Methodology · Revised 30 August 2026

Every number on this site, and where it came from.

SolarAssetIndex is built entirely on public-domain federal data. The sources, the join logic, and the full scoring formula are documented below — including what the score does not account for. Nothing here is proprietary, and the ranking is reproducible by anyone willing to do the work.

01Data sources

Two federal datasets, both public domain, both free of republication restrictions.

SourceProvidesRelease
USPVDB v4.0
United States Large-Scale Solar Photovoltaic Database — USGS with Lawrence Berkeley National Laboratory
6,611 operational facilities of 1 MW and above. Hand-digitized array boundaries, capacity in AC and DC, install year, module technology, tracking configuration, tilt, azimuth, mount type, and co-located battery.14 April 2026
Annual
EIA Form 860M
Preliminary Monthly Electric Generator Inventory — U.S. Energy Information Administration
Operator and owning entity for operational plants. All pre-operational solar projects with development stage codes and expected commercial operation dates.July 2026
Monthly

The operational fleet indexed here totals roughly 162.8 GW-dc across 49 states and the District of Columbia. The pipeline adds 1,065 projects representing about 123 GW of nameplate capacity.

How the two are joined

Every USPVDB record carries an EIA plant identifier. That field is matched directly against the Plant ID in Form 860M, and generator-level rows are aggregated up to the plant level, since EIA reports individual inverter blocks while USPVDB describes whole facilities.

6,568 of 6,611 facilities match — 99.3%. The remaining 43 have no corresponding EIA record and display as "Operator not on record" rather than being dropped or guessed at.

02What is and isn't included

The directory covers solar generating facilities of 1 MW and above. That threshold is inherited from USPVDB and is the reason coverage is what it is.

Mount typeFacilities
Ground-mount5,976
Rooftop407
Canopy148
Canopy and rooftop combined62
Floating2

Not included: residential solar, and commercial rooftop arrays below 1 MW. There is no authoritative public inventory of sub-megawatt systems in the United States, so any attempt to list them would be assembled from vendor claims and permit scraping. We would rather show a complete picture of one segment than a partial one of two.

If you operate a facility below the threshold and want it indexed, use the claim form on any listing.

03Inspection priority scoring

Each operational facility carries a score from 0 to 100 estimating how much it would benefit from aerial thermal inspection. Four components, each derived from an observable property of the asset. No component is weighted by anything we won't show you.

Component 1 — Age 0 to 50 points

Photovoltaic modules fail on a bathtub curve. Infant mortality shows up in the first year or two and is usually caught under workmanship warranty. The middle years are quiet. From roughly year eight onward, cell cracking, backsheet degradation, and bypass diode failure rates climb steadily, and those faults are thermally visible long before they show up in production data.

age ≤ 5      →  age × 2
5 < age ≤ 20 →  10 + (age − 5) × (40 ÷ 15)
age > 20     →  50

Flat and low through the first five years, a steady ramp through the second and third decade, saturating at 50. Age is the heaviest single input because it is also the best-evidenced one.

Component 2 — Scale 0 to 25 points

Two things scale with facility size: the production revenue exposed to an undetected fault, and the economics of aerial survey versus manual inspection. A crew can walk a 2 MW site in a day. Nobody walks a 300 MW site.

scale = min(25, 8 × log₁₀(MW_dc + 1))

Logarithmic rather than linear, so a 400 MW facility does not swamp the ranking entirely.

Component 3 — Configuration 0 to 15 points

Tracking systems add mechanical failure modes that fixed-tilt arrays do not have — stuck rows, misaligned strings, and drive faults, all of which present as thermal anomalies. Thin-film modules follow a different degradation curve than crystalline silicon and are more prone to certain shunting behaviours.

single-axis tracking  +8
dual-axis tracking   +10
thin-film modules     +5
                    capped at 15

Component 4 — Mount type 0 to 10 points

Rooftop arrays carry consequences ground-mount arrays don't. A hot spot over an occupied building is a fire-risk question, not only a production question, and roof access is slow, disruptive, and often requires fall protection. Canopy structures sit above vehicles and pedestrians with similar access problems.

rooftop   +10
canopy     +7
floating   +5

04Priority bands

Bands are percentile cutoffs recalculated from the live fleet on every data refresh, not fixed thresholds. This is a deliberate choice and worth explaining.

Why relative, not absolute

The American solar fleet is young — median install year is around 2019. When we first ran fixed thresholds, 72% of facilities landed in the lowest band and not one qualified as critical. That result is technically defensible and practically useless. Ranking each site against the rest of the fleet answers the question an operator actually asks: where does this asset sit relative to everything else out there?

LOWMODERATEHIGHCRITICAL
BandCutoffPercentileFacilitiesCapacity
Critical≥ 43.7Top 10%67227,716 MW
High≥ 34.3Next 20%1,31858,784 MW
Moderate≥ 25.2Next 30%1,99952,074 MW
Low< 25.2Bottom 40%2,62224,228 MW

Observed scores currently range from 4.4 to 69.2, with a median of 27.9. Colour on the map follows an infrared ramp rather than a traffic-light palette, because the audience for this data reads thermal imagery daily.

05Pipeline classification

Pre-operational projects are not scored — a facility that hasn't been built has no vintage. They are classified by EIA development stage instead, collapsed from EIA's six status codes into four stages.

StageEIA codesProjects
Construction complete, pre-commercialTS85
Under constructionU, V496
PermittedT, L311
PlannedP173

Where a project's generators carry differing status codes, the project inherits the most advanced stage among them.

06Known limitations

Listed plainly, because a scoring model whose weaknesses aren't published isn't a model — it's a marketing claim.

The significant one

The score does not account for weather or hail exposure. Hail is the largest single driver of unscheduled photovoltaic inspection in the United States, and a site in the hail corridor carries risk that vintage alone will never capture. Incorporating it requires a NOAA storm-events join that has not been built. Until it is, treat the score as a measure of intrinsic asset risk, not total risk.

07Corrections and refresh

The dataset is rebuilt when either source publishes — USPVDB annually, EIA Form 860M monthly. Band cutoffs recompute against the fleet at each rebuild, so a facility's band can change without its underlying score moving.

If a listing is wrong, use the Claim this listing action on any facility. Corrections that conflict with the federal source are noted on the listing rather than silently overwritten, so the record stays traceable to its origin.

SolarAssetIndex is operated by SolarATL, the solar division of Eyes Above. Inspection work referred through this directory is fulfilled by that team. The scoring model makes no distinction between facilities inside and outside their service area — the ranking is computed nationally and identically.