Urban Climate Analysis

Quantify the environmental impact of codified urban planning using satellite remote sensing and geospatial analysis. Primary objective: measure vegetation outcomes of planned green space via NASA AG100 NDVI and emissivity rasters across neighboring towns.

PythonRoot Cause AnalysisSustainability

Sole researcher

Aug 2023 - Oct 2023

Analyzed 5,906 raster data points aggregated into 3,854 H3 hexagons. Reston recorded mean NDVI of 52.7 vs. 41.4 in Herndon and 37.5 in McNair, indicating that mid-density housing within preserved canopy corridors maintains higher vegetation coverage without reducing residential density.

Introduction

Reston, Virginia is recognized for its planned urban framework. Green spaces are legally codified in its development charter, producing a built environment distinct from neighboring communities. I wanted to move beyond anecdote and quantify the actual environmental impact of those design choices using satellite remote sensing data.

The study overlays NASA AG100 emissivity/NDVI rasters with GCIS block-level population density across Reston, Herndon, and McNair — three communities sharing similar geography and climate, built around very different planning principles.

Cities

Data Points

5,906

Hexagons

3,854

Reston NDVI

52.7

Resolution

H3 · R10

Methodology

Data sources: NASA AG100 emissivity raster (surface temperature + NDVI) and GCIS 2010 block population density, at incompatible spatial resolutions.

Resolution mismatch: Direct pixel comparison was impossible. H3 hexagonal spatial indexing at resolution 10 (~75 m edge length) gave both datasets a shared spatial unit, with all raster points within each hexagon averaged to a single value.

Outlier filtering: Census data contained anomalies: zones legally designated 1-2 DU/acre reporting ~23,000 persons/km² (two orders of magnitude off). A 1,000 persons/km² threshold isolated low-density residential areas for fair comparison.

Key metric: NDVI-to-density (ndvi_scaled × den_scaled) captures vegetation coverage relative to population, enabling a density-normalized comparison across all three cities.

PythonGeoPandasH3 IndexingNASA AG100GCISscikit-learnrasterioEllipsis Drive

Key Findings

At matched population densities, Reston's NDVI-to-density score is 70% higher than Herndon and 2.3× higher than McNair, confirming that planned green infrastructure produces measurably better environmental outcomes regardless of population count.

Vegetation Health

Reston's mean NDVI of 52.7 exceeds Herndon (41.4) and McNair (37.5) across all density zones, a persistent gap driven by intentional green space placement around multi-unit housing.

Temperature Effect

Higher NDVI correlates with lower scaled surface temperature across the dataset. Reston hexagons with the highest vegetation scores cluster around multi-level housing with surrounding tree canopy.

Best Pattern

Low-rise and mid-rise apartment complexes embedded in green areas produced the highest NDVI-to-density scores, the specific building typology Reston was designed around.

Worst Pattern

Dense suburban development near Dulles (MetroPark at Arrowbrook, The Point at Ridgeline) showed low NDVI despite similar population density, with minimal setback vegetation and large impervious surfaces.

Aggregated Results by City

City	Mean Density (persons/km²)	Mean NDVI	NDVI / Density · low-density zones
Reston	1,635	52.7	0.00743
Herndon	2,441	41.4	0.00438
McNair	3,869	37.5	0.00321

Challenges

Raster resolution mismatch: NASA and GCIS datasets use different grid spacings, making direct overlap meaningless. H3 hexagonal bucketing resolved this by giving both datasets a shared spatial unit.

Population density anomalies: Multiple blocks in low-density zones registered 100+ persons/acre in the census raster, two orders of magnitude beyond zoning data. Identified via histogram analysis and filtered before drawing conclusions.

Correlation vs. causation: No direct causal link between NDVI and temperature was isolated at this scale. Confounding variables like building materials, street width, and impervious surface fraction were not independently controlled.

Selected Observations

Fig 1 · Poor Planning

MetroPark at Arrowbrook (38.955, -77.409). Dense development near Dulles with minimal green setbacks. Low NDVI and elevated temperature despite density comparable to Reston.

Fig 2 · Best Performers

Southgate Square Center (38.943, -77.36). Multi-unit housing surrounded by canopy. Highest NDVI-to-density scores in the dataset.

Fig 3 · H3 Coverage

3,854 hexagons covering Reston, Herndon, and McNair after NaN removal. Blue clusters correspond to Reston's planned green corridors versus red zones in neighboring towns.

Conclusion

Using H3 spatial indexing to normalize two incompatible raster datasets, I demonstrated that Reston's urban planning, specifically its green corridors and intentional placement of multi-unit housing within vegetated areas, produces statistically higher NDVI and lower surface temperatures per unit of population density compared to neighboring towns.

This supports the hypothesis that zoning policy directly shapes environmental outcomes. The pattern identified — mid-rise apartments in maintained green areas — is reproducible and scalable, pointing to a concrete design prescription for sustainable urban infill.

Read Full Paper