Impacts of Urban Agglomerations on Weather Dynamics: A Multifaceted Initiative for Better Understanding, Global Urban Monitoring, and Improved Urban-Rainfall Forecasting

Abstract

Urban areas are expanding rapidly and becoming increasingly connected, which has serious effects on the environment around them. The first objective of this dissertation evaluates this interaction by looking at how these connected urban clusters affect air quality at both local and regional levels. It shows that traditional studies, which often focus on cities as separate units, tend to miss the combined impacts of these larger urban networks. The investigation further pointed to a notable gap in the availability of continuous, long-term, high-resolution, and globally consistent datasets capable of capturing changes in urban areas over time, especially in rapidly developing regions of low- and middle-income countries. This kind of data is particularly important for understanding how cities grow and affect their surrounding weather and climate systems. To fill this gap, the second objective develops a new dataset called the Normalized Difference Urban Index Plus (NDUI+). This dataset uses satellite images and advanced deep learning techniques to produce yearly maps of global urban areas at a high resolution (30 meters) from 1999 to the present. NDUI+ combines data from several satellite sensors (such as VIIRS, DMSP-OLS, and Landsat) and corrects for differences caused by changes in sensors over time, creating a consistent and reliable product. It is well known that cities can affect local weather patterns, but weather prediction models often struggle to simulate these effects accurately. One major reason is the lack of up-to-date, detailed urban data. The third objective of this study tries to solve this issue by using NDUI+ data in the Weather Research and Forecasting (WRF) model to see if it can improve rainfall prediction over cities. Three urban fraction configurations (NLCD 2011 (default to WRF), NLCD-2020, and NDUI+ 2020) were tested over the city of Chicago to simulate a deadly Derecho event from the year 2020. Results showed that improved and updated urban representation from NDUI+ data yields more accurate rainfall forecasts. Overall, NDUI+ urban fraction-based total rainfall forecast achieved a 40.4% improvement in rainfall forecast accuracy compared to the default NLCD 2011 urban fraction case, and 21.6 % better accuracy than NLCD 2020 urban fraction configuration over the city.