An AI-based Early Warning System for Urban Air-Pollution and Environmental Risk Monitoring

Main Article Content

Mohammed Abdul Khaled
Muhammad Nadeem
Ahmed Gamal Elsayed Hamed Ali Mansour
Ahmed Ali Mohammed Kulfod
Mohammed Abdul Aziz

Abstract

Air pollution is a long-standing environmental and public-health problem in urban areas, especially in fast-urbanising areas where the earliest possible detection of dangerous air-quality situations is a key to the efficient mitigation of risks. This paper is an analytical research exploration of the viability of an AI-supported early-warning system of the urban air-pollution and environmental hazards monitoring system with the aid of only secondary historical data. The paper fails to construct new predictive models and fails to provide an operational monitoring system, rather, it assesses the applicability of the previously reported established artificial-intelligence-based analytical procedures to publicly accessible datasets as an aid to early warning generation. Air-quality data observed at various urban monitoring points in 2019– 2023 on a hourly basis was analysed through a tailored working process that included a pre-processing phase, quality control, exploratory temporal space inspection, analysis of short-term trends, identification of pollution events, ADR-assisted trend and anomaly validation, and environmental risk classification services. Back-testing of history was then done to investigate the timeness and stability of the generation of warnings before the observed pollution peaks. The findings show that there are distinct seasonal and diurnal cycles in the quality of the urban air, and particulate matter has become the leading cause of serious pollution events. With the help of analytical workflows provided with AI, 87.6% of the detected pollution events had early warnings in the form of their highest concentrations, with an average lead time of about 9.8 hours. The risk classification also suggests that the occurrence of the high-risk and severe-risk conditions is the highest during winter months and the prolonged periods of pollution. The results indicate that useful earlywarning functionality is possible with analytical application of existing artificial intelligence solutions, and through the use of existing secondary data that are available publicly, and do not rely on building new models, or deploying real-time systems. According to the proposed analytical workflow, the proposed work has a viable and scalable basis to support future urban air-pollution early-warning efforts and risk management of the environment within the current monitoring infrastructures.

Article Details

Section
Articles