Academy of Interdisciplinary Studies Division of Environment and Sustainability 236 Improvement of the Air Quality Forecast by Using Deep-Learning Technique Supervisor: FUNG Jimmy Chi Hung / ENVR Student: PHAM Nguyen Nhat Minh / DSCT Course: UROP 1100, Fall UROP 2100, Spring This study developed an efficient framework to integrate ERA5/CAMS reanalysis data with ground-based PM2.5 measurements across China. A computational pipeline was designed to interpolate high-resolution meteorological and aerosol variables from ERA5/CAMS datasets (721×1440 grid) to the PM2.5 monitoring resolution (329×430 grid). Optimized bivariate spline interpolation techniques were employed with memoryefficient processing to handle the temporal matching of 72 hourly time steps from July 1–3, 2018. The resulting dataset combines 23 meteorological and aerosol variables with PM2.5 measurements, supporting advanced air quality modeling and machine learning applications. An XGBoost-based model was also developed to predict PM2.5 concentrations using the integrated dataset. Air Quality Measurement Techniques Supervisor: GU Dasa / ENVR Student: LEUNG Nga Man / EVMT Course: UROP 1000, Summer This report analyses the concentrations of common greenhouse gases and volatile organic compounds in two primary occupational spaces in a household: living room and bedroom. From July 15th to 20th 2025, 8 air samples were collected from the living room and bedroom in different conditions. After running laboratory tests, it is found that indoor carbon dioxide concentrations are higher than outdoor concentrations, while the concentrations of chosen volatile organic compounds in the bedroom air sample before sleeping were critically higher than other air samples. Machine Learning for Environmental Applications Supervisor: LAU Alexis Kai Hon / ENVR Co-Supervisor: FUNG Jimmy Chi Hung / ENVR Student: LEUNG Hok To / DSCT TAN Rickson Caleb Yap / DSCT Course: UROP 1100, Spring Inspired by the ChatClimate conversational AI framework developed by Vaghefi et al. (Vaghefi et al., 2023), this project seeks to integrate the findings of the Sixth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC AR6) into a large language model (LLM) to enable scientifically grounded responses on climate change for both general and expert audiences. Unlike ChatClimate’s reliance on external vector databases, our approach focuses on fine-tuning a standalone LLM, ensuring data privacy and allowing for tailored adaptations in downstream deployments. The project comprises two parallel workstreams: the first involves pretraining and fine-tuning the model on IPCC AR6 content, while the second develops a comprehensive evaluation framework to assess the model’s performance. Section 2 details the training, and Section 3 describes the evaluation.
RkJQdWJsaXNoZXIy NDk5Njg=