Population density
PopDen
People per square kilometer. The classic lens — low density reads rural, high density reads urban.
Census data · 2023Environmental Science & Technology · 2024
Urbanicity & Rurality,
Redefined.
How researchers classify urban and rural areas can change the magnitude — and even the direction — of environmental exposure disparities. This study compares nine indicators under three classification schemes across Virginia and West Virginia. The classification choice is part of the result.
The one-line takeaway
Same places.
Different conclusions.
Environmental exposures and their health impacts vary substantially between urban and rural areas. But there is no standard way to draw that line. Different indicators and classification schemes produce different spatial patterns, different demographics — and sometimes opposite answers about who is more exposed.
The problem
One word. More than fifteen definitions.
In the United States alone, more than 15 federal definitions of “rural” are in use. Agencies and researchers pick indicators and schemes based on their own objectives, geography and data — and the choice quietly shapes every result that follows.
15+
federal definitions of “rural” in the United States
30%
gap between USDA (46 million) and Census Bureau (~60 million) rural population estimates for 2015
11
different classification systems found across 103 U.S. urban/rural health studies
44%
of the global population — about 3.5 billion people — lived in rural areas in the World Bank estimate cited by the study
The method
One landscape. Nine lenses.
The study uses nine census-tract-level indicators — three population-based and six built-environment-based — to construct alternative classification systems on a five-level scale from very rural to very urban.
Population-based indicators
Population size
TotalPop
Population count within each census tract, with lower values generally indicating greater rurality in this study.
Census data · 2023Rural–urban commuting area
RUCA
USDA codes combining population density, urbanization and daily commuting patterns into a multidimensional classification.
USDA · 2010Built-environment indicators
Nighttime light
NTL
City glow from space. NPP-VIIRS satellite lights trace human activity after dark.
NPP-VIIRS · 2020Points of interest
POIDen
Groceries, schools, parks and hospitals, mapped from OpenStreetMap. Density signals urban life.
OpenStreetMap · 2023Urban area coverage
UrbanC
The share of a census tract that the Census Bureau classifies as urban land.
U.S. Census · 2020Building density
BldgDen
The footprint of the built world — the share of land covered by building outlines.
OpenStreetMap · 2023Road density
RoadDen
Total road length per square kilometer. Infrastructure as a proxy for urbanization.
OpenStreetMap · 2023Distance to urban center
DisUrban
Straight-line distance from a tract’s center to the nearest urban boundary. Farther reads more rural.
U.S. Census · 2020The schemes
Three ways to draw the same line.
Each of the nine indicators is cut into five levels using one of three classification schemes. Changing the scheme can change everything.
Global quantile
Five levels with equal numbers of census tracts. Thresholds are set using indicator values across the entire study region.
The baseline scheme for this study.
Global natural breaks
Groups similar values together, emphasizing the contrasts between classes — a very different geography can emerge.
Maximizes between-class differences.
Local quantile
Quantile thresholds set separately for Virginia and West Virginia, mirroring locally formulated classifications.
One row per state.
The study area
Two states. One contrast.
The study focuses on the neighboring states of Virginia and West Virginia, which had markedly different urbanization rates in the 2020 Census. Analysis is conducted at the census-tract level, a common unit in environmental exposure and health research.
75.6%
Virginia’s urbanization rate — 2020 U.S. Census
44.6%
West Virginia’s urbanization rate — 2020 U.S. Census
70.1 km²
Average census-tract area (SD 128.3 km²)
3
Environmental exposures compared: PM2.5, greenspace, land-surface temperature
The findings
The label changes the answer.
Nine indicators and three schemes repeatedly reclassify the same census tracts. The resulting spatial patterns, demographic compositions and exposure gaps can differ substantially.
Who counts as urban?
Share of the study area’s population classified into “moderately urban”, “highly urban” or “very urban” — by indicator.
Same census tracts, re-classified. Under global quantile, the RUCA system calls 89.43% of the population urban; under global natural breaks, the POIDen system calls just 2.87% urban. Data: Fig. 5a–b, Song et al. 2024.
The direction flips.
Mean PM2.5 exposure (2000–2016) in West Virginia — urban vs rural tracts, by classification system.
Under TotalPop, rural tracts carry higher PM2.5 (7.14 vs 6.86 μg/m³). Under PopDen, urban tracts do (8.21 vs 6.41 μg/m³). Same places — opposite conclusion. RUCA- and NTL-based systems also reverse the disparity. Data: Fig. 6b, Song et al. 2024.
Not all lenses agree.
Pearson correlation (r) between classification systems’ urban/rural levels — higher means more similar spatial patterns.
TotalPop has very low spatial agreement with the other systems (r < 0.02). The scheme matters too: BldgDen and DisUrban are highly correlated under global quantile (r = 0.91) but substantially less so under global natural breaks (r = 0.41). Data: Figs. 3–4, Song et al. 2024.
What it means
Choose the lens. Understand the lens.
The study doesn’t argue for a single “correct” definition — every system captures a real facet of urbanicity and rurality, and none captures all of it. The call is to see classification itself as part of the science.
For researchers
Run multiple systems.
Use several complementary urban/rural classifications and sensitivity analyses before concluding. A single system can mask — or invert — the disparities under study.
For policymakers
Match the label to the goal.
Classification systems shape where resources and policies land. Policymakers should work closely with researchers so the chosen system fits the policy goal and the target population.
The honest caveat
No indicator is complete.
Census data carry demographic biases and delays; nighttime lights are skewed by wildfires, ships and industry; road and building footprints can mislabel mining or power sites. Even multidimensional codes like RUCA have limits.
The paper
Delineating Urbanicity and Rurality:
Impact on Environmental Exposure Assessment
How the line between city and country is drawn — and why it matters for exposure science.
DOI: 10.1021/acs.est.4c06942What they didBuilt nine indicators and three schemes into urban/rural classification systems, applied them to Virginia & West Virginia, and compared spatial patterns, demographics and three environmental exposures.
What they foundClassification systems change the magnitude and direction of urban–rural exposure disparities — in exposure assessment, the label is part of the result.
KeywordsUrban/rural metrics · urban–rural classification · urban–rural disparities · environmental exposure