MENU



→ What Makes NYC Hotter
→ Where The Rent Goes in NYC
→ Unused Floor Area Ratio and Skyscraper Development in Manhattan→ About Me


What Makes NYC Hotter


Yuanhao Wu
Keywords: Urban Heat Island, NYC, Spatial Analysis, Remote Sensing, Climate & Design, Data Visualization
GitHub Repository: https://github.com/KharleWu/UHI_NYC

Overview


New York City’s dense built environment makes it significantly warmer than its surrounding suburbs, a phenomenon known as the Urban Heat Island (UHI) effect. This project investigates what makes NYC hotter, combining satellite remote sensing, spatial statistics, and urban data modeling to visualize and explain the city’s temperature patterns.

Using Landsat 8–9 thermal infrared imagery from 2022–2025, I derived surface temperature maps and aggregated them to the census block group level. By overlaying over 60 clear-day rasters and applying Principal Component Analysis (PCA) and Geographically Weighted Regression (GWR), the study quantified how factors like building density, green coverage, transportation networks, and land use diversity contribute to local heat intensity.

The resulting composite model achieved a strong fit (R² = 0.88), revealing clear spatial trends:

  • High-density and transit-heavy zones, such as Midtown Manhattan, Downtown Brooklyn, and the Financial District, recorded the highest heat levels.
  • Green and open-space areas, including Central Park and Staten Island’s residential districts, consistently appeared cooler.

These insights translate directly into urban design guidance: increasing vegetative cover, integrating reflective and permeable materials, and rebalancing built density can measurably reduce surface heat and improve environmental equity across NYC neighborhoods. This project transforms raw satellite data into actionable spatial intelligence, helping city planners, policymakers, and the public better understand how urban form shapes local climate resilience.

Menu

  • So Close, But So Different
  • Urban Heat Island Effect in NYC
  • What is Contributing the Urban Heat Island Effect?
  • Results and Interpretation



So Close, But So Different


Central Park and White Plains are both located in the state of New York. The distance between these two places is 25 miles, which is approximately a 40-minute car drive. But, despite the close distance between them, do these two places have similar temperatures?


Based on historical temperature data from the National Weather Service (NWS), I obtained the monthly average temperatures for Central Park and White Plains from 2005 to 2024. As shown below, the line chart on the left displays the average temperatures for both locations over time, while the chart on the right illustrates the temperature difference between Central Park and White Plains (The Difference = Central Park temperature – White Plains temperature).




From the line chart, it is clear that the monthly average temperature in Central Park is significantly higher than that in White Plains, with a mean difference of 3.52°F. Despite the close proximity of the two locations and the absence of notable differences in elevation or mountain barriers, such a pronounced temperature gap still exists. 


- So, what could be the reason behind this?
It’s because of the Urban Heat Island (UHI).




Urban Heat Island Effect in NYC

 
The urban Heat island effect means that urbanized areas experience higher temperatures than outlying areas. Structures such as buildings, roads, and other infrastructure absorb and re-emit the sun’s heat more than natural landscapes such as forests and water bodies. Urban areas, where these structures are highly concentrated and greenery is limited, become “islands” of higher temperatures relative to outlying areas. Daytime temperatures in urban areas are about 1–7°F higher than temperatures in outlying areas, and nighttime temperatures are about 2–5°F higher.

- Existing Analysis
Climate Central conducted an in-depth analysis of the Urban Heat Island (UHI) effect across 44 major U.S. cities, using census tracts as the unit of measurement. The resulting Urban Heat Island (UHI) Index quantifies how much hotter each area is, in degrees Fahrenheit, as a result of the built environment’s characteristics. The analysis revealed that New York City exhibits an average UHI Index of 8.6 °F.

The map below illustrates the distribution of UHI Index values across census tracts in New York City. Darker areas indicate zones with the most intense heat island effects. You can also use the slide bar to adjust the minimum UHI Index displayed on the map, allowing you to explore which areas experience more intense heat.

 – Defining My UHI Index

To enable a more granular and reliable analysis of New York City’s Urban Heat Island (UHI) effect, I developed a customized evaluation standard. I used Surface Temperature data from the USGS Landsat 8–9 Collection 2 Level 2 to assess citywide temperature variations. 

From the available spectral bands, I focused on Band 10 (ST_B10.TIF), which records emitted thermal radiation rather than reflected sunlight. This makes it ideal for mapping surface temperature patterns across the city. 

Using Rasterio, I clipped the raster to NYC’s borough boundaries, and with GDAL’s Polygonize tool, I converted pixels (30m × 30m) into vector polygons containing temperature values. Since the dataset stores integer-based Digital Numbers (DN), I applied a standard linear scaling to convert them into Fahrenheit.

After preprocessing, the dataset contained nearly 900,000 polygons (~300 MB). To improve performance and ensure consistency with demographic data, I aggregated pixel-level temperatures to the census block group level using an area-weighted averaging method, reducing the dataset to about 15,000 records (~20 MB). Here is an example of the workflow:

1. Level-2 Reflective Raster
2. Level-2 ST-B10 Raster
3. Clip Raster to NYC
4. Convert the data


To identify consistent spatial and temporal patterns, I compiled approximately 60 surface temperature raster layers with minimal cloud coverage, spanning 2022 to 2025. Each image was captured around 3:30 PM Eastern Time to ensure temporal consistency across the dataset. From this collection, I selected 12 representative dates. Using the Natural Breaks (Jenks) classification method from the Mapclassify library, I standardized the legend categories for each raster to better represent temperature variations. All selected dates are displayed in YYYYMMDD format for clarity and uniformity.



To further process the raster data and conduct the final UHI (Urban Heat Island) assessment, I chose to overlay all the collected raster layers with equal weighting. By aggregating them in this way, I generated a single, composite raster that represents the overall surface temperature pattern. The resulting surface temperature estimates are as follows:





What is Contributing the Urban Heat Island Effect?



Urban heat islands (UHIs) emerge from urban design and land cover patterns that disrupt a city’s natural thermal balance. Key contributing factors include low-albedo surfaces, loss of vegetation, dense building geometry, anthropogenic heat emissions, urban surface composition, air pollution, and the broader influence of climate change. Together, these variables intensify surface temperatures and reduce environmental resilience in densely developed areas.

To investigate how these factors interact across New York City, I compiled 24 distinct datasets from multiple public sources for integrated spatial analysis. The major data sources include:


Because this study focuses on New York City, I assume that the overall temperature distribution is relatively consistent citywide. Based on this assumption, I use the average surface temperature of each census block group as the dependent variable, representing localized UHI intensity. All other variables serve as independent predictors, capturing demographic, built environment, and land cover characteristics, to quantify their respective influence on the city’s heat dynamics.


- Phase 1: Assessing Spatial Autocorrelation
Before exploring the factors driving urban temperature variation, it is essential to determine whether temperature values exhibit spatial clustering. To test this, I applied Moran’s I, a global measure of spatial autocorrelation that quantifies how similar temperature values are distributed across space. A positive Moran’s I indicates that similar temperature values tend to cluster, while the accompanying p-value tests whether this pattern is statistically significant.

In this analysis, the results showed a Moran’s I of 0.7512 (p = 0.001) , confirming a strong and statistically significant spatial autocorrelation. This finding validates the presence of a spatial structure in temperature distribution and supports the use of spatially explicit modeling techniques in subsequent analyses.

- Phase 2: Identifying Key Variables

After confirming the spatial structure of temperature, I examined its statistical relationship with a range of demographic, environmental, and built-environment variables. Using the Pearson correlation matrix, I identified variables most strongly associated with the average temperature at the census block group level.

To reduce noise and improve interpretability, I retained only variables with an absolute correlation coefficient ≥ 0.1, ensuring that only meaningful predictors were carried forward. These selected variables form the foundation for subsequent modeling of the key drivers behind the Urban Heat Island (UHI) effect.


 - Phase 3: Principal Component Analysis (PCA) and Geographically Weighted Regression (GWR)
To reduce multicollinearity and uncover underlying patterns within the explanatory variables, I applied Principal Component Analysis (PCA) to the 14 variables identified through correlation filtering (|r| ≥ 0.1). PCA transforms correlated variables into uncorrelated components that capture the majority of the variance in the dataset. The first eight components explained over 80% of total variance.



Building on this dimensionality reduction, I employed Geographically Weighted Regression (GWR) to model how these principal components influence surface temperature across New York City. Unlike traditional global regressions, GWR captures spatial heterogeneity by allowing the relationship between predictors and temperature to vary locally.
The GWR model achieved a strong overall fit (R² = 0.88), confirming high explanatory power. By mapping the local R² values and spatially varying coefficients, the analysis reveals distinct geographic patterns in the strength and direction of influence across neighborhoods.




Results and Interpretation



Among the eight principal components used in the GWR model, PC4, PC7, and PC8 demonstrated the highest overall influence, with mean absolute coefficients of 9.29, 8.09, and 6.13, respectively. These components represent the dominant latent factors shaping temperature variability across the city.



Principal Component Key Interpretation Urban Context
PC4 High traffic density and newer, high-density construction Intensively developed areas with limited open or vegetated spaces exhibit elevated surface temperatures due to increased heat retention and anthropogenic emissions.
PC7 Greater share of public facilities and green spaces, lower vehicle density These areas show stronger cooling effects through evapotranspiration and reduced waste heat, making them more resilient to UHI effects.
PC8 Moderate building volumes, substantial greenery, and higher-educated residents Reflects planned residential neighborhoods that integrate environmental design elements, leading to localized cooling benefits.

The combined PCA loadings and GWR variable importance analysis confirm that built-environment and land-cover characteristics exert the strongest influence on surface temperature distribution. Dense, modern infrastructure and traffic intensity amplify heat accumulation, while areas with higher vegetation cover and public open space remain comparatively cooler.

These findings underscore the crucial role of urban form, land use, and green infrastructure in mitigating New York City’s Urban Heat Island (UHI) effect, offering spatially grounded insights for climate resilience planning and sustainable design.
© 2025 Yuanhao Wu. All Rights Reserved.