Introduction
Urban planning has traditionally relied on quantity-based metrics to assess infrastructure provision. This "more is better" philosophy assumes that increasing the density of roads, parks, or transit facilities linearly improves urban life. However, this approach often overlooks the negative externalities associated with infrastructure, such as noise, air pollution, and the urban heat island effect. This paper proposes a shift toward Infrastructure Quality (IQ), a metric that accounts for the non-linear impacts of urban features on environmental hazards.
Methodology
The researchers developed a data-driven framework to calculate IQ provision scores across five major U.S. cities: Houston, Dallas, Los Angeles, Chicago, and Detroit. The methodology integrates machine learning with interpretability tools to identify optimal infrastructure thresholds.
- Hazard Classification: An XGBoost algorithm classified census tracts into high or low environmental hazard categories based on air pollution (PM2.5) and extreme heat, using infrastructure features (e.g., road density, walkability, POI density) as predictors.
- Interpretability with SHAP: To move beyond "black box" predictions, SHapley Additive Explanations (SHAP) were used to quantify the objective weight of each feature.
- Threshold Discovery: By analyzing SHAP dependence plots, the study identified "inflection points"—specific thresholds where adding more infrastructure (like railways or high-volume roads) transitions from a benefit to a significant driver of environmental hazards.
- The IQ Metric: The final IQ score measures the proximity of a neighborhood's infrastructure levels to these optimal thresholds, rewarding provision that avoids crossing into "hazard zones."
Key Results
The Quantity-Quality Paradox
The study reveals a stark paradox: neighborhoods with the highest quantity of infrastructure often suffer from the lowest quality. These areas are frequently over-saturated with features that exacerbate heat and pollution, effectively "expiring" their utility. For instance, in Houston, railway coverage beyond 10.68% was found to sharply increase environmental hazard risks.
Infrastructure-Specific Dynamics
Not all features behave the same way. While roads and railways exhibit clear thresholds, features like parks and newer housing showed "uncapped" utility, where increased provision consistently correlated with lower hazards. Furthermore, the importance of these features is highly city-specific; rail infrastructure is a primary hazard predictor in Chicago but is less significant in Dallas or Detroit, highlighting the failure of "one-size-fits-all" urban indices.
Socioeconomic Inequality
The IQ metric exposes deep-seated inequalities. Across all five cities, high-IQ census tracts were strongly correlated with higher median incomes. In Los Angeles, residents in high-quality infrastructure areas earned up to 27% more than those in low-quality areas. Traditional quantity-based metrics often mischaracterize low-income areas as "well-served" because they ignore the environmental burdens of the infrastructure present.
Conclusion
By moving beyond simple counts of urban assets, this research provides a more nuanced understanding of environmental justice. The findings suggest that urban planning must transition from maximizing quantity to optimizing quality through threshold-aware development. Future work should expand this framework to include utilities like water and digital infrastructure to provide a truly comprehensive view of urban equity.
Summary updated