Smartphone-based crowdsourced urban heat monitoring and mapping

Thesis Abstract

Urban heat risk and thermal inequity are amplified in rapidly urbanizing cities, where conventional meteorological networks are sparse and fail to capture microclimatic variability experienced by residents. This study develops and implements a smartphone-based crowdsourced framework to monitor and map urban heat using user-contributed surface temperature estimates and ambient conditions, coupled with device-derived metadata, to produce high-resolution thermal maps and exposure indicators. The aim is to (i) assess the feasibility and reliability of crowdsourced temperature data for urban heat assessment, (ii) generate fine-grained urban heat maps linked to socio-spatial factors, and (iii) evaluate the potential of such crowdsourced data to inform heat mitigation planning and public health interventions. Specific objectives include (a) designing a low-burden mobile app module for standardized temperature reporting and geotagging, (b) calibrating smartphone-derived temperature estimates against a network of fixed sensors and portable thermography measurements, (c) developing a data fusion workflow that integrates crowdsourced observations with satellite-derived land surface temperature and crowd context variables (time of day, occupancy, shading, and surface materials), (d) applying spatial interpolation and machine learning techniques to produce high-resolution urban heat surfaces, and (e) assessing associations between heat exposure, demographics, land use, and built-environment features. The methodology combines a convergent mixed-methods design with a smartphone-based crowdsourcing platform deployed across 12 districts of a mid-sized metropolitan area. The population includes adult residents and workers who consent to contribute voluntary temperature reports over a 12-month period. A purposive sample of 200 participants will be recruited to ensure diversity in age, housing type, and neighborhood characteristics. Data collection instruments comprise a cross-platform mobile application for recording surface and ambient temperatures (with standardized protocols and calibration prompts), a lightweight questionnaire capturing demographic and behavioral context, and a portable reference thermal device for validation in 50 scheduled field sessions. Fixed-sensor data from the city’s urban climate network (n ? 30 stations) and high-resolution Landsat-derived land surface temperature (LST) data will be integrated. Data analysis employs a multi-step approach (i) quality control and calibration of smartphone readings using linear regression models and Bland-Altman analyses to quantify agreement with reference measurements, (ii) spatial interpolation via ordinary kriging and gradient boosting regression to generate 1–10 m resolution heat maps, (iii) data fusion through a Bayesian hierarchical model that combines crowd data, fixed sensors, and satellite data while accounting for spatial autocorrelation and measurement error, (iv) clustering analysis to identify heat-island typologies, and (v) regression analyses (multilevel and geographically weighted regression) to examine links between heat exposure and socio-spatial variables. The analytical framework draws on the theoretical underpinnings of urban climate resilience and crowdsourcing reliability, with relevant theories including the Theory of Planned Behavior to interpret participant engagement and the Urban Climate Governance framework to interpret policy implications. Expected findings indicate that smartphone-based crowdsourced data, when properly calibrated, can capture sub-district thermal heterogeneity with acceptable bias (mean absolute error ? 1.5°C relative to reference sensors) and reveal hot spots associated with low-albedo surfaces, lack of tree canopy, and intensively built-up blocks. The fusion model is anticipated to reduce uncertainty by 25–40% compared to satellite-only or sensor-only approaches, producing actionable heat maps at 5–10 m resolution. Analyses are expected to show significant associations between high heat exposure and vulnerable populations, including lower-income households and elderly residents, as well as stronger heat signals in commercial districts with low greenery. The study anticipates identifying neighborhood-scale heat-island typologies and critical microclimates that respond differently to shading, wind corridors, and material properties. The study contributes to knowledge by demonstrating a scalable, low-cost crowdsourced methodology for urban heat monitoring that complements traditional networks, advancing methods in spatial data fusion, and providing empirical evidence on microclimatic drivers of heat exposure and inequity. It informs urban planning and public health strategies by delivering high-resolution, temporally enhanced heat maps and exposure indicators that can guide targeted greening, reflective surface interventions, and heat-health outreach. The main conclusion supports the viability of smartphone crowdsourcing as a complementary data stream for urban climate monitoring and policy-making, with recommendations emphasizing standardized calibration protocols, sustained participant engagement strategies, integration with municipal planning processes, and continued evaluation of data quality and privacy protections.