How Much Land Do Solar Farms Use? Capacity per Area in Python

On this page

Problem statement

How much land a solar farm needs is one of the most argued questions about renewable energy, and the answer depends on what is measured. The USPVDB's p_area is the area of the panel arrays and inverters only, drawn from imagery, without fences, roads or setbacks, so it gives the core footprint of every facility of a megawatt or more. This page turns it into megawatts per square kilometre and acres per megawatt, compares tracking and fixed arrays and the years they were built, and looks at the records the database itself flags as implausible.

Quick answer

Divide DC capacity by array area for the whole fleet and for each facility, and express it in the acres per megawatt AC that planners use:

dens = pv.p_cap_dc / (pv.p_area / 1e6)                 # MW DC per km² of array
print(f"all facilities: {pv.p_cap_dc.sum() / (pv.p_area.sum() / 1e6):.1f} MW DC per km² of array; median facility {dens.median():.1f}")
acres_per_mw = (pv.p_area / 4046.86) / pv.p_cap_ac
print(f"acres of array per MW AC: median {acres_per_mw.median():.2f}; middle half {acres_per_mw.quantile(0.25):.2f}-{acres_per_mw.quantile(0.75):.2f}")
all facilities: 57.1 MW DC per km² of array; median facility 66.6
acres of array per MW AC: median 4.78; middle half 3.90-5.86

Across all facilities, the arrays hold 57.1 MW DC per km², and the median facility 66.6. In planners' units, the median facility uses 4.78 acres of array per MW AC, and the middle half 3.90 to 5.86. The larger figures that circulate in planning debates count the fenced area around the arrays, with roads and gaps, which this database does not draw.

Histogram of acres of panel array per MW AC, centred around 4 to 6 with a median of 4.78. Histogram of acres of panel array per MW AC, centred around 4 to 6 with a median of 4.78.
Acres of array per megawatt.

Mounting, year and the outliers

1. By mounting

g = pv.groupby("p_axis").agg(n=("case_id", "size"), mw=("p_cap_dc", "sum"), km2=("p_area", lambda a: a.sum() / 1e6))
g["mw_per_km2"] = g.mw / g.km2
print(g[g.n >= 20].round(1).to_string())
                           n        mw     km2  mw_per_km2
p_axis                                                    
dual-axis                 49     798.6    19.9        40.1
fixed-tilt              3109   29222.6   405.4        72.1
fixed-tilt,single-axis    21     366.9     6.8        54.2
single-axis             3146  119029.4  2180.8        54.6

Fixed-tilt arrays are the densest, 72.1 MW DC per km² of array, and single-axis trackers hold 54.6. Trackers turn to follow the sun and must be spaced so their rows do not shade each other when tilted at morning and evening; fixed rows can sit closer. Trackers produce more energy per megawatt, so per unit of energy the gap is smaller than per unit of capacity.

2. By year

y = pv[pv.p_year >= 2010].groupby("p_year").agg(mw=("p_cap_dc", "sum"), km2=("p_area", lambda a: a.sum() / 1e6))
print((y.mw / y.km2).round(1).to_string())
p_year
2010    44.5
2011    57.4
2012    57.3
2013    55.1
2014    55.6
2015    59.0
2016    54.8
2017    59.8
2018    63.1
2019    66.1
2020    56.8
2021    61.6
2022    58.4
2023    54.7
2024    55.1
2025    53.8

Density by year built changes little: between 54 and 66 MW DC per km² since 2011, with no steady trend. Panels have become more efficient, but the shift to trackers, which need more space, has probably offset the gain.

Bars of MW DC per square kilometre of array by year built, between about 45 and 66 with no steady trend. Bars of MW DC per square kilometre of array by year built, between about 45 and 66 with no steady trend.
Capacity per area of array by year built.

3. Outliers the database flags

flag = pv[pv.p_zscore.abs() > 3]
print(f"|z| > 3: {len(flag)}; median MW DC per km² among them: {(flag.p_cap_dc / (flag.p_area / 1e6)).median():.0f}")
print(flag.assign(d=(flag.p_cap_dc / (flag.p_area / 1e6)).round(0)).nlargest(4, "d")[["p_name", "p_state", "p_sys_type", "p_cap_dc", "p_area", "d"]].to_string(index=False))
|z| > 3: 60; median MW DC per km² among them: 444
               p_name p_state p_sys_type  p_cap_dc  p_area      d
           Elm Street      MA    rooftop       3.7     869 4258.0
Maverick Solar 7, LLC      CA     ground     179.0   62529 2863.0
           Howell CSG      NY     ground       3.0    2381 1260.0
        Halifax Solar      MA    rooftop       3.1    2799 1108.0

The database flags its own doubtful records with p_zscore, the z-score of capacity per area. 60 facilities lie more than three standard deviations out, with a median of 444 MW DC per km², several times any real array. The top of the list is a rooftop in Massachusetts with 3.7 MW on 869 m², and Maverick Solar 7 in California, 179 MW on 62,529 m², where the outline evidently covers only part of the plant. Drop or check them before quoting land use.

The land in all US solar arrays

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: the land in all US solar arrays

km2 = pv.p_area.sum() / 1e6
print(f"array area: {km2:,.0f} km² ({km2 * 247.105:,.0f} acres) for {pv.p_cap_ac.sum():,.0f} MW AC")
big = pv[pv.p_cap_ac >= 100]
print(f"facilities of 100 MW AC or more: {len(big)}, {big.p_area.sum() / pv.p_area.sum():.1%} of the array area")
array area: 2,852 km² (704,769 acres) for 131,574 MW AC
facilities of 100 MW AC or more: 406, 58.3% of the array area

Together, the arrays cover 2,852 km², 704,769 acres, for 131,574 MW AC, an area a little larger than Rhode Island's land. The 406 facilities of 100 MW AC or more account for 58.3% of it.

Horizontal bars of MW DC per square kilometre by mounting: fixed tilt 72.1, single-axis 54.6, mixed 54.2, dual-axis 40.1. Horizontal bars of MW DC per square kilometre by mounting: fixed tilt 72.1, single-axis 54.6, mixed 54.2, dual-axis 40.1.
Capacity per area by mounting.