the easy half. The hard half is how hot the cells get, because in full sun they run far above the 25 °C of the rating. Sandia National Laboratories' module temperature model puts cells 24 to 45 degrees Celsius above the air at 1,000 W/m2 and airport-average wind, depending on whether the panel hangs in open air or lies close to the roof. In that Phoenix hour the cell is at 67 °C on an open rack and 87 °C when mounted tight, 19 degrees apart for the same panel, the same sun and the same air. Moving from the cooler mount to the hotter one costs about as many percentage points as moving from a low-coefficient panel to a high-coefficient one.
A peak hour is also not a year. In a PVWatts typical year for Phoenix, 19 percent of the year's sunlight on a roof-mounted array arrives while the cells are at 60 °C or hotter, and 10 percent on an open rack. Most sunlight arrives with cooler cells, which is why a Silfab SIL-410 HC+ that gives up 17 percent in that hot hour on a standoff roof gives up roughly half as much, 8 percent, over twelve months. Neither figure includes the soiling, wiring, inverter and other losses a production estimate stacks on top.
What the datasheet coefficient promises
A module rating is a laboratory number. Silfab and REC both print it at standard test conditions: 1,000 W/m2 of sunlight, air mass 1.5 and a cell temperature of 25 °C. The line that matters for heat is "Temperature Coefficient Pmax", the percentage of that rating the module gives up for each degree Celsius the cells run above 25 (and gains for each degree below). REC states that its coefficients are linear values, and the arithmetic here treats all four sheets that way, so the power at irradiance E and cell temperature Tc is the rating multiplied by E/1,000 and by (1 + coefficient x (Tc - 25)). A Silfab SIL-410 HC+ at -0.36%/°C with cells at 65 °C is 40 degrees above the reference, so it delivers 1 - 0.0036 x 40 = 85.6 percent of 410 W, or 351 W, in full sun. With cells at 10 °C on a cold, bright morning the same arithmetic gives 105.4 percent, which is how a panel can briefly exceed its nameplate.
The four datasheets in the table were chosen because each prints a Pmax coefficient and a nominal operating temperature, and all four describe the same construction, a glass front and a polymer backsheet. They are not a ranking. Price, warranty, availability and wattage differ and are outside the table, and a module with a better coefficient is not automatically the better purchase. What the table does show is spread. At a cell temperature of 70 °C, 45 degrees above the reference, the gap between the lowest coefficient (-0.24) and the highest (-0.377) is 6.2 percentage points of output.
| Module | Cells and construction | STC efficiency | Pmax temperature coefficient | Nominal operating temperature | Datasheet edition code |
|---|---|---|---|---|---|
| REC Alpha Pure-RX, 450 to 470 W | 88 half-cut bifacial heterojunction cells, glass front, polymer backsheet | 21.6 to 22.6% | -0.24%/°C | 44 ± 2 °C (NMOT) | Footer 6.2026 V7, REF PM-DS-12-06/08.04.08.01 |
| Silfab SIL-430 QD | 108 half-cut N-type silicon cells, glass front, polymer backsheet | 22.1% | -0.29%/°C | 45 ± 2 °C (NOCT) | SIL-430-QD-20260804 |
| Silfab SIL-410 HC+ | 132 half-cut mono PERC cells, glass front, polymer backsheet | 20.7% | -0.36%/°C | 45 ± 2 °C (NOCT) | SIL-410 HC+-20260101 |
| Silfab SIL-420 BG | 66 mono PERC MWT cells, glass front, polymer backsheet | 21.9% | -0.377%/°C | 43.5 ± 2 °C (NOCT) | SIL-420-BG-20260101 |
PVWatts's 2014 planning values bracket the four sheets. The PVWatts Version 5 manual, published in September 2014, assigns -0.47%/°C to a "standard" module, -0.35%/°C to a "premium" one and -0.20%/°C to thin film, values it says came from a statistical analysis of more than 11,000 modules in the CEC module database. The standard value is worse than every sheet in the table, the thin-film value is better than every sheet, and only the premium value falls inside the range. Two of the four 2026 sheets beat the premium figure and two are slightly worse, so a -0.3%/°C rule of thumb describes the middle of these four sheets and not their ends. Efficiency does not predict the coefficient. The SIL-420 BG is more efficient than the SIL-410 HC+ (21.9 against 20.7 percent) and has the worse coefficient (-0.377 against -0.36). The cost per square foot guide covers what efficiency does change, the roof area a given wattage needs.
Two cautions apply to every number from a datasheet. The coefficient is measured near standard conditions and applied far from them, which REC states plainly: "The temperature coefficients stated are linear values". And the nominal operating temperature on the sheet, 43.5 to 45 °C for Silfab and 44 °C for REC, describes a mild reference day of 800 W/m2 and 20 °C air, not a Phoenix afternoon. REC prints a second figure that is closer to the hot case, a module operating temperature of 80 °C, defined on the sheet as the 98th percentile operating temperature, with no climate or mount attached to it.
How hot the cells actually get
Sunlight that is not converted to electricity becomes heat, and the cells settle at the temperature where heat leaving the module equals heat arriving. Three things set that balance: how much sun reaches the glass, how warm the air is, and how freely air moves over both faces. Sandia's module temperature model, published by the PV Performance Modeling Collaborative, reduces this to one equation for the back-surface temperature, Tm = E x e^(a + b x WS) + Ta, where E is the irradiance on the module in W/m2, Ta is the air temperature in °C and WS is the wind speed in m/s. The coefficient a sets how many degrees above the air the module sits in still air at full sun, and b sets how fast wind removes that excess. A second step converts back-surface temperature to cell temperature: Tc = Tm + (E/1,000) x ΔT, where ΔT is the small gap between the cell and the back of the module at 1,000 W/m2.
The coefficients come from Table 1 of the 2004 Sandia report by King, Boyson and Kratochvil, fitted to thousands of measurements taken on clear-sky days with the module in near thermal equilibrium and wind measured at the standard height of 10 meters. The PVPMC page lists a and b for the same rows; ΔT appears only in the report, which describes it as typically 2 to 3 °C for flat-plate modules on an open rack and as zero for a module with a thermally insulated back. The first four rows are the flat-plate constructions relevant to a house. The last two are a thin-film-on-steel module and a concentrator, shown for completeness.
| Module type | Mount | a | b | ΔT (°C) |
|---|---|---|---|---|
| Glass/cell/glass | Open rack | -3.47 | -0.0594 | 3 |
| Glass/cell/glass | Close roof mount | -2.98 | -0.0471 | 1 |
| Glass/cell/polymer sheet | Open rack | -3.56 | -0.0750 | 3 |
| Glass/cell/polymer sheet | Insulated back | -2.81 | -0.0455 | 0 |
| Polymer/thin-film/steel | Open rack | -3.58 | -0.113 | 3 |
| 22X linear concentrator | Tracker | -3.23 | -0.130 | 13 |
Two features of the table drive everything that follows. Between the open-rack and close-roof rows for glass/cell/glass, a changes from -3.47 to -2.98 and b from -0.0594 to -0.0471. At 1,000 W/m2 and 3 m/s that is the difference between e^(-3.47 - 0.0594 x 3) x 1,000 = 26 degrees and e^(-2.98 - 0.0471 x 3) x 1,000 = 44 degrees above the air, before the small ΔT step. The mount does not change the sunlight or the air; it changes how freely the heat escapes.
The second feature is accuracy. The report says the model gives the module operating temperature to about ±5 °C. In its words: "The simple model has proven to be very adaptable and entirely adequate for system engineering and design purposes by providing the expected module operating temperature with an accuracy of about ±5°C. Temperature uncertainties of this magnitude result in less than a 3% effect on the power output from the module." (SAND2004-3535, pages 17 and 18.) For the coefficients in the datasheet table, ±5 °C moves the loss by 1.2 to 1.9 percentage points, inside that 3 percent.
The four Silfab and REC modules are glass-and-backsheet products, so the glass/cell/polymer sheet rows are the nearest match. Sandia publishes an open-rack row for that construction and an insulated-back row, but no close-roof row. The only close-roof measurement in the table is for double-glass modules. That gap is why the roof cases below are presented as a range and not as one number.
Testing the roof rows against NOCT
A model that claims to describe roofs should reproduce what manufacturers and PVWatts already assume for roofs. The nominal operating conditions give a common yardstick: 800 W/m2, 20 °C air and 1 m/s wind, the definition REC prints for its NMOT. Silfab's sheets print the 800 W/m2 and air mass but not the air temperature or the wind, so the comparison assumes the conventional 20 °C and 1 m/s. Running each flat-plate Sandia row at those conditions gives the first column of the table.
| Sandia row | Cell temperature at NOCT conditions | Degrees above the 20 °C air | Reference value |
|---|---|---|---|
| Glass/cell/polymer sheet, open rack | 43.5 °C | 23.5 | Silfab NOCT 43.5 to 45 °C and REC NMOT 44 °C (datasheets); PVWatts open rack 45 °C |
| Glass/cell/glass, open rack | 45.9 °C | 25.9 | Same references; no double-glass module is among the four datasheets |
| Glass/cell/glass, close roof mount | 59.6 °C | 39.6 | PVWatts roof mount with 4-inch standoffs: about 49 °C |
| Glass/cell/polymer sheet, insulated back | 66.0 °C | 46.0 | No datasheet or PVWatts value; boundary case with no rear airflow |
The open-rack glass/polymer row lands at 43.5 °C, within 1.5 degrees of the Silfab and REC values of 43.5 to 45 °C. Agreement on that row says the equation and the unit conversions are right. The roof rows are another matter. The glass/glass close-roof row gives 59.6 °C and the insulated-back row 66.0 °C, 16 to 23 degrees hotter than the open rack. PVWatts takes a far smaller step. Its Version 5 manual keeps an installed nominal operating cell temperature of 45 °C for an open rack and estimates "roughly 49 C for 4 inch standoffs" on a roof, a difference of 4 degrees. PVWatts version 8 keeps the same two fixed-array choices, open rack and roof mounted, though its documentation page does not restate the temperatures.
The reading that fits both is that Sandia's close-roof and insulated-back rows describe modules lying nearly flat on the roof surface or built into it, with little or no airflow behind them, and that a residential rack with a few inches of gap sits between those rows and the open rack. The report itself says the model "is appropriate for other mounting configurations, for example directly attached to a roof surface, but different coefficients appropriate to the situation must be determined." None of the six Sandia rows is labelled for a rack with a four-inch standoff.
The hot-hour tables therefore use three cases. Open rack is Sandia's glass/polymer row. A standoff roof adds 6 degrees to the open-rack cell temperature, the average gap between PVWatts's roof-mounted and open-rack runs at the hottest hour of the typical year in four cities (6.0, 6.0, 5.8, 6.3 degrees for Phoenix, Houston, Atlanta and Denver). That is larger than the 4-degree gap between the nominal ratings, and the hottest hours in those files are bright (964 to 1,072 W/m2) and nearly windless (0.3 to 0.5 m/s). A tight roof is Sandia's glass/glass close-roof row, the only row in the table labelled for a roof. The insulated-back row is shown as an outer bound and is not a case for a standard rack.
Six July afternoons from NOAA normals
The air temperature and wind come from the National Centers for Environmental Information's 1991 to 2020 U.S. Climate Normals, which NCEI describes as the latest in a series of decadal normals. Six airport stations give a spread of climates: a desert, a Gulf coast city, a Central Valley city, the Southeast, a high-plains city and a cool coastal one. The monthly normals set the scene, showing how hot July is on average and how many days reach 90 and 100 °F.
| City (station) | July average daily high | July average temperature | Days at or above 90 °F | Days at or above 100 °F |
|---|---|---|---|---|
| Phoenix (Phoenix Sky Harbor International Airport, AZ) | 106.5 °F (41.4 °C) | 95.5 °F | 30.8 | 28.5 |
| Houston (Houston Intercontinental Airport, TX) | 94.5 °F (34.7 °C) | 85.1 °F | 27.3 | 1.8 |
| Sacramento (Sacramento Executive Airport, CA) | 92.6 °F (33.7 °C) | 75.9 °F | 20.6 | 4.8 |
| Atlanta (Atlanta Hartsfield-Jackson International Airport, GA) | 90.1 °F (32.3 °C) | 80.9 °F | 16.4 | 0.4 |
| Denver (Denver International Airport, CO) | 89.9 °F (32.2 °C) | 75.1 °F | 18.8 | 1.3 |
| Seattle (Seattle-Tacoma International Airport, WA) | 77.4 °F (25.2 °C) | 67.1 °F | 1.8 | 0.1 |
Phoenix averages 28.5 days at or above 100 °F in July, Houston 1.8 and Seattle 0.1. The monthly high is not the hour that matters, though. It is the average of each day's afternoon peak, and the cell temperature depends on the sun and wind in the same hour as the heat. The hourly normals supply that hour.
The hour itself: air, wind and sun
NCEI computes each hourly normal from "450 possible values": the same date and hour plus or minus 7 days, across the 30 years of the period (26 of them with enough data at Denver International), reported in local standard time. Besides the normal, the hourly product gives a 90th percentile, the temperature that one hour in ten of that date and hour exceeded. The tables use the 90th percentile as the hot afternoon, together with the average wind from the same product, for 1 p.m. local standard time on July 15. In Phoenix that hot hour is 107.1 °F, above the month's average daily high of 106.5 °F.
| City | Normal temperature | Hotter-than-normal hour (90th percentile) | Average wind | Hours with wind of 3 mph or less |
|---|---|---|---|---|
| Phoenix | 101.5 °F (38.6 °C) | 107.1 °F (41.7 °C) | 6.7 mph (3.00 m/s) | 5.8% |
| Houston | 90.7 °F (32.6 °C) | 95.0 °F (35.0 °C) | 6.9 mph (3.08 m/s) | 9.1% |
| Sacramento | 87.6 °F (30.9 °C) | 97.0 °F (36.1 °C) | 8.3 mph (3.71 m/s) | 3.1% |
| Atlanta | 86.6 °F (30.3 °C) | 91.9 °F (33.3 °C) | 8.0 mph (3.58 m/s) | 7.5% |
| Denver | 87.5 °F (30.8 °C) | 96.1 °F (35.6 °C) | 8.6 mph (3.84 m/s) | 4.0% |
| Seattle | 72.0 °F (22.2 °C) | 81.9 °F (27.7 °C) | 8.4 mph (3.76 m/s) | 1.8% |
Sunlight is an input, not a measurement. In the PVWatts typical-year files for Phoenix, Houston, Atlanta and Denver, the 1 p.m. hour in July had median irradiance of 812 to 962 W/m2 on a south-facing 20 degree plane, and maxima of 955 to 1,010 W/m2 across the 31 days, so 1,000 W/m2 describes a clear hour and not an average one. A bright but hazy hour near 800 W/m2 is worked as a sensitivity below.
Wind is the input to treat with the most suspicion. The normals are airport observations, and the Sandia coefficients were fitted against wind at a 10 meter reference height; airport wind here is an input and not a measurement at any roof. The right-hand column shows how often the wind is 3 mph (1.3 m/s) or less, which NCEI counts as calm: from 1.8 percent of 1 p.m. hours in Seattle to 9.1 percent in Houston. Calm hours are rare at an airport in the early afternoon, yet the hottest hours in the PVWatts typical-year files are nearly windless, and a roof in a quiet neighborhood can be calmer than an airport, so the calm case is worked separately.
Cell temperature on three kinds of mount
Putting the hot-hour air and average wind into the model gives the cell temperatures in the table. On the open rack the six cities range from 52 to 67 °C, with Phoenix the hottest at 67.4 °C. The standoff roof adds 6 degrees by construction, and the tight roof adds about 19 degrees in every city, since the wind shifts it by less than a degree. Seattle's tight-roof cells at 71.3 °C are hotter than Houston's open-rack cells at 60.6 °C: a mount that traps heat can outweigh a 7 degree advantage in air temperature.
| City | Air | Open rack, glass/polymer | Standoff roof (open rack + 6 °C) | Tight roof, glass/glass | Insulated back |
|---|---|---|---|---|---|
| Phoenix | 41.7 | 67.4 | 73.4 | 86.8 | 94.3 |
| Houston | 35.0 | 60.6 | 66.6 | 79.9 | 87.3 |
| Sacramento | 36.1 | 60.6 | 66.6 | 79.8 | 87.0 |
| Atlanta | 33.3 | 58.0 | 64.0 | 77.2 | 84.4 |
| Denver | 35.6 | 59.9 | 65.9 | 79.0 | 86.2 |
| Seattle | 27.7 | 52.2 | 58.2 | 71.3 | 78.5 |
Wind and sunlight move these numbers more than most people expect. Holding Phoenix's hot-hour air temperature fixed, an open-rack cell is 6.8 degrees cooler in a 5 m/s breeze than in a 1 m/s one. At the calm boundary of 1.3 m/s, the Phoenix open-rack cell reaches 70.5 °C and the tight-roof cell 90.5 °C. The PVWatts typical years contain hot hours with wind of 0.3 to 0.5 m/s whose cell temperature reaches 70 to 73 °C on an open rack and 76 to 80 °C on a roof. Using 800 W/m2 in place of 1,000 W/m2 takes Phoenix's open-rack cell down to 62.3 °C and the tight-roof cell to 77.8 °C.
A heat wave adds degrees directly, because each degree of air temperature is a degree of cell temperature in this model. The table already uses a hotter-tenth air temperature, so it describes a warm day and not a record. The model's ±5 °C accuracy applies on top of all of it. For a Phoenix roof the defensible summary is a range: a cell near 67 °C on an open rack and 73 °C on a standoff roof as central cases, and 87 °C at the tight-roof extreme, all in a hotter-than-normal hour. REC's 98th-percentile operating temperature of 80 °C sits inside that range, a cross-check from a manufacturer and not from a model. The Silfab sheets print an operating temperature of -40 to +85 °C; the open-rack and standoff cases fall inside it and the tight-roof extreme of 87 °C does not.
One Phoenix hour, in kilowatts
Take an 8 kW DC array in Phoenix at 1 p.m. local standard time on a hotter-than-normal July day, with 1,000 W/m2 on the array. At the standoff roof's 73.4 °C the cells are 48.4 degrees above the reference. For the Silfab SIL-410 HC+ the factor is 1 - 0.0036 x 48.4 = 0.8256, so the array delivers 6.60 kW at its terminals, 1.40 kW below the 8 kW rating. For the REC Alpha Pure-RX at -0.24%/°C the factor is 0.8837 and the output is 7.07 kW. Same sun, same roof, same air: 0.47 kW more, or 7.0 percent, from the lower coefficient alone.
| Module (coefficient) | Open rack | Standoff roof | Tight roof | Insulated back |
|---|---|---|---|---|
| REC Alpha Pure-RX (-0.24%/°C) | 10.2%, 7.19 kW | 11.6%, 7.07 kW | 14.8%, 6.81 kW | 16.6%, 6.67 kW |
| Silfab SIL-430 QD (-0.29%/°C) | 12.3%, 7.02 kW | 14.0%, 6.88 kW | 17.9%, 6.57 kW | 20.1%, 6.39 kW |
| Silfab SIL-410 HC+ (-0.36%/°C) | 15.3%, 6.78 kW | 17.4%, 6.60 kW | 22.3%, 6.22 kW | 24.9%, 6.01 kW |
| Silfab SIL-420 BG (-0.377%/°C) | 16.0%, 6.72 kW | 18.3%, 6.54 kW | 23.3%, 6.14 kW | 26.1%, 5.91 kW |
Show the bar values as a table
| Mount | Cell temperature | REC Alpha Pure-RX (-0.24%/°C) | Silfab SIL-410 HC+ (-0.36%/°C) |
|---|---|---|---|
| Open rack | 67.4 °C | 10.2% | 15.3% |
| Standoff roof | 73.4 °C | 11.6% | 17.4% |
| Tight roof | 86.8 °C | 14.8% | 22.3% |
| Insulated back | 94.3 °C | 16.6% | 24.9% |
The table's power figures stop at the module terminals. Inverter losses, wiring, soiling, mismatch and the rest of the PVWatts loss stack come off afterward, and so does inverter clipping, a separate effect covered in the production estimate guide. What the table isolates is the heat term.
Mount and module are comparable levers. Moving the REC array from an open rack to a tight roof costs 4.7 percentage points of the hour's output, and the SIL-410 array 7.0. Swapping the highest-coefficient module for the lowest at a standoff roof is worth 6.6 points. Both are single-digit points of one hour's output. The sizing guide covers the loss stack an estimate carries around them.
The same hour in five more cities
The Phoenix result generalizes in a predictable way. A cooler city shifts every case down, but a hot mount can erase that advantage. Seattle's tight-roof SIL-410 loses 16.7 percent in its hotter-than-normal July hour, more than a Phoenix SIL-410 on an open rack, which loses 15.3 percent. Houston, Sacramento, Atlanta and Denver fall between, with open-rack losses of 8 to 13 percent across the lowest-coefficient and highest-coefficient datasheets.
| City | Open rack: REC / SIL-420 | Standoff roof: REC / SIL-420 | Tight roof: REC / SIL-420 |
|---|---|---|---|
| Phoenix | 10.2% / 16.0% | 11.6% / 18.3% | 14.8% / 23.3% |
| Houston | 8.5% / 13.4% | 10.0% / 15.7% | 13.2% / 20.7% |
| Sacramento | 8.6% / 13.4% | 10.0% / 15.7% | 13.1% / 20.6% |
| Atlanta | 7.9% / 12.5% | 9.4% / 14.7% | 12.5% / 19.7% |
| Denver | 8.4% / 13.2% | 9.8% / 15.4% | 13.0% / 20.4% |
| Seattle | 6.5% / 10.2% | 8.0% / 12.5% | 11.1% / 17.4% |
These are hot-hour losses, not annual ones, and they assume 1,000 W/m2. They show that heat is a real term even in Seattle's cool July, a large one on Phoenix summer afternoons, and that the mount is a variable worth asking about in either. They do not rank the cities on annual energy, because Phoenix also has far more sunlight: the PVWatts run gives it 2,362 kWh/m2 a year on the array plane against 1,891 in Houston and 1,888 in Atlanta. The lower-than-expected guide covers how to separate a seasonal dip from a fault.
A peak hour is not a year
Annual energy needs a full-year model, and the hot-hour numbers above are not one. A proper model needs hourly irradiance, air temperature and wind for the site, a thermal model run for every hour, and an electrical model with the module's own parameters. PVWatts is the free tool that does the first two for a typical year. Two PVWatts version 8 runs per location (open rack and roof mount, 8 kW DC, 20 degree tilt, south-facing, 14 percent system losses) return hourly cell temperature and plane-of-array irradiance, which allows a direct statement about how much of the year's sunlight arrives hot. The runs use the NOAA station coordinates for Phoenix, Houston, Atlanta and Denver International.
Weighting the hourly cell temperature by hourly irradiance gives the temperature at which the average kilowatt-hour of sunlight is converted: 45.5 °C on an open rack and 48.3 °C on a roof in Phoenix, 40.8 and 43.4 °C in Houston, 39.1 and 42.1 °C in Atlanta, 34.4 and 37.1 °C in Denver. Multiplying the excess over 25 °C by a datasheet coefficient gives a linear approximation of the annual temperature loss. In Phoenix it is 4.9 percent for the REC on an open rack and 8.8 percent for the SIL-420 on a roof, against 11.6 to 18.3 percent in the hot hour on a standoff roof. For the SIL-410 on a roof the hot hour is 2.1 times the annual figure in Phoenix (17.4 against 8.4 percent), 2.3 times in Houston and 2.3 times in Atlanta.
| Location and mount | Irradiance-weighted mean cell temperature | Hottest hour in the file | Share of the year's sunlight arriving with cells at 60 °C or more | Approximate annual loss, REC (-0.24) to SIL-420 (-0.377) |
|---|---|---|---|---|
| Phoenix, open rack | 45.5 °C | 72.2 °C | 9.7% | 4.9% to 7.7% |
| Phoenix, roof mount | 48.3 °C | 78.1 °C | 19.1% | 5.6% to 8.8% |
| Houston, open rack | 40.8 °C | 69.6 °C | 4.0% | 3.8% to 5.9% |
| Houston, roof mount | 43.4 °C | 75.6 °C | 9.5% | 4.4% to 6.9% |
| Atlanta, open rack | 39.1 °C | 72.4 °C | 4.7% | 3.4% to 5.3% |
| Atlanta, roof mount | 42.1 °C | 78.2 °C | 11.3% | 4.1% to 6.4% |
| Denver, open rack | 34.4 °C | 73.3 °C | 3.2% | 2.3% to 3.6% |
| Denver, roof mount | 37.1 °C | 79.5 °C | 6.9% | 2.9% to 4.6% |
Show the monthly values as a table
| Month | Open rack | Roof mount |
|---|---|---|
| Jan | 0.789 | 0.780 |
| Feb | 0.788 | 0.779 |
| Mar | 0.763 | 0.753 |
| Apr | 0.752 | 0.743 |
| May | 0.731 | 0.722 |
| Jun | 0.711 | 0.702 |
| Jul | 0.711 | 0.703 |
| Aug | 0.712 | 0.702 |
| Sep | 0.723 | 0.714 |
| Oct | 0.753 | 0.743 |
| Nov | 0.776 | 0.767 |
| Dec | 0.804 | 0.795 |
In all four locations the roof run is warmer than the open-rack run, by 2.6 to 3.0 degrees on the irradiance-weighted average and 5.8 to 6.3 degrees at the hottest hour. Over a year the roof penalty is therefore about 0.6 to 1.1 percentage points across the coefficient range, a real but modest term. In the hot hour the same 6-degree step costs 1.4 to 2.3 points. The two statements do not conflict. They describe the same physics weighted differently: a few hours carry a large loss and most hours carry a small one.
The seasonal shape is why a hot summer is not a bad summer. The Phoenix run's monthly output per unit of plane-of-array sunlight, in the figure, is 0.804 in December and 0.711 in June and July on an open rack, a dip of 11.6 percent. The run does not separate temperature from the other seasonal terms, such as the sun's angle on the glass, but temperature is the one that follows the thermometer. The dip is smaller than the swing in sunlight: the same run produces 1,433 kWh in May and 849 kWh in December. In Phoenix, heat trims a summer month and does not reverse the ranking of summer and winter.
Four limits keep these annual numbers honest. The approximation applies each datasheet's coefficient to PVWatts's cell temperatures, whereas PVWatts itself uses its own module-type coefficient and its own thermal model, so PVWatts's own annual temperature loss would differ. The two mount settings also differ in plane-of-array sunlight by 0.6 to 0.9 percent, so the AC energy gap between the two runs is not attributable to temperature alone and is not quoted. Sacramento and Seattle have a hot-hour result and no annual one. And a typical year smooths out hot years, so any particular summer will land above or below these averages. The default loss line of 14 percent in the PVWatts Version 5 manual has no temperature item; the guide to the losses line lists the ten items that make it up, and heat is modelled separately, hour by hour.
What the mount and the module change
Both levers are decided once, at installation, and they are documented very differently. The module choice comes with a datasheet: the coefficient, the nominal operating temperature and the efficiency are printed, and the numbers above show a lower coefficient is worth about 7 points of output in the hottest hours on a standoff roof. The mount choice can come with no number at all. A quote can say "roof mount" without the height of the standoff, the clearance behind the modules or the temperature model behind the production estimate. The ground versus roof mount guide covers the other trade-offs between mounting types; heat is one more to add to that comparison, with the open rack at the cool end of the table.
Heat does not outweigh sunlight in the four locations run. The hottest, Phoenix, produced the most: 14,005 kWh for the 8 kW roof-mounted run against 11,437 kWh for Atlanta, with the same system losses applied. Heat also has a slower cost. The guide to panel lifespan covers the evidence that hotter operating conditions are one of the steady pressures behind faster degradation, which no hot-hour or annual temperature figure here captures.
Heat in your own production estimate
A production estimate either models heat or buries it. PVWatts models it: its array-type input offers "Fixed - Open Rack" and "Fixed - Roof Mounted", and its module-type input offers standard, premium and thin film, each with a built-in coefficient. A proposal built in another tool may use the module's own coefficient and a thermal model the proposal never names. Ask which was used. The answer decides whether the printed annual figure already contains the roof penalty of about 0.6 to 1.1 points shown above or ignores it.
A cheap cross-check uses the same arithmetic. Take the module's coefficient from its datasheet, take a July afternoon cell temperature for the nearest city from the tables, and compute the hot-hour percentage. A production reading from a bright July afternoon that sits far below the datasheet-based figure points to a cause other than heat, such as soiling, shade, clipping or an inverter limit. The shortfall diagnostic guide sorts those causes by the shape of the curve.
Four heat checks for the installer
- Which module model is on the proposal, and what is the Pmax temperature coefficient on its current datasheet?
- What are the standoff height and the clearance behind the modules, and does the production estimate use open-rack or roof-mount settings?
- Which tool, and which module-type or coefficient setting, produced the annual kWh figure?
- Does the estimate state a nominal operating temperature, and is it the manufacturer's value or a default?
Solar panel heat questions
Do solar panels stop working when it is too hot?
No. Output falls gradually with cell temperature at the datasheet's coefficient, 0.24 to 0.377 percent per degree on the four sheets compared here. In the Sandia model a Phoenix open-rack cell at 67 °C in a hotter-than-normal July hour gives 10 to 16 percent less than the rating for that hour. Silfab prints an operating temperature of -40 to +85 °C and REC a 98th-percentile operating temperature of 80 °C; the open-rack and standoff cases (67 and 73 °C) are inside both, and the tight-roof extreme of 87 °C is above both.
How much hotter than the air does a solar panel get?
In the Sandia model at 1,000 W/m2 and airport-average wind, about 24 to 26 degrees Celsius hotter than the air on an open rack, and up to 45 degrees hotter for a module mounted tight to the roof. The rise shrinks with wind and with lower sunlight and grows with a tighter mount. The model's accuracy is about ±5 °C.
Is a roof-mounted panel hotter than a panel on an open rack?
Yes. In PVWatts typical-year runs for Phoenix, Houston, Atlanta and Denver, the roof setting raises the irradiance-weighted cell temperature by 2.6 to 3.0 degrees and the hottest hour by 5.8 to 6.3 degrees. How much hotter in practice depends on the standoff gap, which Sandia's published rows do not cover for a typical rack.
Which temperature coefficient is good?
Closer to zero is better. Among the four 2026 datasheets compared, the range is -0.24 to -0.377 percent per degree Celsius. The 2014 PVWatts manual used -0.35 for a premium module and -0.47 for a standard one. Efficiency, price, warranty and wattage differ between modules, so the coefficient is one input to a comparison, not a ranking.
How much energy does heat cost over a year?
With the linear approximation described above and PVWatts typical-year cell temperatures, between 5 and 9 percent in Phoenix across the four datasheets, 4 to 7 percent in Houston and 2 to 5 percent in Denver. A full annual loss for your roof needs a roof-specific PVWatts or installer model, which this page does not replace.
Do cold days make solar panels more efficient?
Per unit of sunlight, yes. A cell at 10 °C on the SIL-410 HC+ coefficient delivers 105.4 percent of its rating at 1,000 W/m2. In the Phoenix PVWatts run, December has the highest output per unit of plane-of-array sunlight (0.804 against 0.711 in June), while December's 849 kWh is well below May's 1,433 kWh, because the sunlight is smaller.
Two lookups frame the question for a quote: the Pmax temperature coefficient and nominal operating temperature on the module's datasheet, and the mount settings behind the production estimate. Set them beside the hot-hour and annual figures here. The result is not by itself a reason to reject a quote. It is a reason to ask one more specific question.
Sources and verification
- D. L. King, W. E. Boyson and J. A. Kratochvil, Photovoltaic Array Performance Model, Sandia National Laboratories SAND2004-3535 (December 2004), pages 17 to 20 and Table 1: Thermal model equations 11 and 12, Table 1 coefficients a, b and ΔT, the ±5 °C accuracy statement, wind at 10 meters, and the statement that other mounts need their own coefficients. Checked November 3, 2026.
- PV Performance Modeling Collaborative (Sandia), Sandia Module Temperature Model: Equation Tm = E x e^(a + b x WS) + Ta and the table of a and b by module type and mount. Checked November 3, 2026.
- PV Performance Modeling Collaborative (Sandia), Sandia Cell Temperature Model: Tc = Tm + E/E0 x ΔT with E0 = 1,000 W/m2. Checked November 3, 2026.
- REC Alpha Pure-RX datasheet, 450 to 470 W, US edition: Pmax coefficient -0.24%/°C, NMOT 44 ± 2 °C defined at 800 W/m2, 20 °C and 1 m/s, T98 operating temperature 80 °C, efficiency 21.6 to 22.6%; footer 6.2026 V7, REF PM-DS-12-06/08.04.08.01. Checked November 3, 2026.
- Silfab Solar SIL-430 QD datasheet, edition 20260804: Pmax coefficient -0.29%/°C, NOCT 45 ± 2 °C, 22.1% efficiency, 108 half-cut N-type cells. Checked November 3, 2026.
- Silfab Solar SIL-410 HC+ datasheet, edition 20260101: Pmax coefficient -0.36%/°C, NOCT 45 ± 2 °C, 20.7% efficiency, 132 half-cut mono PERC cells, operating temperature -40 to +85 °C. Checked November 3, 2026.
- Silfab Solar SIL-420 BG datasheet, edition 20260101: Pmax coefficient -0.377%/°C, NOCT 43.5 ± 2 °C, 21.9% efficiency, 66 mono PERC MWT cells. Checked November 3, 2026.
- NOAA National Centers for Environmental Information, 1991 to 2020 U.S. Climate Normals, monthly (NCEI Data Service, dataset normals-monthly-1991-2020): July values for stations USW00023183 (Phoenix), USW00012960 (Houston), USW00023232 (Sacramento), USW00013874 (Atlanta), USW00003017 (Denver International) and USW00024233 (Seattle); the URL shows the Phoenix query. Checked November 3, 2026.
- NOAA NCEI, 1991 to 2020 Hourly Normals, station files, and Hourly Normals Documentation: Station files at ncei.noaa.gov/data/normals-hourly/1991-2020/access/
.csv for USW00023183, USW00012960, USW00023232, USW00013874, USW00003017 and USW00024233: HLY-TEMP-NORMAL, HLY-TEMP-90PCTL, HLY-WIND-AVGSPD (miles per hour) and HLY-WIND-PCTCLM (percent) at 07-15T13:00:00; 450 possible values per hourly normal; calm defined as 3 mph (1.3 m/s) or less. Checked November 3, 2026. - NOAA NCEI, U.S. Climate Normals overview: Describes 1991 to 2020 as the latest in the series of decadal U.S. normals. Checked November 3, 2026.
- A. P. Dobos, PVWatts Version 5 Manual, NREL/TP-6A20-62641 (September 2014): Module types and coefficients (Table 3: -0.47, -0.35 and -0.20 %/°C), Fuentes thermal model, installed nominal operating cell temperature of 45 °C for an open rack and roughly 49 °C for 4-inch standoffs, loss items (Table 6). Checked November 3, 2026.
- National Laboratory of the Rockies (formerly NREL), PVWatts API version 8: Eight hourly runs (software version 8.5.0): 8 kW DC, standard module, array type 0 (open rack) and 1 (roof mounted), 14 percent losses, 20 degree tilt, 180 degree azimuth, DC to AC ratio 1.2, 96 percent inverter, NSRDB typical-year weather, at the NOAA station coordinates for Phoenix, Houston, Atlanta and Denver. Checked November 3, 2026.
The Sandia coefficients are generic fits from 2004 and the report states that other mounting configurations need their own; the standoff roof case is the open-rack temperature plus 6 degrees taken from PVWatts, an assumption and not a Sandia measurement. Irradiance of 1,000 W/m2 and wind taken from airport normals are inputs, not measurements of any roof. The hot hour is the 90th-percentile hourly temperature for July 15 plus or minus 7 days over 1991 to 2020, not a forecast or a record. The module comparison covers four datasheets from two manufacturers and is not a ranking. PVWatts runs cover Phoenix, Houston, Atlanta and Denver; the hot-hour tables cover Phoenix, Houston, Sacramento, Atlanta, Denver and Seattle, so Sacramento and Seattle have no PVWatts run. Denver's hourly normals rest on 26 years of data, the others on 30. Annual loss figures are linear approximations and not a full-year production model. Your module datasheet, your production estimate and your installer's mounting drawing control. Nothing here is a quote, a forecast or financial advice.