School of Science Department of Mathematics 50 Retrieving of Urban Morphology from Satellite Images and Machine Learning Supervisor: FUNG Jimmy Chi Hung / MATH Student: YANG Yuang / DSCT Course: UROP1100, Summer Urban morphology is a study that has a valuable application on meteorological modeling, carbon emission estimation, city environment studies, and city planning. Among all urban surface composition information, the information of buildings in the city is especially important. This project aims to develop a machine learning approach to retrieve building locations and dimensions from satellite images, which would significantly reduce the cost of gaining urban building data and facilitate environmental studies worldwide. In the first phase of this project, a model based on FCN-restnet101 is used to handle the semantic image segmentation task to exploit satellite images to generate a bitmap that implies the locations of buildings. After a short session of preliminary training, the model achieves an average overall precision of 80%. Data processing and techniques to prevent model outputting the same value are also discussed. Use Machine Learning Technique to Retrieve Air Quality Data of SO2 Supervisor: FUNG Jimmy Chi Hung / MATH Co-supervisor: LU Xingcheng / ENVR Student: WU Zhiang / COSC Course: UROP1100, Fall Sulphur dioxide (SO2) is one of the most important factors of air quality, which takes part in some chemical reactions in the atmosphere. Prediction based on past observation is on the topic of environmental science. With the development of machine learning in computer science, more and more deep learning techniques are applied in this area. In this project, we tried to use an ensemble BP neural network to predict SO2. Many experiments are conducted to find a reasonable and good model with the data provided by some scientific institutions. Use Machine Learning Technique to Retrieve Air Quality Data of SO2 Supervisor: FUNG Jimmy Chi Hung / MATH Co-supervisor: LU Xingcheng / ENVR Student: YI Sophia / COMP Course: UROP1100, Fall Sulfate and nitrate constitute a significant portion of air pollutants emitted into the atmosphere. Source apportionments of have shown that among many factors including industrial emissions, biomass burning, and residual combustion, secondary sulfate and secondary nitrate were the most abundant sources of PM2.5(Huang et al., 2018), contributing to 33% and 23% of PM2.5 mass annually (Wang et al., 2019). Therefore, monitoring the concentration of sulfate and nitrate in the air is crucial for the management. This project builds up a backpropagation model to estimate and in 2005 and 2015 based on 13 parameters including meteorological variables and pollutants variables. By training these datasets, the model is able to reflect secondary sulphate and secondary nitrate over a particular region.