Abstract
Water is essential for life, human activities, and ecological balance. Understanding the spatiotemporal dynamics of urban water bodies is critical for sustainable water management, especially in fast-growing cities like Zhengzhou. This study investigates the evolution of water bodies in Zhengzhou from 2014 to 2023, using multi-seasonal Landsat 8 imagery and GIS analysis. Three water extraction methods-thresholding, random forest (RF), and a hybrid RF model incorporating spectral indices (NDWI, MNDWI, EWI) and GLCM-based texture features-were evaluated. The hybrid RF approach demonstrated the highest classification accuracy and was applied to generate seasonal water body maps for the target years. The results show a general increase in surface water area, with seasonal peaks in July–September, and spatial expansion mainly in the northern and northeastern regions. These changes are driven by both natural (e.g., precipitation) and anthropogenic (e.g., urban development, artificial lake construction) factors. Compared to previous studies, this research offers three key innovations: (1) a decade-long, seasonal-scale analysis of water dynamics; (2) a robust multi-feature classification framework for complex urban settings; and (3) an integrated interpretation of natural and human influences. The findings provide valuable references for water resource planning and urban ecological restoration, and the methods used may be transferable to similar inland cities under monsoon climates.