Why Did My Solar Production Drop This Year? LA Year-to-Year Data, 2023 and the Output That Didn’t Come Back
On the same Los Angeles-area systems, a year’s solar production usually lands within a few percent of the year before: half of 488 year-over-year changes on two monitoring platforms fell between −5.4% and +1.5%. Big drops come when the whole fleet moves together. In 2023 the median system made 10.8% less than in 2022 and 94% of systems fell, in a year with about 7% less sunlight. What followed is the open question: by 2025 the sunlight was within 3% of 2022, but the same systems still made 10–13% less, and October 2024 ran about a fifth below the year before with nearly the same sun.
How much annual solar output changes from one year to the next
488 year-over-year changes on 209 systems, each compared with itself.
| Year | Systems | Median change | Middle half | Up / down | SolarEdge | Enphase | Sunlight, NASA | Sunlight, ERA5 |
|---|---|---|---|---|---|---|---|---|
| 2021 vs 2020 | 35 | +2.5% | +0.2% to +5.5% | 77% / 23% | +2.3% | — | +1.4% | +0.7% |
| 2022 vs 2021 | 58 | +0.6% | −1.1% to +2.9% | 55% / 45% | +1.1% | 0.0% | +0.5% | +0.6% |
| 2023 vs 2022 | 85 | −10.8% | −12.3% to −9.0% | 6% / 94% | −10.7% | −11.1% | −7.3% | −7.0% |
| 2024 vs 2023 | 135 | −2.2% | −5.3% to +0.4% | 29% / 71% | −1.8% | −2.5% | +3.6% | +4.7% |
| 2025 vs 2024 | 175 | +0.2% | −1.6% to +1.9% | 54% / 46% | +0.6% | −0.1% | −0.2% | −0.1% |
| Sunlight: annual change at our San Fernando Valley weather point, NASA POWER satellite data and ERA5 reanalysis. 2020 vs 2019 omitted (10 systems). | ||||||||
https://cali-energy.com/research/solar-production-year-to-year. Aggregated data are free to use under CC BY 4.0 with a link to this page.2023: a year with less sun
Same systems, 2023 against 2022, month by month.
| Month | Production change, median | Sunlight, NASA | Sunlight, ERA5 |
|---|---|---|---|
| January | −16.6% | −15.9% | −16.2% |
| February | −17.3% | −17.2% | −15.5% |
| March | −21.4% | −19.5% | −24.1% |
| April | −2.3% | −2.5% | −1.6% |
| May | −23.5% | −16.1% | −14.7% |
| June | −23.8% | −14.8% | −12.4% |
| July | −2.5% | +1.0% | +2.3% |
| August | −0.9% | −1.5% | −4.0% |
| September | −10.3% | −4.2% | −3.3% |
| October | +5.1% | +4.6% | +4.6% |
| November | −7.0% | 0.0% | +0.3% |
| December | +4.4% | +5.5% | +9.4% |
| 85 systems with complete 2022 and 2023. | |||
2023 is the clearest case. Production fell in January, February, March, May and June, the months in which both satellite and weather-model sunlight were 12–24% lower than a year earlier; in April, July, August, October and December the two were close to 2022. When nearly every system falls in the same months the sun does, the cause is the year, not the equipment. May and June fell further on rooftops than in the sunlight data, which fits low coastal cloud that a coarse grid only partly sees; how the marine layer shapes a year is measured in solar output by area, and the wet winter of 2022–23 in winter production.
The output that didn’t come back
Systems with complete years 2022 through 2025, median kWh per kW.
| Year | SolarEdge (50 systems) | Enphase (18 systems) | Sunlight, ERA5 (kWh/m²) | Sunlight, NASA (kWh/m² a day) |
|---|---|---|---|---|
| 2022 | 1,534 | 1,623 | 2,110 | 5.76 |
| 2023 | 1,409 | 1,426 | 1,963 | 5.34 |
| 2024 | 1,367 | 1,396 | 2,055 | 5.53 |
| 2025 | 1,377 | 1,416 | 2,052 | 5.52 |
| Sunlight at our San Fernando Valley weather point. Each platform column follows the same systems in all four years. | ||||
Sunlight came back after 2023: ERA5 shows +4.7% in 2024 and a 2025 within 3% of 2022. Production did not. Against 2022, the same SolarEdge systems made 10% less in 2025 and the same Enphase systems 13% less. Three years of normal aging account for part of that gap, roughly 1.5–4 points at the rates measured in how fast solar systems lose output; dust after the long dry spell of 2024 for another part (dirty panels, measured). The rest, several percent, is not explained by sunlight. It shows on two independent monitoring platforms, so it is not a reporting artifact, and it is shared by nearly all systems, so it is not a fault on any one roof.
Autumn 2024: low output with ordinary sun
Median kWh per kW of systems with September–November data in all four years, and output per unit of sunlight.
| Month | Enphase, kWh/kW (48) | SolarEdge, kWh/kW (66) | Sunlight, ERA5 (kWh/m²) | Enphase per unit of sunlight | SolarEdge per unit of sunlight | Aerosol (CAMS) | Rain (ERA5) |
|---|---|---|---|---|---|---|---|
| September 2022 | 129 | 128 | 174 | 0.74 | 0.74 | 0.14 | 24 mm |
| September 2023 | 116 | 116 | 168 | 0.69 | 0.69 | 0.13 | 24 mm |
| September 2024 | 113 | 113 | 187 | 0.60 | 0.60 | 0.11 | 2 mm |
| September 2025 | 113 | 115 | 173 | 0.65 | 0.67 | 0.13 | 25 mm |
| October 2022 | 112 | 110 | 148 | 0.76 | 0.75 | 0.11 | 5 mm |
| October 2023 | 119 | 118 | 155 | 0.77 | 0.77 | 0.07 | 4 mm |
| October 2024 | 93 | 94 | 152 | 0.61 | 0.62 | 0.10 | 3 mm |
| October 2025 | 113 | 112 | 152 | 0.75 | 0.74 | 0.07 | 59 mm |
| November 2022 | 97 | 92 | 109 | 0.89 | 0.84 | 0.07 | 59 mm |
| November 2023 | 90 | 88 | 109 | 0.82 | 0.80 | 0.06 | 23 mm |
| November 2024 | 82 | 76 | 106 | 0.77 | 0.71 | 0.07 | 9 mm |
| November 2025 | 76 | 71 | 92 | 0.83 | 0.77 | 0.08 | 138 mm |
| Weather at our San Fernando Valley point. Per unit of sunlight: kWh per kW divided by kWh per square meter. | |||||||
October 2024 is the sharpest case. The same Enphase systems made −22% against October 2023 and the SolarEdge systems −21%, while ERA5 sunlight changed −2%. September and November 2024 were also low for their sunlight. None of the usual explanations covers it:
Heat: the October 1–3 heat wave cost about 3% per unit of sunlight on those days, and the milder days after it were just as low (heat waves, measured). Smoke and haze: CAMS aerosol in October 2024 was ordinary, and each 0.1 of it costs about 2% (smoke and haze, measured). Cloud and rain: almost no rain, and both weather sources saw a normal October sky. A data problem: two independent monitoring platforms show the same drop. Dust: the panels had gone without a soaking rain since spring, and the first big storm, in late January 2025, lifted output noticeably, so dust is a real part of it; but dust builds slowly and on its own would not put October twice as far below its usual level as September.
Is your drop unusual? Compare your system with the fleet
Subtract the fleet’s change in the same year from your own. What is left is your system’s own change.
| Your change vs the fleet | Where it sits among measured results | Suggested next step |
|---|---|---|
| −2.1 to +2.1 points | Middle half: in line with the fleet | None from this signal alone |
| −4.7 to −2.1 points | Below the middle half, within the usual spread | Compare month by month with last year |
| −7.0 to −4.7 points | Lowest tenth | Find the months where output was lost; check alerts and shade |
| Below −7.0 points | Lowest 5% | Look for flat or zero days, offline panels or inverter faults; consider a check |
| Based on 488 year-over-year comparisons on two platforms. A screening signal, not a diagnosis of the equipment. | ||
Example: a system that made 9,000 kWh in 2022 and 7,900 in 2023 fell 12.2%. The fleet fell 10.8% that year, so the system was only 1.4 points below the fleet — an ordinary result in a poor solar year. The same 12% drop in a year when the fleet was flat would put the system in the lowest 5%.
Six systems, year by year
Each system against its own typical year (= 100), next to the fleet. SolarEdge systems with the longest records.
| System | 2019 | 2020 | 2021 | 2022 | 2023 | 2024 | 2025 |
|---|---|---|---|---|---|---|---|
| System A | — | 107 | 104 | 85 | 97 | 102 | 98 |
| System B | — | 107 | 102 | 105 | 98 | 97 | 90 |
| System C | — | 108 | 109 | 108 | 92 | 92 | 92 |
| System D | — | 103 | 105 | 106 | 97 | 95 | 95 |
| System E | 102 | 99 | 102 | 101 | — | 81 | 92 |
| System F | — | 101 | 103 | 102 | 95 | 96 | 99 |
| Fleet median | 100 | 101 | 105 | 108 | 100 | 97 | 98 |
Most of these system-years (64%) sit within 5 points of the fleet line: in a good year everyone is up, in a poor one everyone is down. The exceptions are the useful part. System A fell to 85 in 2022 while the fleet stood at 108, then returned to its usual level in 2023. System E dropped to 81 in 2024 against the fleet’s 97 and moved back toward the fleet in 2025. System B ended 2025 at 90, 9 points below the fleet. One system moving apart while the others hold steady is exactly what a year-to-year check is designed to catch.
Methodology
- Systems
- 209 systems on Enphase and SolarEdge monitoring with at least two consecutive complete years in 2019–2025, under the rules of the platform comparison: 12 valid months, no downtime year, a plausible output per kW and a verified DC rating.
- Year-over-year change
- The same system’s annual kWh compared with its previous complete year. Years with fewer than 20 comparisons are left out of the pooled figures.
- Your change vs the fleet
- A system’s change minus the median change of all systems in the same year.
- Sunlight
- NASA POWER all-sky irradiance (satellite-derived, about a 1° grid) and ERA5 reanalysis surface irradiance (about 25 km), both at our San Fernando Valley weather point, 34.24°N 118.50°W. Aerosol: CAMS daily mean optical depth.
- Trajectories
- Six SolarEdge systems with at least six complete years, picked at even steps from lowest to highest output.
Frequently asked questions
Why did my solar production drop this year?
Most often because the year had less sun. Half of year-over-year changes on our measured systems fell between −5.4% and +1.5%. Compare your change with the fleet’s change in the same year; a gap of more than about 5 points is the signal worth checking.
Is a 10% drop in solar production normal?
It can be. In 2023 the median system fell 10.8% and 67% of systems fell more than 10%. In a year when the fleet was flat, a 10% drop would put a system in the lowest 5%.
Why was 2023 a bad year for solar in LA?
Less sunlight: about 7% less over the San Fernando Valley than in 2022, concentrated in January–March and May–June, the same months in which production fell.
Why didn't my production recover in 2024 and 2025?
Across our fleet it did not fully recover either: the same systems made 10–13% less in 2025 than in 2022, while sunlight was only 3% lower. Aging and dust explain part; the rest is not explained by the sunlight data.
Why was October 2024 so low?
Not the sun. The same systems made about 22% less than in October 2023 with nearly the same sun; heat, smoke, cloud and data problems do not explain it, and dust explains only part.
More research
Next steps
Your system fell 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.