Date of Defense
13-4-2026 3:00 PM
Location
Microsoft Teams
Document Type
Thesis Defense
Degree Name
Master of Science in Internet of Things
College
CIT
Department
Computer and Network Engineering
First Advisor
Dr. Bassem Mokhtar
Keywords
Air quality monitoring, LSTM, ARIMA, hybrid forecasting, zonal approach, satellite remote sensing, Abu Dhabi, UAE.
Abstract
Air pollution is one of the most important and devastating environmental issues that heavily affects public health around the world and it’s the cause of approximately 4.2 million early deaths. This thesis focuses on further improving forecasting models by introducing a zonal approach and satellite-based spatial validation. The main objective is to explore a zonal approach with the ground station data and to add a spatial component using satellite imagery to improve the accuracy of the results. It follows a four-stage evolution framework while focusing on the three different zones chosen. The four stages introduced different aspects which include a long horizon baseline, seasonal hybridization, horizon reduction, then finally multivariate optimization. ARIMA, LSTM, ARIMA-LSTM, and SARIMA-LSTM were chosen as the main models with Hamdan Street, Liwa, and Musaffah as the three zones for this thesis. The study's results showed that no single model was universally best, it heavily relied on the zone and pollutant combination. For example, SARIMA-LSTM performed best for NO2 in the urban zone while standalone ARIMA performed best for PM10 in the industrial zone. With some pollutants, the complexity of the model did not matter which points to the importance of the zonal approach used here. When comparing satellite imagery to the ground station results, there was a high correlation between them. That showed the potential it had for combining it with conventional methods to get more accurate results in the future. The contribution of this study is the potential for improving early warning systems and policy decisions concerning air pollution by using 24 hours ahead forecasting. This study addresses the gaps in research around the world that focus on single urban centers or regions rather than specifically focusing on different activity zones and the gap of combining spatial approaches with conventional ground-based approaches.
Included in
A COMPARATIVE STUDY ON PERFORMANCE OF IOT-DRIVEN ML-ENABLED FORECASTING MODELS FOR EFFICIENT AIR QUALITY MONITORING
Microsoft Teams
Air pollution is one of the most important and devastating environmental issues that heavily affects public health around the world and it’s the cause of approximately 4.2 million early deaths. This thesis focuses on further improving forecasting models by introducing a zonal approach and satellite-based spatial validation. The main objective is to explore a zonal approach with the ground station data and to add a spatial component using satellite imagery to improve the accuracy of the results. It follows a four-stage evolution framework while focusing on the three different zones chosen. The four stages introduced different aspects which include a long horizon baseline, seasonal hybridization, horizon reduction, then finally multivariate optimization. ARIMA, LSTM, ARIMA-LSTM, and SARIMA-LSTM were chosen as the main models with Hamdan Street, Liwa, and Musaffah as the three zones for this thesis. The study's results showed that no single model was universally best, it heavily relied on the zone and pollutant combination. For example, SARIMA-LSTM performed best for NO2 in the urban zone while standalone ARIMA performed best for PM10 in the industrial zone. With some pollutants, the complexity of the model did not matter which points to the importance of the zonal approach used here. When comparing satellite imagery to the ground station results, there was a high correlation between them. That showed the potential it had for combining it with conventional methods to get more accurate results in the future. The contribution of this study is the potential for improving early warning systems and policy decisions concerning air pollution by using 24 hours ahead forecasting. This study addresses the gaps in research around the world that focus on single urban centers or regions rather than specifically focusing on different activity zones and the gap of combining spatial approaches with conventional ground-based approaches.