Efficient air quality monitoring relies on reliable data preprocessing to ensure the precision and stability of predictive models. In urban environments, air pollution exhibits both temporal and spatial variability, necessitating specialized preprocessing techniques to capture these dynamics accurately. This study presents a data preprocessing framework that applies temporal and spatial clustering to refine air quality monitoring data. Temporal clustering is used to group data according to time-based patterns, such as daily and seasonal fluctuations, helping to reveal periodic trends and anomalies. Spatial clustering, on the other hand, organizes data by geographic location, allowing for the identification of localized pollution patterns and sources. By combining these two clustering approaches, the preprocessing framework enables a comprehensive understanding of air pollution distribution, providing a foundation for subsequent modeling efforts. This approach is validated on a large-scale urban air quality dataset, showing improved data consistency and enhanced model performance in predicting pollutant levels.