How Fast Do Solar Systems Lose Output? Measured on 60 LA Roofs Over Five to Nine Years
Corrected for sunlight, the typical Los Angeles-area rooftop system with five or more years of data lost 1.3% of its output a year (median of 60 systems; 1.0% for the 35 with six years or more). That matches the 1.3% a year NREL measured on residential rooftops, a figure that includes dirt and shade, and sits above the roughly 0.5% a year typical of panels alone. The decline was not steady: output stayed within a few percent from 2019 to 2023, then fell across nearly every system at once from September 2024 to early 2025, and has partly recovered in 2026. Checked against a second sunlight source and on a second monitoring platform, the answer stays in the same range, roughly 1–1.5% a year: 0.9% with NASA’s sunlight data, 1.7% on the 14 Enphase systems with five years or more.
This is a whole-system figure: everything that changes what the meter sees, not only the panels. Panel aging, dust on the glass, a tree that grew, a failed panel-level device, heat and smoke all count. The method is the one NREL uses for fleets of systems: compare each month with the same month a year earlier, after correcting both for sunlight, and take the median.
Sunlight-corrected output, year by year
Same systems, each compared with its own typical month. 100 = the system’s usual level.
| Year | Systems | Index (median) |
|---|---|---|
| 2019 | 17 | 103.4 |
| 2020 | 36 | 100.5 |
| 2021 | 56 | 102.7 |
| 2022 | 72 | 104.1 |
| 2023 | 74 | 101.4 |
| 2024 | 73 | 95.5 |
| 2025 | 70 | 96.3 |
| 2026* | 62 | 98.0 |
| 2026*: January–August. A year counts for a system with at least eight valid months (six for 2026). | ||
From 2019 to 2023 the index moved between 100.5 and 104.1 with no clear direction. In 2024 it dropped to 95.5 and stayed there in 2025 (96.3); January–August 2026 is back up to 98.0. Most of the loss in the per-year rate comes from that one shared step, not from a steady slide on every roof.
Month by month: the dips every system shared
Monthly data show the drops are shared, not individual. September and October 2020 (89 and 85) coincide with that fall’s wildfire smoke, including the Bobcat Fire in the San Gabriel Mountains, and January 2025 (91) with the Palisades and Eaton fires. Smoke cuts the light that reaches the panels and leaves ash on the glass, and the sunlight dataset does not fully capture it. The lowest month, October 2024 (83), has no counterpart in the sunlight data and showed up in every region we monitor, including systems well outside Los Angeles. If your production fell in late 2024, it was most likely not your system alone — compare it with the fleet in why did my solar production drop.
How fast each system lost output
60 systems with five or more years of data.
| Data per system | Systems | Median | Middle half | 95% interval of median | Losing under 0.5%/yr | Losing over 2%/yr |
|---|---|---|---|---|---|---|
| 3+ yrs | 79 | −1.43 | −2.9…−0.4 | −1.9…−1.0 | 27% | 37% |
| 4+ yrs | 70 | −1.41 | −2.7…−0.4 | −1.9…−1.0 | 29% | 34% |
| 5+ yrs | 60 | −1.29 | −2.1…−0.3 | −1.8…−0.7 | 33% | 28% |
| 6+ yrs | 35 | −0.97 | −1.9…−0.1 | −1.9…−0.3 | 43% | 23% |
| Longer records dilute the 2024–2025 step, which sits near the end of every record; that is why the median moves toward zero as records get longer. | ||||||
https://cali-energy.com/research/solar-degradation-rate-los-angeles. Aggregated data are free to use under CC BY 4.0 with a link to this page.A third of the systems (33%) lost less than 0.5% a year, about what the panels alone are expected to lose. At the other end, 28% lost more than 2% a year. NREL found the same pattern on home rooftops: systems with more shade declined faster, a median 2.1% a year against 1.2% for little or no shade. A system losing more than 2% a year after sunlight correction is worth checking for a fixable cause — new shade, heavy dirt or a failed panel-level device — before anyone looks at a panel warranty.
How this compares with national field data
| Population | Systems | Median loss a year | Source |
|---|---|---|---|
| Panels alone (modules) | — | about 0.5% | Deceglie et al., summarizing prior work |
| US commercial and utility systems | ≈1,700 sites | 0.75% | Jordan et al., 2022 |
| in hotter climates | 0.88% | same | |
| US residential rooftops | 387 systems | 1.3% | Deceglie et al. |
| with medium or high shade | 2.1% | same | |
| This study, 5+ years | 60 | 1.3% | Cali Energy monitoring, 2019–2026 |
| This study, 6+ years | 35 | 1.0% | same |
| NREL’s residential figure comes from the same year-on-year method (RdTools), applied to sub-hourly data with sunlight modeled on the plane of the panels. | |||
Panels age slowly; whole systems on real roofs lose more because the meter also sees what happens around the panels. Premium panel warranties promise about 89–92% of rated power after 25 years, which is a promise about the panel under test conditions, not about a roof’s yearly output. For warranty terms and what year 10 or 25 looks like, see solar panel degradation and how long solar panels last.
Checked another way
The same method on a second monitoring platform, with a second sunlight source, and without the shared 2024–25 dip.
| Systems and record length | Systems | ERA5 sunlight (main) | NASA sunlight | ERA5, without Sept 2024 – Apr 2025 |
|---|---|---|---|---|
| SolarEdge, 5+ years | 60 | −1.29 | −0.89 | −1.15 |
| SolarEdge, 6+ years | 35 | −0.97 | −0.90 | −0.51 |
| Enphase, 5+ years | 14 | −1.69 | −0.86 | −1.57 |
| Enphase, 3+ years | 96 | −2.70 | −1.75 | — |
| NASA POWER data run through 2025. Leaving out eight months removes too many comparisons on the shorter Enphase records to report a 3-year figure. | ||||
Three things come out of this. First, the sunlight source matters by about half a point: NASA’s satellite data, which include the year’s actual haze, give 0.9% on SolarEdge and 0.9% on Enphase, against 1.3% and 1.7% with ERA5. Second, the shared dip of late 2024 adds a little but does not make the rate: without it, SolarEdge systems with five years or more lost 1.1% a year. Third, short records mostly measure the period they cover. The 96 Enphase systems with three or more years of data, most of them started in 2022 or 2023, show 2.7% a year, because their records begin just before the 2023 dip and run through 2024–25; what those years did is in why solar production drops from year to year. For planning, a whole-system loss of roughly 1–1.5% a year is what our data support, less on an unshaded roof that is kept clean.
New panels can lose more in their first year, from light-induced degradation. On 90 Enphase systems that started in 2021–2023, comparisons against the first year of operation ran at a median −7.1% a year and later ones at about zero. That pattern is what light-induced degradation would look like, but it is also what the calendar would produce: those first years fell on 2022–23, and the comparisons that follow land on 2024–26. In these records a first-year loss cannot be told apart from the calendar.
How much is left after 10, 20 or 25 years
Output as a share of the first year, in a year of average sun, at a constant yearly loss.
| System age | Panels alone, 0.5% | Our fleet, NASA sunlight, 0.9% | Our fleet (ERA5) and NREL rooftops, 1.3% |
|---|---|---|---|
| 5 years | 98% | 96% | 94% |
| 10 years | 95% | 91% | 88% |
| 15 years | 93% | 87% | 82% |
| 20 years | 90% | 83% | 77% |
| 25 years | 88% | 80% | 72% |
Buying a house that already has solar? Ask for the monitoring login or two to three years of production, and compare it with the first year or with the system size times the LA benchmark in real-world solar production. A ten-year-old system making about 90% of its early output is normal; much less than that points to shade, a failed panel-level device or an inverter that needs attention, which is cheaper to find before closing. Inverter outages are not part of these rates; they are measured in what an inverter failure costs.
Why real systems lose more than panels
Dust and ash
Dirt builds on the glass between rains, and smoke seasons add ash. Rain or a rinse removes much of it, so this part comes and goes.
Growing shade
A tree that adds a few feet a year shades a little more of the array each winter. NREL found shaded home systems declining fastest.
Panel-level electronics
A failed optimizer or microinverter takes one panel out while the rest keep producing, so output drops a few percent without an outage.
The panels themselves
Slow chemical and mechanical aging of the cells, about half a percent a year on typical modules. This is the part a panel warranty covers.
For a 6 kW system making about 8,740 kWh a year, a 1.3% yearly loss is about 110 kWh a year. That is small next to ordinary weather: on the same systems, the middle half of year-to-year changes ran from −6.3% to +2.0%. One weak year says little; a gap that keeps widening against the same months of earlier years, after a wet winter has washed the panels, is the signal to check. The monthly benchmark and the year-to-year comparison are built for that.
Methodology
- Systems
- Rooftop systems in our SolarEdge monitoring fleet with at least 24 month-to-month comparisons a year apart: 79 systems; 60 have five or more years of data and 35 six or more (longest 8.5 years). The same four systems as in the production benchmark are excluded (one we did not install, one expanded, two whose output does not match their recorded capacity).
- Sunlight correction
- Each month’s kWh per kW is divided by that month’s sunlight on a horizontal surface from the ERA5 reanalysis (Open-Meteo Historical API), taken at the center of the system’s region (San Fernando Valley, Los Angeles basin, Santa Clarita Valley, Ventura County; systems elsewhere use the Los Angeles basin point). Regional centers, not roof locations.
- Outage months
- The first (partial) month and months below 75% of the system’s own median for that calendar month are removed, so outages and inverter failures do not count as wear. They are covered in inverter repairs.
- Rate
- For every month with a valid month a year earlier, the change between the two; a system’s rate is the median of those changes (the year-on-year method in NREL’s RdTools). The fleet figure is the median across systems; its 95% interval comes from resampling systems.
- Index
- A month’s sunlight-corrected output divided by the same system’s median for that calendar month; the fleet line is the median across systems reporting that month (at least 15).
- Checks
- With NASA POWER sunlight data instead of ERA5 (available through 2025), the median for 5+ years is −0.89%/yr (44 systems). Leaving out September 2024 – April 2025, the median is −1.15%/yr for 5+ years and −0.51%/yr for 6+. The same method on Enphase monitoring (98 systems with 24 or more comparisons, monthly totals from daily data, sunlight from ERA5 at the system’s area): −1.69%/yr for 14 systems with 5+ years; outage months removed: 0.6%.
- Limits
- Monthly data and horizontal sunlight at a regional point, with no correction for panel temperature or roof direction, so single-system rates are uncertain; the fleet median is the robust number. The causes in “why” are not separated in the data. Not a random sample of Los Angeles roofs.
Frequently asked questions
How much do solar panels degrade per year in Los Angeles?
On our monitored rooftop systems, whole-system output corrected for sunlight fell a median 1.3% a year over five to nine years (1.0% for systems with six years or more). That includes dirt, shade and equipment; panels alone typically lose about 0.5% a year.
Is losing 1% a year normal for a home solar system?
Yes. NREL measured a median 1.3% a year on 387 residential rooftops. In our data 33% of systems lost less than 0.5% a year; a system losing more than 2% a year is worth checking for a cause such as new shade, heavy dirt or a failed panel-level device.
Why did my solar production drop in late 2024 and 2025?
Across nearly all the systems we monitor, sunlight-corrected output fell from September 2024 into early 2025 (October 2024 was 83% of the usual level), and has partly recovered in 2026. A drop shared by the whole fleet points to conditions, not to your system.
Do solar panels lose efficiency over time?
Slowly. A year’s aging is smaller than ordinary weather: on the same systems the middle half of year-to-year changes ran from −6.3% to +2.0%.
I'm buying a house with 10-year-old solar. How much output is left?
At about 1% a year, roughly 90% of the first year’s output in an average-sun year; at 0.5% about 95%. Ask for two or three years of monitoring data and compare; much less than that points to a fixable cause such as shade or a failed device.
What does a solar degradation curve look like?
Close to a straight line falling about 1% a year on our fleet, from 100% to about 78% after 25 years, but in real years it moves in steps with the weather: flat from 2019 to 2023, a shared dip in late 2024, a partial recovery in 2026.
How do I know if my solar system is degrading too fast?
Compare each month with the same month in earlier years and with the fleet, after a wet winter has cleaned the panels. A gap that keeps widening year after year is the signal; a single bad year usually is not.
More research
Next steps
Production slipping more than the fleet?
Prepared by Cali Energy, October 15, 2026. Figures describe systems in our monitoring fleet and are not a guarantee of what any specific roof will produce; orientation, tilt, shade, equipment and weather change results. Not engineering, financial or utility advice. See our Content Disclaimer.