“Listening to the City, Seeing the Planet”: Integrating Environmental Complaints and AlphaEarth for Multi-Scale Visual Analytics
Authors
Yunchao Wang (zhejiang university of technology), Jinghui Chu (Zhejiang University of Technology), Kaijun Yang (Zhejiang University of Technology), zhe chen (Zhejiang University of Technology), Gefei Zhang (Zhejiang University of Technology), Guodao Sun (Zhejiang University of Technology), Ronghua Liang (Zhejiang University of Technology)
Presentation
- Session
- Story time
- Time
- Thursday, Nov 12, 09:12 – 09:24 (US/Eastern) · session 08:00 – 09:30
- Location
- Hall America north
Keywords
Visual analysis, environmental complaints, spatial analysis, urban governance
Abstract
Citizen environmental complaints provide a valuable, citizen-centered perspective on urban environmental conditions, complementing traditional sensor-based monitoring systems. However, extracting actionable insights from such large-scale, heterogeneous data remains challenging due to the entanglement of episodic events and persistent anomalies, as well as the highly diverse spatial propagation patterns across different pollution types. Moreover, effective multi-scale visual analytics tools that can jointly integrate subjective public feedback with objective urban physical context are still lacking, limiting comprehensive cross-level understanding and interpretation. In this paper, we present EnvLens, a visual analytics system that unifies geo-referenced citizen complaints with satellite-derived surface semantic embeddings within H3 cells. To capture the nuanced spatial dynamics of disturbances, we introduce a directional propagation analysis method that drives a data-driven anisotropic diffusion heatmap based on a von Mises parametric kernel. Based on Jensen–Shannon Divergence (JSD) and Mean Resultant Length (MRL), we propose a JSD-MRL spatial-behavior typology that systematically classifies complaint patterns into three behavioral types, population-following, structural-diffusive, and spatial-specific. EnvLens employs a dual-embedding comparison workflow to recommend and compare urban regions, revealing pairs with highly similar built environments but markedly different complaint profiles. Through case studies utilizing 42,755 environmental complaints from Shenzhen and domain expert feedback, we demonstrate that EnvLens successfully pinpoints type-specific spatial anomalies, interprets directional propagation footprints, and generates robust cross-regional hypotheses for urban governance.
For Practitioners
Practitioners Statement. This paper will be of interest to municipal environmental managers, urban-governance agencies, GIS and remote-sensing analysts, public-sector data scientists, and civic-technology teams who work with citizen-reported environmental problems. Practitioners can apply the EnvLens workflow to combine environmental complaints with satellite-derived urban context, identify complaint-type-specific spatial anomalies, examine directional propagation patterns, and compare physically similar areas with different complaint profiles. These capabilities can help prioritize field inspections and monitoring resources, identify locations requiring further investigation, and generate cross-district hypotheses about possible sources, infrastructure, or reporting differences. The approach is intended to support exploratory analysis and decision-making rather than replace direct environmental measurements or establish causal relationships; practitioners should therefore validate detected patterns with sensor observations, facility records, and meteorological data.