Deep Learning-Based Short-Term Weather Forecasting and Optimal Data Quantity Analysis for Urban Air Mobility

Urban Air Mobility (UAM) is an innovative aviation system designed for efficient operation in urban areas, providing solutions for rapid transportation of passengers and cargo. The safe operation of UAM depends on accurate predictions of weather phenomena such as turbulence, gusts, and icing conditions. However, the lack of observational data in the atmospheric boundary layer, which is UAM's primary operational area, has limited research on real-time weather prediction (nowcasting). To address this issue, this study utilizes high-resolution meteorological data from the Boseong Weather Observation Tower, which provides detailed insights into atmospheric boundary layer conditions. This study improves prediction accuracy by combining Long Short-Term Memory (LSTM) networks, specialized in time series data analysis, with Fully Connected Layers. Additionally, through power spectrum analysis, we investigate the impact of meteorological variable characteristics on prediction performance and identify the optimal amount of data required for deep learning-based real-time weather prediction models. The results of this study are expected to contribute to the development of real-time weather prediction models that can enhance the safety and efficiency of UAM operations. By addressing observational data gaps and utilizing deep learning technology, this study aims to contribute to establishing UAM as a reliable urban transportation solution. Furthermore, the methodology proposed in this study can be applied to building weather prediction systems for UAM operations in various urban environments, which is expected to play a crucial role in the development of sustainable transportation infrastructure for smart cities.