Integrating deep learning and machine learning methods to track air pollution amounts and impacts on vegetation cover, using remote sensing data

Keywords: Deep learning, machine Learning, remote sensing, vegetation degradation, air pollution
Abstract Type: Paper Abstract

Abstract

Using machine learning and deep learning approaches, this article seeks to monitor and forecast vegetation degradation brought on by air pollution in the metropolis of Dhaka, Bangladesh. This region has experienced an above average increase in air pollution and so discussion and analysis of related degradation of vegetation, related to this air pollution, is needed. Consequently, it is crucial to monitor and forecast vegetation health and change, especially degradation, across this landscape. This trajectory of vegetation change and degradation was monitored using eight deep learning segmentation models (Unet, Unet++, Manet, Linknet, FPN, PSPNet, DeepLabV3, DeepLabV3+). Seven machine learning methods were evaluated to link the air pollution zones to the trends of vegetation change and degradation, utilizing multiple algorithms, including Support Vector Machine (SVM), Extreme Gradient Boosting (XGB), K-Nearest Neighbor (KNN), Naive Bayesian (NB), Decision Tree (DT), Random Forest (RF), and Neural Network (NN). This research determined that there has been a decrease in dense vegetation close to higher air pollution zones. Additionally, according to research employing the machine learning technique, Dhaka may have some of the worst air quality in the world. It appears a significant loss of vegetation cover, as a result of these high air pollution levels, is underway. The use of an integrated machine learning and deep learning strategy as highlighted in this research, appears to be a practical and economical method for tracking the vegetation degradation and change, and in establishing the causal links to air pollution.