The emergence of urban heat traps and human mobility in 20 US cities

Urban heat (UH) is a major environmental hazard with impacts on mortality, morbidity, and energy consumption. While UH has been widely studied, most prior work relies on index-based, location-specific measures and pays limited attention to human network dynamics that can extend or alleviate exposure beyond static hotspots. Human mobility shapes urban socio-spatial networks and can expand the reach of environmental hazard exposures, as shown previously for air pollution. In the context of UH, existing mobility-aware indices (e.g., dynamic thermal exposure) capture exposure levels but do not characterize emergent properties arising from the intersection of mobility patterns and the spatial distribution of heat. This study defines three emergent properties at that intersection: heat traps (populations residing in high-UH areas visiting other high-UH areas), heat escapes (populations in high-UH areas visiting low-UH areas), and heat escalates (populations in low-UH areas visiting high-UH areas). The research questions are: to what extent does human mobility exacerbate UH exposure (prominence of heat traps), alleviate exposure (heat escapes), or expand the reach of exposure (heat escalates)? Using aggregated and anonymized mobility data, we construct origin–destination networks at the census tract level for 20 U.S. metropolitan areas and examine trip proportions among high and low UH areas to assess cross-city similarities and differences. SUHI (Surface Urban Heat Island) derived from LST is used to represent relative heat intensity, with recognition of limitations in its correspondence to ambient air temperature. Annual mean SUHI across urbanized areas is employed to capture citywide heat patterns for comparison with mobility.

Prior studies identify multiple drivers and correlates of urban heat, including tree canopy and vegetation reducing UH, transportation-related effects, and population density changes. Mobility datasets have been widely used to study responses and exposures for hurricanes, floods, and infectious diseases, demonstrating that movement patterns influence hazard exposure. However, the literature insufficiently examines mobility–UH relationships. A mobility-aware UH index (DTEx) has been proposed, but it does not capture emergent properties like traps, escapes, and escalates. Remote sensing advances provide SUHI/LST data enabling tract-level UH characterization, though SUHI relates to surface rather than air temperature. This study builds on these strands by explicitly integrating human mobility networks with SUHI to reveal emergent exposure patterns.

Study scope and period: Human mobility was analyzed for February 2020 (pre-COVID-19 disruptions) across 20 U.S. metropolitan areas. Heat exposure metric: Annual mean Surface Urban Heat Island (SUHI) values were obtained from the United States Surface Urban Heat Island database, which synthesizes MODIS and terrain data for census tracts within urbanized areas. SUHI values were split via quantile breaks into low, median, and high UH clusters; analyses focused on low and high UH tracts. Mobility data: Aggregated, privacy-protected location data from Spectus provided anonymized device-level positions and dwell times. Home tracts were inferred based on dwell time patterns; visits aggregated by unique device IDs yielded origin (home tract) to destination (visitation tract) counts. Data cover intra-county trips only. Mobility network construction: For each metro area, a directed network was formed where nodes are census tracts and weighted edges represent monthly visit counts from home tracts to visitation tracts. Ratios for emergent properties: For each home tract i, three ratios were defined by dividing trips in each category by total trips from tract i: (1) heat escalates (low→high): R_low→high = trips from low-UH i to high-UH j / total trips from i; (2) heat escapes (high→low): R_high→low = trips from high-UH i to low-UH j / total trips from i; (3) heat traps (high→high): R_high→high = trips from high-UH i to high-UH j / total trips from i. City classification: For each metro, percentages of trips in each category were aggregated and used to classify cities. If more than half of trips were trap, escalate, or escape type, the city was classified accordingly (some cities exhibit multiple classifications). Visualization: UH categories were mapped using quantile breaks; Python Matplotlib produced maps illustrating SUHI and mobility ratios.

- Urban heat traps are prevalent: Most metropolitan areas display strong high→high mobility, indicating populations in high-UH tracts predominantly visit other high-UH tracts. - Pronounced traps in Los Angeles and Chicago: Los Angeles has 52% of tracts in high UH, and 81% of high-UH tracts exhibit high→high trips (ratios up to 0.92). Chicago has 49% of tracts in high UH, with 78% of high-UH tracts showing high→high trips (ratios up to 0.91). Traps in Chicago are more spatially clustered than in Los Angeles. - Lower trap prevalence in Boston and Atlanta: Boston has 36% high-UH tracts, with 37% of high-UH tracts showing high→high trips and very low high→low (escape) ratios; Atlanta has 21% high-UH tracts, with 40% of high-UH tracts showing high→high trips and some escapes (high→low up to 0.12). Despite lower trap intensity, Table 2 classifies Boston and Atlanta as trap cities based on aggregate thresholds. - High heat escapes identified: Minneapolis shows strong escapes, with 56% of high-UH tracts having high→low trips (ratios up to 0.13) and 45% of tracts in high UH. Dallas similarly shows notable escapes, with 49% of high-UH tracts having high→low trips (ratios up to 0.17) and 50% of tracts in high UH. - Heat escalates observed: Escalation (low→high) is present in several metros; Los Angeles exhibits notable low→high visitation ratios in northern tracts (0.22–0.35). Table 2 lists multiple metros as escalate & trap. - Cross-city classification (Table 2 highlights): Many metros are classified as trap or escalate & trap (e.g., Los Angeles, Chicago, Houston, Detroit, Philadelphia). Minneapolis and Dallas are tagged with escape (Minneapolis: escape; Dallas: trap & escape). DC and Phoenix show extreme values in Table 2 (e.g., DC low→high 1.0 and high→high 1.0; Phoenix low→high 0.95 and high→high 1.0). - Emergent nature of traps: Supplementary analysis found no statistically significant association between heat traps and demographic segregation or facility centrality, suggesting traps arise from mobility-behavioral dynamics rather than simple structural or demographic attributes.

Findings demonstrate that the intersection of human mobility and spatial UH distribution yields emergent exposure patterns that can exacerbate (traps), alleviate (escapes), or expand (escalates) heat exposure. In many cities, high-UH residents predominantly visit other high-UH areas, reinforcing exposure. Some cities, notably Minneapolis and Dallas, show greater movement from hot to cooler tracts, indicating potential for mobility-enabled risk reduction. Heat escalation patterns reveal that residents of cooler tracts can increase exposure by visiting hotter zones, underscoring that risk is not confined to residence location. The lack of significant correlation between traps and demographic segregation or facility distribution suggests emergent, behavior-driven dynamics. For planning, these metrics offer actionable insights: prioritize greening and cooling interventions in frequently visited high-UH hotspots, improve access and transit options to cooler areas, and target mobility-influenced escalated-risk zones. Recognizing mobility-induced heat risks is vital for designing adaptive, equitable urban environments.

This study advances understanding of urban heat exposure by integrating large-scale mobility networks with tract-level SUHI to identify three emergent properties—heat traps, escapes, and escalates—across 20 U.S. metropolitan areas. Results show heat traps are widespread, with pronounced cases in Los Angeles and Chicago; escapes are notable in Minneapolis and Dallas; and escalation occurs in several metros. The framework provides planners with metrics to diagnose mobility-mediated heat risks and to guide targeted interventions in high-visit hotspots and along common travel corridors. Future research should incorporate cross-county mobility, temporal patterns of visitation (duration and time of day), seasonal analyses (including winter vs. summer SUHI/LST and ambient air temperature), indoor exposure considerations (e.g., air conditioning availability), and refined heat metrics that better capture human thermal experience.

- Data representativeness: Mobility data are based on smartphone users who opt in to location sharing, underrepresenting populations without smartphones or those who opt out (e.g., some children, elderly, lower-income groups). Weighting/stratification could improve representativeness. - Temporal detail: Destination dwell times and visit purposes are not captured, potentially affecting interpretation of exposure contexts. - Indoor vs. outdoor exposure: Mobility to hotter tracts may not translate to outdoor exposure if activities occur in air-conditioned environments; AC availability varies across tracts. - SUHI vs. air temperature: SUHI reflects surface temperature and may diverge from ambient air temperature, especially daytime; results based on SUHI may not directly equate to human thermal stress. - Seasonal context: Mobility data are from February 2020; winter conditions in colder cities (e.g., Chicago) might influence interpretation of higher SUHI values. - Spatial coverage of mobility: Dataset excludes cross-county trips, limiting analysis to intra-county mobility. - Classification thresholds: City-level classifications based on majority trip type may oversimplify nuanced intra-city patterns.