Research Projects

This project aims to analyze the chemical, geo-physical, spatio-temporal dynamics and emission inventory of atmospheric pollutants in urban areas using demography, socio-economic, and meteorological conditions. The goal is to develop a city-based 3-Dimensional real-time air pollutant propagation model using a component-based digital twin model and develop an Unmanned Aerial Vehicle and satellite observation-based 3-Dimensional pollutant propagation validation methodology. The proposed modeling framework uses five mechanisms: multimodal data assimilation and advanced machine learning algorithms for high resolution spatial estimation of pollutant concentrations; development of multivariate spatiotemporal pollutant concentration forecasting models based on ground monitoring sensor network locations; development of a 3-Dimensional vertical profiler of pollutants; and development of a Physics Informed knowledge enhancer for capturing pollutant transportation and dispersion based on wind speed and other underlying meteorological conditions. The project also involves a thorough investigation of historical pollutant propagation incidents and validating the model capacity. Real-time source contribution related investigations and identification of pollutant deposition-related factors and impacts of it on the local ecology are also discussed. Collaboration between partner institutions is planned for future collaboration and publication of key findings. Key questions include how to assimilate multimodal pollutant data from different sources and develop data-dependent digital-twin algorithms for tracking real-time 3-dimensional atmospheric pollutant trajectories for urban areas, how machine learning/deep learning algorithms can be used to capture gaseous propagation of atmospheric tropospheric pollutants on an urban scale or beyond, how to use modern multispectral-hyperspectral satellite observations for capturing high-resolution atmospheric pollutant propagation dynamics, how to establish connections between real-time atmospheric pollutant concentration provided by sensors and data-knowledge enriched digital twin-based atmospheric pollutant distribution models, and how to use Physics Informed Neural Networks to capture the physical properties of pollutants.