Snow Accumulation and its Influence on Social Interaction in Built Environments

A Philadelphia Study

Kwadwo Afari Gyan

Urban environments are shaped not only by their physical structures but also by the seasonal conditions that influence how people move, interact, and occupy space. Winter, in particular, introduces a set of environmental constraints that can dramatically alter the usability of public spaces. Snow accumulation, surface icing, and reduced mobility can transform otherwise vibrant streets and plazas into difficult, inaccessible terrains. These seasonal dynamics are especially important in cities like Philadelphia, where winter storms are frequent enough to disrupt daily life but not so constant that the built environment is fully adapted to them.

A central observation motivating this study is that impervious surfaces — streets, sidewalks, parking lots, and paved plazas — tend to accumulate snow in ways that directly affect walkability and social interaction. Unlike vegetated or porous surfaces, impervious materials do not absorb meltwater or allow snow to dissipate naturally. Instead, they create large, continuous surfaces where snow can settle, compact, and persist. This increases the area of obstruction during winter events and can make everyday activities — walking to transit, accessing services, or simply moving through a neighbourhood — significantly more difficult. In this sense, the built environment does not merely host snow; it amplifies its effects.

Yet, despite the intuitive nature of this relationship, the spatial patterns of snow accumulation across different urban forms are not well documented at the parcel or neighbourhood scale. Most studies focus on regional snow cover, meteorological patterns, or transportation impacts. Few examine how snow interacts with specific urban parcels, especially in relation to their built-up characteristics. This gap is important because winter accessibility is experienced at the human scale: the sidewalk in front of a home, the plaza outside a school, the path through a park.

This study addresses that gap by using multi-year Sentinel-2 satellite imagery (2015–2026) to quantify snow accumulation and built-environment characteristics across selected sites in Philadelphia. The analysis focuses on two key questions:

• How much snow accumulates on each site during peak winter conditions?
• Does the proportion of impervious surface influence the extent or intensity of snow cover?

The study focuses on eight parcels of 0.25-mile walkable distance each within Philadelphia, selected by random sampling. These sites represent real locations of interest in ongoing urban design, geography, and urban planning work. They provide a range of parcel sizes, land uses, and micro-environmental conditions. Using parcel-level polygons allows snow and built-environment metrics to be extracted at a human-scale resolution, which is appropriate for studying walkability and winter accessibility.

Sentinel-2 Level-2A surface reflectance imagery was used because it offers high spatial resolution (10–20 m) suitable for neighbourhood-scale analysis, a frequent revisit time (5 days) ensuring multiple winter observations, and the spectral bands needed to compute snow and built-up indices (NDSI, NDVI, NDWI, NDBI). Using a consistent, long-term satellite archive (2015–2026) allows snow patterns to be compared across years and ensures that results are not biased by a single winter season.

The "snowiest day" for each site was identified by computing the Normalized Difference Snow Index (NDSI) for every winter image and selecting the one with the highest mean NDSI. Four snow metrics were computed from this snowiest winter image:

• Winter Mean NDSI — captures overall snow intensity
• Winter Max NDSI — identifies the brightest, densest snow areas
• Winter Min NDSI — shows where snow is thinnest or mixed
• Winter Snow Fraction (NDSI > 0.4) — measures how much of the site is snow-covered

A free cloud-based summer image was used to compute summer NDSI values as a control, confirming that winter snow signals are not caused by bright roofs, concrete, or misclassification.

A supervised land-cover classification was performed using NDVI, NDWI, and NDBI indices, followed by a Random Forest classifier. Summer vegetation is captured to ensure that vegetation is depicted and built surfaces are not obstructed by snow. The output built fraction quantifies imperviousness, which is central to this research question.

Correlation matrices, scatterplots, and linear regression models were used to examine relationships between built fraction and snow metrics. All metrics were exported from Google Earth Engine as a CSV file and analyzed in R. Figures were exported as high-resolution PNGs and PDFs.

Across all sites, Winter Snow Fraction ranged from 0.58 to 0.90, while Winter Mean NDSI ranged from 0.44 to 0.60. This indicates strong and consistent snow presence across all sampled locations. The data confirms that during peak winter conditions, a substantial proportion of each site is snow-covered.

Built fraction values were extremely low, ranging from 0.00 to 0.02. This reflects the low-density character of the sampled parcels. Table 3.1 presents land acreage data across all eight sites.

Source: Author's construct, 2026. Data derived from parcel-level analysis and Sentinel-2 imagery.

Built fraction showed weak or near-zero correlation with snow metrics. This is attributable to the lack of variation in built fraction across sites — all sites were essentially open or low-density, preventing the statistical model from detecting the effect of impervious surfaces on snow retention. Scatterplots showing built fraction vs. Winter Mean NDSI and Winter Snow Fraction showed flat or weak trends because built fraction was nearly constant across all eight sample sites.

Although statistical relationships between imperviousness and snow accumulation were weak — due to the uniformly low built fraction values in the dataset — the conceptual mechanism remains valid. Impervious surfaces expand the area where snow accumulates and remains exposed, increasing obstruction and reducing walkability and social interaction in built spaces.

This interpretation takes into consideration the dominant attention given to automobile usage. During snowstorms, there is high emphasis on clearing roads before considering other spaces that serve other modes of transport such as bike lanes and walkways. These are places of convergence and conversation — a thought likened to arguments made by Jane Jacobs (1961). When there is snow, there is increasing obstruction of these spaces that harbour social interaction.

The snow metrics confirm that snow covers a large proportion of each site. If these sites had higher imperviousness, more of that snow would fall on walkable surfaces, directly affecting accessibility.

This study successfully quantified snow accumulation and built-environment characteristics using multi-year Sentinel-2 imagery. While snow metrics show strong winter coverage, built fraction values were too low and too similar to reveal a statistical relationship between imperviousness and snow accumulation. The results highlight the importance of designing public spaces that remain functional during snow events and underscore the need for winter-responsive planning strategies that support social interaction and mobility.

Copernicus. (2023). Sentinel-2 MSI Level-2A Surface Reflectance Product. European Commission.

Cutter, S. L., Boruff, B. J., & Shirley, W. L. (2003). Social vulnerability to environmental hazards. Social Science Quarterly.

ESA. (2023). Sentinel-1 Synthetic Aperture Radar (SAR) Level-1 GRD Product. European Space Agency.

Gehl, J. (2011). Life Between Buildings: Using Public Space. Island Press.

Hall, D. K., Riggs, G. A., & Salomonson, V. V. (1995). Development of methods for mapping global snow cover using MODIS data. Remote Sensing of Environment.

Jacobs, J. (1961). The Death and Life of Great American Cities. Random House.

Loukaitou-Sideris, A. (1995). Urban form and social interaction in public spaces. Journal of Planning Education and Research.

NASA. (2023). MODIS Snow Cover Daily Product (MOD10A1). National Aeronautics and Space Administration.

National Centers for Environmental Information (NCEI). Daily Snowfall Data, Philadelphia International Airport. https://www.ncei.noaa.gov/cdo-web/datasets

NOAA. (2023). National Snow Analysis (NSA) Snow Depth and Snow Water Equivalent Grids. National Oceanic and Atmospheric Administration.

Shook, E., & Wang, S. (2015). A geospatial cyberinfrastructure for urban snow monitoring. Computers, Environment and Urban Systems.

USGS. (2023). National Land Cover Database (NLCD). United States Geological Survey.

Project Snow X  ·  Kwadwo Afari Gyan  ·  May 2026  ·  Philadelphia, PA