What question did this study set out to answer?

This research aims to develop a forecasting framework for urban PM2.5 levels without relying on traditional meteorological data.

June 17, 2026Applied SciencesOpen Access

STEA-Net: An Endogenous Multi-Pollutant-Driven Spatio-Temporal Framework for Urban PM2.5 Forecasting

View Full Paper

Ask AI

Bookmark

Authors

Randomized trial demonstrates enhanced PM2.5 forecasting in urban areas, suggesting a meteorology-free approach for better public health management.

This research aims to develop a forecasting framework for urban PM2.5 levels without relying on traditional meteorological data.
Introduced STEA-Net, which utilizes a hybrid adjacency matrix for mapping pollutant transport pathways.
Employed a Graph Attention Network and Bidirectional LSTM to analyze spatial and temporal dynamics.
Validated model using a multi-year dataset from India (Jan 2015–Feb 2020).
Achieved a peak Coefficient of Determination (R2) of 0.9294 with a peak RMSE of 14.38 µg/m3.
Outperformed traditional machine learning models and was competitive with advanced deep learning approaches.
Demonstrated architectural stability with a verified deviation of ±2.22 µg/m3.

Kaur et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark

Randomized trial demonstrates enhanced PM2.5 forecasting in urban areas, suggesting a meteorology-free approach for better public health management.