A Transformer-Based Multimodal Learning Framework for Spatiotemporal Change Detection and Urban Air Quality Estimation Using Satellite Imagery, Ground Sensors, and Traffic Data

Read full text Summarise with VIRAT
Authors: M.Thangavel, A.M. Kalpana, S.Ananth
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This study aims to address the pressing challenge of environmental monitoring in rapidlyexpanding urban areas by developing a reliable system for spatiotemporal land-use changedetection and air quality estimation, both of which are essential for sustainable urban planning anddevelopment. A transformer-based multimodal learning framework is proposed, which jointlyprocesses satellite imagery, ground-level sensor readings, and traffic data. Leveraging the selfattention mechanism of Transformers, the model captures and integrates complex spatial andtemporal dependencies across these heterogeneous data sources within a unified predictivearchitecture. Experimental results demonstrate that the proposed model achieves state-of-the-artperformance in both land-use change detection and pollutant level prediction (e.g., PM2.5, NO₂),outperforming a wide range of traditional and deep learning baselines. The framework also showsstrong generalization capability, maintaining high performance when applied to previously unseenurban regions, as confirmed through an urban transferability study. This work is the first tocombine satellite, sensor, and traffic data within a transformer-based multimodal framework forenvironmental monitoring. Its ability to scale and generalize across urban settings, along with itsinterpretable modelling of spatiotemporal dependencies, makes it a promising tool for enhancingdata-driven smart city initiatives and sustainable development efforts.

Fields of Study & Subjects

Air Quality Monitoring and Forecasting Remote-Sensing Image Classification Environmental Quality and Pollution

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
1 January 2026
Language
EN