1.1. Research Background

Extreme heat weather is one of the deadliest environmental hazards in the United States, the heat extreme heat events have significant negative impacts on urban public health and urban sustainable development. In dense, urban metropolitan cities like New York, extreme heat interacts with not just the built environment and local infrastructure conditions, but also the socioeconomic climate—extreme heat acts as one of the many architects shaping where service disruptions, complaints, and other surrounding stressors occur, degrading quality-of-life (QoL) for urban residents.

Within the context of this degradation is New York City’s 311 service, a complaint system that accepts reports via calls, emails, and website submissions that can reflect heat induced QoL behavior; it is a granular, real-time lens that provides understanding of how heat-related aggravation and other aspects can translate into observable, negative resident sentiment. This project in particular seeks to connect extreme heat versus normal heat weeks with environmental factors, socioeconomic conditions, and urban morphology to potentially explain QoL issues by different factors and their different performance during hotter periods.

In this regard, the project is built upon geospatial data science techniques for modeling weekly QoL outcomes in extreme heat and normal heat conditions, as proxied by selected 311 report categories, using ordinary least squares (OLS) regression modeling as a baseline followed by modern machine learning (ML) models, and SHapley Additive exPlanations (SHAP) method to interpret.

1.2. Research Gap

A substantial body of literature has established the correlation between rising temperatures and increased frequency of 311 service requests, specifically regarding noise, energy, and water consumption (Harlan et al., 2006; Hsu et al., 2021). And some other urban research also identified how different socioeconomic, environmental, and urban built metrics patterns shape the spatial heterogeneity of 311 calls (Uejio et al., 2010). However, fewer studies connect them and investigate how the impact of these factors on QoF performance shifts when the thermal environment crosses into extreme thresholds.

Transitional approaches like OLS in 311 analysis generally struggle to capture the non-linear behaviors of human-environment interactions. While machine learning offers improved predictive power, (Kontokosta & Tull, 2017) it generally lacks interpretability. So, by integrating SHAP approach to compare extreme versus normal heat weeks, this study addresses a critical gap, where it takes a step further from simple prediction to interpret and explain socioeconomic, environmental, and urban built drivers under two different heat regimes, of which is explained in detail in section 2. (Lundberg & Lee, 2017)

1.3. Research Objective

With the research gap's context, this study asks: how do environmental, socioeconomic, and urban morphology factors influence the QoL in New York City, defined as QoL-related 311 report rate per capita, during extreme heat weeks versus normal heat weeks?

Heat-related academic literature suggests that discomfort rises with temperature, so it is hypothesized that the QoL rate per capita will align with those findings. However, the objective of this research is to produce SHAP model values that can help reveal the drivers of QoL complaints in New York City.