From sidewalks to graffiti: tracking neighborhood conditions across time with street view

The built environment exerts a powerful influence on behavior and well-being. Features such as sidewalks, street lighting, building maintenance, and graffiti not only shape opportunities for physical activity but also influence perceptions of safety. Population health researchers often use Google Street View (GSV) audits to assess neighborhood conditions, which is a cost-effective and reproducible alternative to in-person observation. Yet one persistent challenge has been the irregular timing of GSV imagery collection, which raises questions about whether using imagery from the “wrong” year might bias results. Published in the Journal of Urban Health, a new study addresses this issue directly with epidemiologist and lead author Dr. Stephen Mooney describing the work as “the first study to really look systematically at how much visual indications of neighborhood conditions on Street View change over time.”

Dr. Robert Kaplan, an epidemiologist and population health scientist at Fred Hutch Cancer Center, and his colleagues, sought to determine the extent to which neighborhood features visible in Street View change over time, and to evaluate whether modest mismatches between imagery dates and periods of epidemiological interest pose a serious threat to validity. The research team focused on four U.S. cities—New York, Chicago, Miami, and San Diego—where the Hispanic Community Health Study/Study of Latinos (HCHS/SOL) cohort has been recruited. These cities provide diverse cultural, economic, and physical contexts, making them an ideal setting for testing temporal stability in neighborhood features.

Researchers sampled nearly 5,000 sites across the four metropolitan areas and ultimately analyzed 2,118 locations with both older and newer high-quality imagery. The images spanned from 2007 to 2023, with a median gap of 10.5 years between the oldest and most recent photographs. Trained auditors, using the Computer Assisted Neighborhood Visual Assessment System (CANVAS), followed a “drop-and-spin” protocol to record twenty built environment features while rotating 360° from a single fixed Street View vantage point. These features included indicators of disinvestment such as vacant lots and buildings in disrepair, as well as pedestrian safety elements such as crosswalks and sidewalk quality.

When the data were analyzed, a striking pattern emerged: most neighborhood features had barely changed over time. Seventeen of the twenty variables assessed were unchanged at over 90 percent of locations. Structural features such as sidewalks, vacant lots, and street lighting were especially consistent, while more transient features such as litter, graffiti, and building repair needs exhibited greater variability over time. This variability was not patterned by place, though–for example, whether any litter was visible changed in about a quarter of locations–but overall prevalence at the city level remained nearly constant. The clearest consistent change over time was the expansion of bicycle infrastructure, which increased across all four cities in line with national investments in cycling facilities and bike-sharing programs.

Dr. Mooney emphasized that: “generally speaking, not much changed over 10 years. This is important because, as scientists, we want to measure conditions at the time people experience them. For folks using Street View to assess neighborhood conditions, it feels like a challenge that we can't control when Google drove the car past a location. But it turns out that for a lot of measures in a lot of places, it really doesn't matter much.” In other words, while dramatic neighborhood transformations—such as those seen in Seattle’s South Lake Union over the past two decades—certainly occur, many residential areas, like Queen Anne or Beacon Hill, remain visually consistent, reducing the risk of bias in longitudinal studies not specifically targeting residents of areas undergoing rapid change..

The analysis also explored whether neighborhood income levels influenced the rate of change. While features of disinvestment were more common in lower-income areas at any single point in time, the pace of change was similar across income levels. This suggests that although inequality in neighborhood conditions persists, it does not manifest as differences in temporal stability of features.

What does all this mean for public health research? By demonstrating that most built environment features remain stable over a decade, the study provides reassurance that modest mismatches between imagery dates and research timelines are unlikely to compromise findings in neighborhoods not undergoing rapid redevelopment. This allows researchers to leverage GSV as a scalable and efficient tool for large epidemiological studies, linking neighborhood conditions to health outcomes such as diabetes incidence or physical activity.

At the same time, the study raises new questions. As Dr. Mooney reflects, “Is there a way to identify where changes have happened, so we can update our neighborhood measures only where it really matters?” This question points to the future direction of research: developing strategies to detect change hotspots, allowing resources to be focused where temporal mismatches could truly influence results. He also highlights the broader significance of linking Street View-based measures of disinvestment to health outcomes within the HCHS/SOL cohort. Such work will allow researchers to explore how environmental neglect—visible in vacant lots or poorly maintained buildings—may affect behaviors such as physical activity and contribute to chronic health conditions.

This study was supported by the National Institute of Environmental Health Sciences.

Mooney, S. J., Smith, C. M., Spalt, E. W., Piepmeier, L., Gassett, A. J., Gunning, G., Carlson, J. A., Evenson, K. R., Chambers, E. C., Daviglus, M., Lovasi, G. S., Gullón, P. T., Hirsch, J. A., Plascak, J. J., Rundle, A. G., Fry, D., Bader, M. D. M., Kau. (2025). Built Environment Change over Time Using Google Street View Assessments of Hispanic Community Health Study/Study of Latinos (HCHS/SOL) Cities. Journal of urban health : bulletin of the New York Academy of Medicine, 102(3), 670–679.