How to Read the US Large-Scale Solar Photovoltaic Database in Python

On this page

Problem statement

The US Large-Scale Solar Photovoltaic Database (USPVDB) maps every solar facility in the country with a capacity of a megawatt or more: 6,611 of them in version 4.0, from community solar fields in Massachusetts to Gemini Solar in the Nevada desert. Each record is an outline of the panel arrays, digitised from aerial imagery, joined to the plant's entry in the Energy Information Administration's Form 860, with its AC and DC capacity, year, mounting, panel type and land use. This page reads the GeoJSON, checks the ids, capacities, points and dates, and shows what the table holds.

Quick answer

Read the GeoJSON with GeoPandas; each row is one facility with its outline:

print(pv[["case_id", "p_name", "p_state", "p_county", "p_year", "p_cap_ac", "p_cap_dc", "p_area", "p_axis"]].head(3).to_string(index=False))
print(pv.crs, pv.geom_type.value_counts().to_dict())
 case_id             p_name p_state p_county  p_year  p_cap_ac  p_cap_dc  p_area      p_axis
  400011            Mustang      OK Canadian    2015       2.5       3.0   50723 single-axis
  407013        Comfort CSG      NY     Erie    2024       4.0       5.5   77051 single-axis
  404351 Viking Solar (CSG)      IL Macoupin    2021       2.0       3.0   56943 single-axis
EPSG:4326 {'Polygon': 3529, 'MultiPolygon': 3082}

The first rows are small: Mustang, a 2.5 MW AC array in Canadian County, Oklahoma, and two community solar fields in New York and Illinois. The geometries are in EPSG:4326, 3,529 single polygons and 3,082 multipolygons. p_area is the array area in square metres, already computed in an equal-area projection.

Map of large-scale solar facilities in the lower 48 sized by capacity: large single-axis tracker plants across Texas, the Southwest, California and Florida, many small fixed-tilt arrays in the Northeast, North Carolina and Minnesota. Map of large-scale solar facilities in the lower 48 sized by capacity: large single-axis tracker plants across Texas, the Southwest, California and Florida, many small fixed-tilt arrays in the Northeast, North Carolina and Minnesota.
Large-scale solar facilities by mounting, sized by MW AC.

The records, the capacities and the dates

1. One record per facility

print(f"case_id unique: {pv.case_id.is_unique}; eia_id unique: {pv.eia_id.is_unique}; states: {pv.p_state.nunique()}")
print(pv.multi_poly.value_counts().to_dict())
print(f"polygons in the multi-part records: {pv[pv.multi_poly == 'multi'].geometry.apply(lambda g: len(getattr(g, 'geoms', [g]))).describe()[['mean', 'max']].round(1).to_dict()}")
case_id unique: True; eia_id unique: True; states: 50
{'single': 3529, 'multi': 3082}
polygons in the multi-part records: {'mean': 5.4, 'max': 126.0}

case_id is the database's own stable id and eia_id the EIA plant code; both are unique, so each EIA plant appears once. The 6,611 facilities carry 50 different state codes. 3,082 records are multi-part, a plant whose arrays are separate fields; they average 5.4 parts, and one has 126.

2. Two capacities

print(f"MW AC: {pv.p_cap_ac.sum():,.0f}; MW DC: {pv.p_cap_dc.sum():,.0f}; DC/AC: {pv.p_cap_dc.sum() / pv.p_cap_ac.sum():.3f}")
print(pv.p_cap_ac.describe(percentiles=[0.25, 0.5, 0.75, 0.99]).round(1).to_string())
MW AC: 131,574; MW DC: 162,802; DC/AC: 1.237
count    6611.0
mean       19.9
std        50.2
min         0.2
25%         1.8
50%         3.2
75%         7.0
99%       250.0
max       690.0

The table has two capacities: 131,574 MW AC, what the inverters deliver to the grid, and 162,802 MW DC, the panels' rating, 1.237 times as much. Sizes are very uneven. The median facility is 3.2 MW AC, three quarters are 7.0 MW or smaller, and the largest is 690 MW; the mean of 19.9 MW describes few facilities. Sum one column only; the AC and DC page shows how far mixing them goes wrong.

Histogram on a log scale of facility capacity, crowded between 1 and 6 MW AC with a long tail to 690 MW, median 3.2 MW and mean 19.9 MW. Histogram on a log scale of facility capacity, crowded between 1 and 6 MW AC with a long tail to 690 MW, median 3.2 MW and mean 19.9 MW.
Facility capacity, MW AC.

3. The point and the outline

pts = gpd.GeoSeries(gpd.points_from_xy(pv.xlong, pv.ylat), crs=4269).to_crs(pv.crs)
inside = pts.within(pv.geometry.reset_index(drop=True))
print(f"representative point inside its outline: {inside.mean():.1%}; outside: {(~inside).sum()}")
print(pv[~inside.values].multi_poly.value_counts().to_dict())
representative point inside its outline: 99.1%; outside: 61
{'multi': 52, 'single': 9}

ylat and xlong are a representative point for each facility, in NAD83. 99.1% lie inside their own outline. Of the 61 that don't, 52 belong to multi-part facilities, where the point is placed between separate fields. For a dot map the points are fine; for anything about land, use the outline.

4. Dates

print(f"operating since: {pv.p_year.min()} to {pv.p_year.max()}")
img = pd.to_datetime(pv.p_img_date.astype(str), format="%Y%m%d", errors="coerce")
print(f"image dates that are not dates: {img.isna().sum()} {pv.loc[img.isna(), 'p_img_date'].tolist()}")
print(f"imagery: {img.min():%Y-%m-%d} to {img.max():%Y-%m-%d}; imaged before the year it began operating: {(img.dt.year < pv.p_year).sum()}")
operating since: 1985 to 2025
image dates that are not dates: 1 [20240231]
imagery: 2009-04-17 to 2026-02-08; imaged before the year it began operating: 136

The facilities began operating between 1985 and 2025. Each outline was checked against an aerial image, dated in p_img_date as an integer, and one of those dates, 20240231, is 31 February: parse with errors="coerce" or the whole column fails. The images run from 2009 to February 2026, and 136 facilities were imaged before the year they began operating, so their outline shows the site under construction.

Capacity by state and by mounting place

The data used on this page

The US Geological Survey and Lawrence Berkeley National Laboratory's US Large-Scale Solar Photovoltaic Database (USPVDB), version 4.0 of 14 April 2026 (Fujita and others, doi:10.5066/P9IA3TUS), public domain, as GeoJSON with the metadata and changelog from the download. USGS replaces the files at the same address with each version, so version 4.0 was saved on this site on 28 October 2026. Capacities are megawatts, areas square metres of panels and inverters.

from pathlib import Path
import zipfile

import geopandas as gpd
import numpy as np
import pandas as pd
import requests

Path("data").mkdir(exist_ok=True)
# USGS/LBNL US Large-Scale Solar Photovoltaic Database (USPVDB) version 4.0 of April 2026 (doi:10.5066/P9IA3TUS),
# array outlines as GeoJSON, with the Census Bureau's 2024 counties and states. Saved 28 October 2026: each new
# version replaces the files at the same address.
# Live: https://energy.usgs.gov/uspvdb/assets/data/uspvdbGeoJSON.zip
ARCHIVE = "usgs-uspvdb-v4-0-2026-10-28.zip"
if not Path(f"data/{ARCHIVE}").exists():
    Path(f"data/{ARCHIVE}").write_bytes(requests.get(f"https://data.spatialworkflow.io/{ARCHIVE}", timeout=600).content)
Z = zipfile.ZipFile(f"data/{ARCHIVE}")
for f in ["uspvdb_v4_0_20260414.geojson", "cb_2024_us_county_20m.zip", "cb_2024_us_state_20m.zip"]:
    if not Path(f"data/{f}").exists():
        Path(f"data/{f}").write_bytes(Z.read(f))
pv = gpd.read_file("data/uspvdb_v4_0_20260414.geojson")
print(f"USPVDB 4.0: {len(pv):,} solar facilities, {pv.p_cap_ac.sum():,.0f} MW AC")
USPVDB 4.0: 6,611 solar facilities, 131,574 MW AC

Example 1: capacity by state

st = pv.groupby("p_state").agg(facilities=("case_id", "size"), mw_ac=("p_cap_ac", "sum")).sort_values("mw_ac", ascending=False)
print(st.head(8).round(0).astype(int).to_string())
print(f"top 5 states' share of capacity: {st.mw_ac.head(5).sum() / st.mw_ac.sum():.1%}")
         facilities  mw_ac
p_state                   
TX              187  23033
CA              893  22881
FL              202  11622
NC              760   6697
AZ              110   5488
GA              152   5364
NV               66   4759
VA              124   4710
top 5 states' share of capacity: 53.0%

Texas and California lead with 23,033 and 22,881 MW AC, but California has 893 facilities against Texas's 187: Texas builds fewer, much larger plants. Florida, North Carolina, Arizona, Georgia, Nevada and Virginia follow, and the five largest states hold 53.0% of the capacity. North Carolina's 760 facilities are small ones: 6,697 MW between them.

Horizontal bars of solar capacity by state: Texas 23.0 GW from 187 facilities, California 22.9 GW from 893, Florida 11.6 GW, then North Carolina, Arizona and Georgia. Horizontal bars of solar capacity by state: Texas 23.0 GW from 187 facilities, California 22.9 GW from 893, Florida 11.6 GW, then North Carolina, Arizona and Georgia.
Solar capacity by state, the twelve largest.

Example 2: where the panels sit

print(pv.groupby("p_sys_type").agg(facilities=("case_id", "size"), mw_ac=("p_cap_ac", "sum"), median_mw=("p_cap_ac", "median")).sort_values("facilities", ascending=False).round(1).to_string())
                       facilities     mw_ac  median_mw
p_sys_type                                            
ground                       5976  130336.2        4.0
rooftop                       407     782.7        1.5
canopy                        148     293.4        1.5
canopy,rooftop                 62     120.1        1.5
canopy,ground                   7      15.0        1.5
canopy,ground,rooftop           5      15.2        3.7
ground,rooftop                  4       6.8        1.8
floating                        2       4.1        2.0

Ground-mounted systems are 5,976 facilities and 130,336 MW AC, nearly all the capacity. 407 rooftop and 148 parking-canopy systems are mostly small, with medians of 1.5 MW, and two floating arrays add 4.1 MW. The database's p_sys_type comes from the imagery, so it describes what the panels sit on, not what EIA reports.