Smartphone-based Crowd-Sourced Geohazard Mapping and Early Warning System

Thesis Abstract

The rapid expansion of mobile technologies and social sensing presents an opportunity to democratize geohazard monitoring, yet urban and peri-urban communities remain vulnerable when authoritative monitoring is sparse or slow to disseminate warnings. This study addresses the need for timely, scalable, and accessible geohazard data collection and early warning dissemination by operationalizing a smartphone-based crowd-sourced platform that integrates user-generated observations with real-time sensor feeds. The aim is to design, implement, and evaluate a crowdsourced geohazard mapping and early warning system that enhances situational awareness, community resilience, and official hazard response. Specific objectives are to (1) develop a cross-platform mobile application enabling geo-tagged hazard reports, multimedia evidence, and user reliability indicators; (2) fuse citizen reports with remotely sensed and in-situ sensor data to produce probabilistic risk maps for landslides, floods, and soil liquefaction; (3) implement an automated alerting mechanism aligned with local authorities’ decision thresholds; (4) assess data quality, user engagement, and perceived usefulness through mixed methods; and (5) evaluate system performance under simulated and historical hazard scenarios to derive practical deployment guidelines. A mixed-methods research design is employed, combining a development study with an evaluative field trial in a seismic- and flood-prone metropolitan region. The population includes 1) city residents aged 18–65 who use smartphones, 2) volunteers recruited via community organizations (target n = 300) for crowdsourced data contributions, and 3) municipal disaster management officials (n = 20) for system integration feedback. Data collection instruments comprise (i) a custom Android/iOS app capturing geolocated hazard reports, photos, and audio notes, plus an integrity scoring module; (ii) a web dashboard aggregating crowd-sourced data with satellite-derived precipitation and LiDAR-derived slope metrics; (iii) semi-structured interviews and focus group discussions with users and officials; and (iv) a series of simulated hazard events to test alert timeliness and accuracy. Validity and reliability are ensured through pilot testing (n = 50 users), triangulation across data streams, test-retest reliability for survey instruments (Cronbach’s alpha ? 0.80), and inter-rater reliability for qualitative coding (Cohen’s kappa ? 0.75). Data analysis employs (i) spatial-temporal data fusion using Bayesian hierarchical modeling to produce probabilistic hazard maps, (ii) regression analysis to identify predictors of user engagement and report accuracy, (iii) signal detection theory to evaluate the balance between false alarms and missed detections, (iv) time-series analysis for alert latency, and (v) thematic analysis for qualitative data guided by the Social Amplification of Risk framework. Theoretical grounding draws on the Technology Acceptance Model (TAM) to interpret adoption, and the Community-Based Disaster Risk Management (CBDRM) paradigm to frame local empowerment and resilience outcomes. Expected findings indicate that the integration of crowd-sourced reports with sensor and remote-sensing data yields improved spatial precision and faster hazard detection than conventional monitoring alone, with user-contributed data achieving 70–85% confirmation rates when corroborated by sensor feeds. The probabilistic hazard maps are anticipated to deliver timely alerts within 3–7 minutes of major triggering signals, reducing decision latency for authorities by up to 40%. Key determinants of sustained user participation are projected to include perceived usefulness, ease of use, and trust in data quality, moderated by demographic and geographic factors. The study is expected to reveal systematic variances in report reliability by terrain type (steep slopes vs. flat basins) and by time of day, necessitating adaptive weighting schemes in the fusion model. Contributions to knowledge include (i) a scalable, participatory geohazard monitoring framework that formalizes data quality control for citizen-generated reports; (ii) a robust data fusion architecture combining crowdsourcing with multi-source sensing under a Bayesian framework; (iii) empirically derived guidelines for deploying smartphone-based warning systems within CBDRM contexts; and (iv) empirical validation of a social-technical model linking technology acceptance to hazard governance outcomes. The study concludes that smartphone-based crowd-sourced geohazard mapping can meaningfully augment official monitoring and early warning, provided that data quality mechanisms, clear alert protocols, and sustained community engagement are in place. Recommendations emphasize integrating the platform with municipal information systems, conducting ongoing user training, refining reliability scoring algorithms, and expanding to multi-hazard scenarios with governance interoperability across agencies.