Journal of Environmental Biology

Patent landscape study on deep learning models for spatio-temporal air quality forecasting

S.P. Gandhale^1,2, T.V. Kale² and S.D. Pohekar¹*     

¹Symbiosis Centre for Research and Innovation, Symbiosis International (Deemed University), Pune- 412 115, India

²Department of Artificial Intelligence and Machine Learning, Symbiosis Institute of Technology, Symbiosis International (Deemed University), Pune-412 115, India

Received: 24 October 2025                   Revised: 24 February 2026                   Accepted: 28 February 2026

*Corresponding Author Email : sdpohekar@gmail.com                  *ORCiD: https://orcid.org/0000-0002-5846-6714

Abstract

Accurate air quality prediction is urgently required due to the harmful impacts of air pollution on human health, sustainability, and livability. Air pollutants’ nonlinear, dynamic, and spatiotemporal characteristics are hard to model using traditional statistical and physical models, particularly in complex urban areas. Therefore, deep learning-based methods have emerged as important ones, where LSTM models effectively modeled temporal dependencies and CNN models modeled spatial dispersion patterns. By learning both temporal and geographical correlations from large and diverse datasets, hybrid CNN-LSTM models have immensely improved the accuracy of predictions.

The complexity of the system has a risen due to the growing adoption of deeper architectures, data fusion techniques, adaptive learning strategies, and real-time system deployment platforms, resulting in a rise in patenting activities. There is a growing trend of using patents as a means to safeguard novel CNN-LSTM architectures, data pipelines, and end-to-end intelligent forecasting systems. Therefore, this paper will examine the patenting landscape of spatio-temporal deep learning models for air quality forecasting with a focus on innovation trends and the impact of system complexity on patent development.

Key words: Air quality forecasting, Air quality index, CNN-LSTM models, Patent landscape, Spatio-temporal prediction

Copyright © 2026 Triveni Enterprises. All rights reserved. No part of the Journal can be reproduced in any form without prior permission. Responsibility regarding the authenticity of the data, and the acceptability of the conclusions enforced or derived, rest completely with the author(s).