17 October 2025
“Solar Doesn’t Work on Cloudy Days”: True or False?
The short answer, from two years of daily data
Solar panels keep working under cloud, because they turn light into electricity and cloud dims light rather than switching it off. How much light gets through is a measured quantity, and for this town it has been measured every day. NASA’s POWER service publishes, for any point on earth, the sunlight that reached the ground each day and the sunlight that would have reached it under a clear sky. For San Fernando, Pampanga, at 15.03 north and 120.69 east, across every day of 2023 and 2024, the ratio between the two says this.
On the median day, 81 percent of clear-sky sunlight arrived. In the rainy months of July to September the median day fell to 71 percent, and the worst tenth of those days got 31 percent or less. On a storm day it falls to almost nothing: 7 percent on 24 July 2024, and 10 and 11 percent on 5 and 6 September 2024.
So the myth is false on an ordinary day and nearly true on a storm day, and the honest page is the one that says both.
What the figure is, and what it is not
Three limits, and the page is only honest with all three on it. First, the ratio is sunlight arriving at the ground, which engineers call irradiance. It is the input to a solar system and not its output. Panel output tracks it closely but not exactly, because panels lose output in heat, the inverter has its own behaviour, and angle, shading, dust and wiring take their share. None of that is in the NASA figure, and a day at 81 percent of clear-sky sunlight is not a promise of 81 percent of the system’s best output. The performance ratio on the calculator page is the separate allowance for what the system does with the light.
Second, the figures are an area average. NASA’s grid cell is about half a degree of latitude by five eighths of a degree of longitude, roughly 55 by 65 kilometres, so the numbers describe the sky over a district rather than over one roof. Third, they are daily totals. A whole day at 71 percent can hold an hour of downpour at a small fraction of that, so a reader expecting their panels to hold at 71 percent through the rain itself will be disappointed. The daily figure smooths over the worst hours, and the worst hours are when people look at the meter.
The solar panel calculator, and the performance ratio it allows for losses
A typical day, the rainy season, and a storm
Across the 731 days, 488 reached more than three quarters of clear-sky sunlight, 64 fell below half, and 19 fell below a quarter. The ten worst days all fell in July or September. In the dry months from November to April the median day stood at 88 percent; in the wet months from May to October it was 75 percent. The two worst stretches were 24 and 25 July 2024, at 7 and 13 percent, and 4 to 6 September 2024, at 18, 10 and 11 percent. On days like those a grid-tied system makes little and the house runs on the utility, and a battery that was full the evening before carries the night but is not refilled the next day.
The yearly figure a system is sized on already carries all of this. Over the two years the sun over San Fernando averaged 5.10 kilowatt-hours per square metre a day with the cloud in, against 6.61 if every day had been clear. That is a check on the other figure this site uses: the Global Solar Atlas, read on 23 September 2026, puts the long-term average at the same point at 5.23 a day, from a different method, a Solargis model at a single set of coordinates averaged over many years rather than a satellite grid cell over two. The two agree within about two and a half percent, so 2023 and 2024 were not an unusual pair of years for this town, and a system sized on the atlas figure has the rainy season inside it. What the figure cannot do is make July produce like April.
The peak sun hours at five points around here, and where the figure comes from
What it means for sizing, and for what to expect
Size on the year and expect the months to differ. The calculator uses the yearly figure, so the system it sizes covers the bill over twelve months rather than in every month: a July bill shows more bought from the grid than an April bill, and the net metering credit earned in the dry months, which now carries forward, is what evens it out.
A house that must have power on a storm day needs a battery and a backed-up circuit, because the usual cause of a dark house in a storm is the grid going down rather than the cloud, and a grid-tied system is required to switch off with it. Panel choice changes little here: the datasheet lines that matter on a hot roof, efficiency and the temperature coefficient, are the same ones that matter in the rainy season, and cleaning the glass in the dry season stops dust and cloud stacking their losses.
Grid tied, hybrid and off grid: which keeps the lights on
The correction
Until 8 October 2026 this page said that output under cloud is typically 10 to 25 percent less than on a clear day. That sentence had no source. The figure most often quoted elsewhere, 10 to 25 percent of normal output on a heavily overcast day, could be traced only to installer and retailer sites rather than to a measurement, so neither version is used here. The NASA data say the old sentence was roughly right for an ordinary partly cloudy day and badly wrong for the overcast rainy-season day it claimed to describe.
Source: NASA POWER, daily point data, community RE, parameters ALLSKY_SFC_SW_DWN and CLRSKY_SFC_SW_DWN, latitude 15.03, longitude 120.69, 1 January 2023 to 31 December 2024, retrieved on 8 October 2026 from power.larc.nasa.gov/api/temporal/daily/point. The ratios on this page are this site’s arithmetic on that file, worked twice by two routes, and the file is kept in the site’s source archive so anyone can pull the same numbers.