Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Spatiotemporal Air Quality Mapping in Urban Areas Using Sparse Sensor Data, Satellite Imagery, Meteorological Factors, and Spatial Features

Osama Ahmad, Zubair Khalid, Muhammad Tahir, Momin Uppal

TL;DR

This work presents a Graph Neural Network–based framework for high-resolution spatiotemporal AQI mapping in urban areas using sparse sensor data, satellite-derived AOD, meteorological variables, and various urban features. It combines temporal CNN streams with spatial GCN processing on a graph built from a 1 km grid, with multi-resolution inputs and a fusion mechanism that yields AQI estimates at unmonitored locations via a semi-supervised learning objective. The Lahore case study demonstrates the feasibility of integrating diverse data sources (PM2.5, AOD, ERA5, POIs, population, UGS, and road networks) to generate detailed AQI maps and capture urban dynamics like rush-hour pollution. Key findings indicate that daily-resolution meteorological features and multi-resolution temporal data significantly boost accuracy, while some spatial features have varying influence due to distribution sparsity; the approach provides a practical tool for policymakers to assess pollutant distributions and target interventions.

Abstract

Monitoring air pollution is crucial for protecting human health from exposure to harmful substances. Traditional methods of air quality monitoring, such as ground-based sensors and satellite-based remote sensing, face limitations due to high deployment costs, sparse sensor coverage, and environmental interferences. To address these challenges, this paper proposes a framework for high-resolution spatiotemporal Air Quality Index (AQI) mapping using sparse sensor data, satellite imagery, and various spatiotemporal factors. By leveraging Graph Neural Networks (GNNs), we estimate AQI values at unmonitored locations based on both spatial and temporal dependencies. The framework incorporates a wide range of environmental features, including meteorological data, road networks, points of interest (PoIs), population density, and urban green spaces, which enhance prediction accuracy. We illustrate the use of our approach through a case study in Lahore, Pakistan, where multi-resolution data is used to generate the air quality index map at a fine spatiotemporal scale.

Spatiotemporal Air Quality Mapping in Urban Areas Using Sparse Sensor Data, Satellite Imagery, Meteorological Factors, and Spatial Features

TL;DR

This work presents a Graph Neural Network–based framework for high-resolution spatiotemporal AQI mapping in urban areas using sparse sensor data, satellite-derived AOD, meteorological variables, and various urban features. It combines temporal CNN streams with spatial GCN processing on a graph built from a 1 km grid, with multi-resolution inputs and a fusion mechanism that yields AQI estimates at unmonitored locations via a semi-supervised learning objective. The Lahore case study demonstrates the feasibility of integrating diverse data sources (PM2.5, AOD, ERA5, POIs, population, UGS, and road networks) to generate detailed AQI maps and capture urban dynamics like rush-hour pollution. Key findings indicate that daily-resolution meteorological features and multi-resolution temporal data significantly boost accuracy, while some spatial features have varying influence due to distribution sparsity; the approach provides a practical tool for policymakers to assess pollutant distributions and target interventions.

Abstract

Monitoring air pollution is crucial for protecting human health from exposure to harmful substances. Traditional methods of air quality monitoring, such as ground-based sensors and satellite-based remote sensing, face limitations due to high deployment costs, sparse sensor coverage, and environmental interferences. To address these challenges, this paper proposes a framework for high-resolution spatiotemporal Air Quality Index (AQI) mapping using sparse sensor data, satellite imagery, and various spatiotemporal factors. By leveraging Graph Neural Networks (GNNs), we estimate AQI values at unmonitored locations based on both spatial and temporal dependencies. The framework incorporates a wide range of environmental features, including meteorological data, road networks, points of interest (PoIs), population density, and urban green spaces, which enhance prediction accuracy. We illustrate the use of our approach through a case study in Lahore, Pakistan, where multi-resolution data is used to generate the air quality index map at a fine spatiotemporal scale.
Paper Structure (15 sections, 7 equations, 4 figures, 2 tables)

This paper contains 15 sections, 7 equations, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Proposed Methodology
  3. Problem Formulation
  4. System Model
  5. Learning Framework
  6. Implementation
  7. Data Acquisition
  8. Particulate Matter Data
  9. Meteorological Data
  10. Road Networks
  11. Point-of-Interests (POIs)
  12. Population Count
  13. Urban Green Spaces (UGS)
  14. Results
  15. Conclusions

Figures (4)

  • Figure 1: Block diagram of the proposed learning network. Temporal features pass through CNNs, producing $\mathbf{H}^{(0)}_k$, while spatial features are processed using GCNs, producing $\mathbf{H}_k$.
  • Figure 2: Distribution of 30 AQI Sensors in Lahore.
  • Figure 3: AQI spatiotemporal map for the Lahore region. For optimal visualization, we recommend using Adobe Reader.
  • Figure 4: Comparison of ground truth and predicted AQI values for sensors at nodes (a) 657 and (b) 1435 from December 19 to 31, 2024, at an hourly resolution. The shaded regions represent rush hours: 7:00–10:00 (orange) and 17:00–20:00 (pink).