United States (52 Cities): How Urbanization Reshapes Greenness, Temperature & Cooling (1995–2024)

Summary

  • Question. Across the U.S., how does urbanization change greenness (NDVI), land-surface temperature (LST), and vegetative cooling compared with nearby non-urban sites—and how do water balance and urban form control those changes?

  • What we did. Built a 29-year Landsat time series (1995–2024) for 52 cities + paired reference areas; compared means, spatial variability, and trends; related differences to precipitation, climate water deficit, tree canopy, and building volume.

  • Outcome. Identified precipitation thresholds (~312–386 mm) where urban and non-urban responses diverge; found cities warm faster and green more slowly than references, with a nationwide decline in vegetative cooling and consistent fragmentation of urban vegetation—all crucial for heat-mitigation and canopy planning.

Six scatter plots showing relationships between environmental factors and plant or climate variables, with regression lines for urban and reference models, including R-squared values.

Research Challenge

City leaders hear conflicting messages: in some places cities are greener and cooler than surroundings (irrigated “oasis” effect), while elsewhere cities are hotter and less green. Planners needed a clear, continental-scale answer that separates climate and water availability from urban form and shows how these relationships evolve over decades. This meant harmonizing satellites, fixing urban footprints through time, and comparing each city to a nearby natural analog—not just looking at cities in isolation.

Approach

Data

  • Landsat 5/7/8/9 (Level-2, Collection 2) → NDVI, LST (May–Sep each year).

  • TerraClimate (precipitation, PET, CWD, VPD) and SPEI drought.

  • NLCD tree canopy & impervious cover; JRC GHSL building volume (3-D form).

  • SRTM elevation; paired non-urban reference sites per city.

Methods

  • Fixed 1995 urban boundaries to isolate within-city change.

  • Cloud-masked Landsat; cross-sensor reflectance harmonization; tract/area aggregation.

  • Computed urban–reference deltas (ΔNDVI, ΔLST, Δcooling), spatial variability (CV), and linear trends (1995–2024).

  • ANCOVA / breakpoint regressions to test precipitation & water-deficit thresholds and urban vs. reference slope differences.

Tools

  • Google Earth Engine (data ingest, masking, harmonization), Python for modeling/QC, ArcGIS Pro for spatial operations.

Scientific graph panel with four scatter plots showing correlations between environmental variables and vegetation or climate metrics, including tree canopy cover, vapor pressure deficit, NDVI, and climate water deficit.

Key Findings

  • Thresholds & decoupling. Above ~312 mm (NDVI) and 386 mm (LST) annual-precipitation equivalents, cities decouple from rainfall compared with their references—consistent with supplemental irrigation and built constraints.

  • Means & variability. Urban areas show higher NDVI variability in 100% of cities (fragmented vegetation). LST variability is mixed and depends on tree canopy and cooling strength. Higher built volume → more NDVI variability.

  • Trends (1995–2024).

    • Warming: 99% of cities warmed; mean +0.18 °C/yr (about 30% faster than references, +0.14 °C/yr).

    • Greenness: Cities, on average, green six-times more slowly than references; losses strongest in desert cities and where building volume is high.

    • Vegetative cooling: All cities show declines (≈ −0.18 °C per NDVI unit per year, ~76% faster decline than references).

  • Drivers. Inter-urban patterns are dominated by biophysical controls (water balance, canopy, building volume); socio-demographics add little at the national scale (though intra-city equity remains critical).

Four scientific graphs showing relationships between precipitation, temperature, and climate sensitivity with various data points and trend lines.

Outputs

  • Deliverables:

    • National workbook: ΔNDVI, ΔLST, Δcooling maps & tables (city vs. reference).

    • Threshold charts (precipitation, water deficit) and driver plots (canopy, building volume).

    • Reproducible code (GEE + Python) and city-level CSVs/GeoTIFFs.

  • Status: Manuscript in review at Landscape and Urban Planning (preprint/shareable summary on request).

Impact / How it’s used

  • Heat mitigation strategy: identifies where vegetation won’t deliver expected cooling without water management, and where canopy does most to stabilize temperatures.

  • Canopy & irrigation policy: uses the ~350 mm precipitation threshold to tailor irrigation rules and planting targets by climate zone.

  • Benchmarking: provides city vs. natural-analog baselines to track progress on cooling and greening goals over time.