Across fifty-two paired urban and natural landscapes across the conterminous United States, multidecadal satellite observations reveal that differences in vegetation greenness and land surface temperature reverse along continental hydroclimatic gradients rather than following a uniform nationwide pattern. In moist, temperate environments, metropolitan footprints were warmer and less vegetated than nearby reference landscapes, whereas in arid environments, cities were frequently greener and cooler than undeveloped surroundings. Across fifty-two pairs examined from 1995 through 2024, the median metropolitan footprint exhibited a normalized difference vegetation index deficit of negative 0.174 and an elevated surface temperature of positive 5.09 degrees Celsius relative to its paired reference, with sample-wide mean differences of negative 0.143 and positive 3.90 degrees Celsius. Furthermore, annual composite evaluations revealed that urban surfaces warmed approximately 22 percent faster than adjacent natural counterparts, while urban greenness changed minimally or declined slightly as reference landscapes greened.
Published in Landscape Ecology by researchers Dion Kucera, Diane Pataki, and G. Darrel Jenerette, the investigation evaluated these dynamics through an inter-urban water-deficit framework. Urbanization alters surface energy partitioning, water availability, and ecological functioning through impervious surfaces, managed vegetation, altered hydrological runoff, and supplemental irrigation. The authors tested six core predictions examining how urban-reference contrasts vary across hydroclimatic gradients, whether urban vegetation dependence on precipitation is attenuated, whether empirical precipitation crossovers occur, how spatial heterogeneity differs between land covers, how long-term trajectories diverge within fixed historical footprints over three decades, and whether city-scale socio-demographic indicators provide explanatory power beyond biophysical variables. The water-deficit perspective integrates established biophysical mechanisms rather than proposing novel hydrological physics, examining how regional evaporative demand, moisture supply, and built morphology interact across continental scales.
To isolate internal environmental change from outward metropolitan expansion over time, the researchers established fixed 1995 spatial footprints for fifty-two cities across the conterminous United States using averaged 1994–1995 Landsat imagery and municipal boundaries. These urban footprints included impervious surfaces and urban parks while excluding airports, agricultural parcels, and large natural reserves. Each urban footprint, with a median dissolved area of 109 square kilometers in the NAD83 Conus Albers projection, was paired with the closest available undeveloped, non-agricultural natural landscape large enough for satellite extraction. These reference parcels, possessing a median dissolved area of 27.4 square kilometers and a median centroid separation of 14.9 kilometers, represented regional ecological analogs rather than unobservable pre-settlement conditions. Forty-nine of the fifty-two pairs shared a common Köppen–Geiger climate class and forty-five shared an identical ecological biome, ensuring comparisons reflected local baseline settings rather than regional macroclimatic disparities.
Satellite greenness and thermal measurements were generated in Google Earth Engine using Tier 1 Level-2 surface reflectance and surface temperature archives from Landsat 5, 7, 8, and 9 spanning May through September between 1995 and 2024 to capture deciduous leaf-on conditions. Reflectance bands across sensors were aligned using published transformation coefficients, while Landsat 7 observations recorded after 2018 were excluded due to orbital drift. Spatial statistics received metric-specific multiplicative adjustments calibrated against overlap periods between 2013 and 2018, and Landsat 5 thermal data were corrected for solar-zenith angles using 1999–2011 overlap records with Landsat 7. Processing algorithms masked cloud, shadow, water, snow, and radiometrically saturated pixels using native quality flags, while atmospheric opacity and aerosol metrics filtered poor scenes. Extractions retained only scenes with less than 25 percent land cloud cover where at least 75 percent of the maximum polygon pixel support remained valid.
Principal multivariable repeated-measures models fit by maximum likelihood showed that precipitation, atmospheric evaporative demand, and tree canopy cover strongly govern surface climate across the continental gradient. Across the sample, tree canopy cover was the sole statistically supported individual predictor of urban greenness and land surface temperature, yielding adjusted model R-squared values of 0.84 for greenness and 0.65 for surface temperature. Conversely, reference greenness was associated with cumulative precipitation, climate water deficit, and canopy cover (adjusted R-squared of 0.97), whereas reference surface temperature was associated with climate water deficit and canopy (adjusted R-squared of 0.87). When conditioning on reference covariate values, modeled urban-reference differences narrowed to negative 0.0434 plus or minus 0.0015 for greenness and positive 0.442 plus or minus 0.124 degrees Celsius for surface temperature. In cross-city regressions, relative greenness differences accounted for 89 percent of modeled surface temperature differences, with every 0.10 increase in relative urban greenness corresponding to a 3.31 degrees Celsius reduction in relative urban surface temperature.
Spatial dispersion analyses revealed that complete urban footprints exhibited consistently greater internal variability in vegetation greenness than their paired natural reference counterparts across every sampled metropolitan area. Urban greenness coefficients of variation and standard deviations were positive for all fifty-two pairs, with full-footprint standard deviation differences averaging 0.0792 (95 percent confidence interval 0.0691 to 0.0890), while interquartile ranges were higher in fifty of fifty-two cities. Urban surface temperature standard deviations were higher in 71 percent of metropolitan footprints. Because whole-polygon metrics can be influenced by boundary extent and geometry, the team evaluated equal-area sensitivity samples using circular 3.14-square-kilometer windows paired by exact acquisition date and sensor. In all thirty-six qualifying equal-area comparisons, urban greenness standard deviation remained consistently elevated, with a mean difference of 0.0804 and an average coefficient of variation increase of 26.6 percentage points, confirming that heightened urban greenness heterogeneity is an intrinsic structural feature rather than an artifact of sampling boundaries.
Evaluation of moisture relationships demonstrated that while both landscape types responded to antecedent rainfall, urban vegetation consistently exhibited flatter precipitation response curves than adjacent natural reference systems. Statistical repeated-measures models identified significant urban interaction terms with both four-month cumulative precipitation and its quadratic term (p < 0.001), indicating diminished responsiveness and reduced downward concavity across urban ecosystems. Discontinuous piecewise-linear regressions placed empirical sensitivity breakpoints at 267 millimeters of four-month precipitation for urban areas and 202 millimeters for natural references, illustrating how rainfall sensitivity was strongest under dry conditions and diminished toward wetter baselines. However, the researchers emphasized that these breakpoint values varied across alternative functional specifications and minimum segment constraints, and several climate groupings lacked stable empirical crossovers. As a result, the authors characterized fitted change points as descriptive features of mathematical curves along continuous hydroclimatic gradients rather than fixed biological or process-based ecological thresholds.
Long-term multidecadal trajectories within fixed spatial footprints revealed a pronounced divergence in vegetation dynamics and surface temperature trends between urban centers and surrounding natural lands. Across fifty-one city-reference pairs meeting annual-composite criteria—requiring at least fifteen paired years spanning two decades or more, which excluded Miami—natural reference landscapes exhibited a mean greening trend of positive 1.29 × 10⁻³ per year, whereas urban greenness stagnated or declined slightly at negative 1.88 × 10⁻⁴ per year. This produced a statistically significant mean paired divergence of negative 1.48 × 10⁻³ per year (p < 0.001), a pattern corroborated by first-order autoregressive generalized least squares and Theil–Sen non-parametric estimators. Urban browning was most acute in desert cities, consistent with the idea that maintaining vegetated landscapes becomes more difficult under drought and heat stress, underscoring that baseline greenness alone does not predict long-term change.
Concurrently, satellite thermal data confirmed that while both developed footprints and undeveloped reference landscapes experienced warming between 1995 and 2024, urban surfaces warmed substantially faster on average. Mean annual-composite warming rates reached positive 0.141 degrees Celsius per year within cities compared to positive 0.115 degrees Celsius per year in reference sites, yielding a paired divergence of positive 0.0252 degrees Celsius per year (p < 0.001), equivalent to approximately 22 percent faster surface warming in urban areas. This divergence was reproduced by autoregressive generalized least squares (+0.0247 degrees Celsius per year) and Theil–Sen estimators (+0.0238 degrees Celsius per year), maintaining stability across cloud-screening, sensor-harmonization, and observation-frequency sensitivity tests. However, accelerated urban warming was not universally observed across all arid locations, where natural backgrounds also warmed rapidly, and statistical significance diminished when restricting analyses to a seventeen-pair subset requiring all five warm-season months, demonstrating that continental warming differentials vary by regional climatic regime.
Investigating city-level socio-demographic indicators from decennial U.S. Census tract records from 2000, 2010, and 2020, the researchers discovered that demographic factors provided limited explanatory power for cross-city environmental variations after controlling for baseline biophysical and geographic predictors. In exploratory spatial slope regressions, numerical conditioning based on dispersion ratios excluded six arid reference series (Phoenix, Las Vegas, Salt Lake City, El Paso, Reno, and Riverside). Across the remaining forty-six pairs, urban footprints displayed less negative spatial slopes between temperature and greenness (+4.05 degrees Celsius per unit NDVI), but this difference became indistinguishable from zero when restricted to shared NDVI support or discrete vegetation thresholds.
The authors emphasized critical methodological boundaries, noting that proximity-based selection of reference sites provides contemporary ecological analogs rather than historical pre-development reconstructions, and that fixed 1995 perimeters capture internal landscape shifts rather than post-1995 spatial expansion. Furthermore, Landsat sensors measure radiometric surface skin temperature during morning orbital overpasses rather than ambient air temperature or direct human physiological thermal stress. The authors also highlighted that while the inter-urban water-deficit framework successfully integrates observed empirical patterns, supplemental irrigation volumes and parcel-level water consumption were not directly measured, leaving specific hydrological mechanisms unpartitioned.
Subject of Research: Continental-scale satellite analysis of urban-reference greenness and surface temperature patterns across U.S. climates
Article Title: Continental-scale urban–reference patterns in greenness and land surface temperature across U.S. cities
Article References: Kucera, D., Pataki, D., & Jenerette, G. D. (2026). Continental-scale urban–reference patterns in greenness and land surface temperature across U.S. cities. Landscape Ecology, 41(11), Article 221. https://doi.org/10.1007/s10980-026-02475-z
Image Credits: AI Generated
DOI: 10.1007/s10980-026-02475-z
Keywords: Landsat, urban heat island, land surface temperature, normalized difference vegetation index, hydroclimate, climate water deficit, urban greenness, tree canopy cover, spatial heterogeneity, inter-urban water-deficit framework, precipitation sensitivity, conterminous United States, TerraClimate, decadal warming trends, ecological homogenization, Google Earth Engine, Theil-Sen estimator, Köppen-Geiger classes
News Source: Drew Townsend. (November 17, 2026). U.S. Cities Warm Twenty-Two Percent Faster as Regional Greenness Differences Reverse. Scienmag.



