The impact of urban spatial environment on COVID-19: a case study in Beijing

Zhen Yang, Jiaxuan Li, Yu Li, Xiaowen Huang, Anran Zhang, Yue Lü, Xu Zhao, Xueyan YangFrontiers in Public Health 2024 · DOI

This study, published in Frontiers in Public Health, investigates how the "bones" of a city—its buildings, parks, and transit hubs—influenced the frequency of COVID-19 outbreaks in Beijing. By analyzing data at both the city-wide and neighborhood scales, the researchers highlight a critical link between urban design, socio-economic status, and public health resilience.

TL;DR

The researchers used geographically weighted regression and clustering analysis to show that high building density and facility concentration increased COVID-19 outbreak frequency in Beijing, while greenery and water provided a protective effect. Crucially, the impact of neighborhood-level factors like household count and housing prices varied significantly across socio-economic tiers, highlighting the need for "spatial justice" in urban planning.

What Problem Does This Paper Address?

While the link between population density and virus transmission is well-known, the specific nuances of the urban spatial environment are often overlooked. Urban planners and public health officials need to know: Does a park nearby actually help? Do commercial hubs drive outbreaks more than transit stations? And why do some neighborhoods seem more resilient than others despite similar densities?

This paper addresses the gap between macro-level city statistics and micro-level neighborhood characteristics. It seeks to provide a theoretical basis for "spatial justice"—the idea that urban resources (like green space and healthcare) should be distributed fairly to ensure all citizens have an equal shot at staying healthy during a pandemic.

Methods

The authors employed a multi-stage analytical framework:

  1. Spatial Grid Analysis: They divided the area within Beijing’s fifth ring road into 2,769 rectangular cells (500m x 500m) to map outbreak frequency over a one-year period (Nov 2021 – Nov 2022).
  2. Geographically Weighted Regression (GWR): Unlike standard global regression, GWR allows the relationship between variables to change across space. This was used to analyze "external" factors like building density, volume ratio (floor area ratio), and the density of commercial, service, and transportation facilities.
  3. K-Means Clustering: They categorized 344 epidemic-affected neighborhoods into three grades—low-end, mid-end, and high-end—based on housing prices, building age, and household counts.
  4. Binomial Logistic Regression (BLR): This was used to identify which "internal" neighborhood factors (e.g., number of households vs. building age) were the strongest predictors of outbreaks within each socio-economic cluster.

Datasets

  • Epidemic Data: Sourced from the Beijing Municipal Government Data Resource Network, the Health Commission, and Beijing Daily.
  • Urban Spatial Data: Building outlines, heights, and Point of Interest (POI) data for facilities were obtained from Gaode Map.
  • Environmental Data: Water and greenery coverage were extracted from Landsat-8 satellite imagery using supervised classification in ArcGIS.

Key Insights and Ideas

  • The "Greenery Shield": Water and greenery coverage were the only external factors negatively correlated with outbreak frequency. This suggests that urban nature isn't just an aesthetic luxury; it acts as a buffer, potentially by improving air quality or providing lower-density spaces for movement.
  • Facility Density as a Mobility Proxy: The density of commercial and transportation facilities showed a strong positive correlation with outbreaks. These areas act as "hubs" where high human mobility and frequent contact create ideal conditions for viral spread.
  • The Socio-Economic Buffer: In high-end neighborhoods, internal factors like building age or household count had no significant correlation with outbreak frequency. The authors attribute this to better healthcare access, stricter management, and more spacious living conditions that "decouple" the physical environment from infection risk.
  • Low-End Vulnerability: In low-end neighborhoods, higher housing prices (within that tier) and higher household counts were positively correlated with outbreaks. This points to the "shared apartment" phenomenon, where high-density, low-cost housing leads to unavoidable close contact.

Key Results

  • Model Accuracy: The GWR model achieved an $R^2$ of 0.79, indicating that nearly 80% of the spatial variation in outbreaks could be explained by the external environmental factors studied.
  • External Drivers: Building density, volume ratio, and facility densities (commercial, service, transport) were all positively correlated with COVID-19 frequency.
  • Neighborhood Tiers: The clustering analysis identified 185 low-end, 101 mid-end, and 58 high-end neighborhoods.
  • Internal Drivers: For mid-end neighborhoods, the number of households was the primary predictor of outbreak frequency ($p < 0.05$).

Limitations and Future Work

The authors acknowledge several limitations:

  • Temporal Resolution: The study uses a 7-day cycle for outbreak records, which may smooth over rapid daily changes in transmission.
  • Behavioral Factors: The model does not account for individual behaviors (mask-wearing, vaccination status) or real-time human mobility data (e.g., cell phone pings).
  • Beijing Specificity: The results are heavily influenced by Beijing’s unique urban form, such as high-density "hutong" areas and large-scale wholesale markets.
  • Future Directions: The researchers suggest exploring "informal blue-green spaces" (like street-side greenery) and investigating how to better distribute medical resources to low-income communities to improve overall urban resilience.

Summary updated

Abstract

Epidemics are dangerous and difficult to prevent and control, especially in urban areas. Clarifying the correlation between the COVID-19 Outbreak Frequency and the urban spatial environment may help improve cities' ability to respond to such public health emergencies. In this study, we firstly analyzed the spatial distribution characteristics of COVID-19 Outbreak Frequency by correlating the geographic locations of COVID-19 epidemic-affected neighborhoods in the city of Beijing with the time point of onset. Secondly, we created a geographically weighted regression model combining the COVID-19 Outbreak Frequency with the external spatial environmental elements of the city. Thirdly, different grades of epidemic-affected neighborhoods in the study area were classified according to the clustering analysis results. Finally, the correlation between the COVID-19 Outbreak Frequency and the internal spatial environmental elements of different grades of neighborhoods was investigated using a binomial logistic regression model. The study yielded the following results. (i) Epidemic outbreak frequency was evidently correlated with the urban external spatial environment, among building density, volume ratio, density of commercial facilities, density of service facilities, and density of transportation facilities were positively correlated with COVID-19 Outbreak Frequency, while water and greenery coverage was negatively correlated with it. (ii) The correlation between COVID-19 Outbreak Frequency and the internal spatial environmental elements of neighborhoods of different grades differed. House price and the number of households were positively correlated with the COVID-19 Outbreak Frequency in low-end neighborhoods, while the number of households was positively correlated with the COVID-19 Outbreak Frequency in mid-end neighborhoods. In order to achieve spatial justice, society should strive to address the inequality phenomena of income gaps and residential differentiation, and promote fair distribution of spatial environments.
BibTeX
@inproceedings{yang2024impact,
  author = {Zhen Yang and Jiaxuan Li and Yu Li and Xiaowen Huang and Anran Zhang and Yue Lü and Xu Zhao and Xueyan Yang},
  title = {{The impact of urban spatial environment on COVID-19: a case study in Beijing}},
  booktitle = {Frontiers in Public Health},
  year = {2024},
  doi = {10.3389/fpubh.2023.1287999},
}
Figure