Accurate differentiation between water and land is crucial for flood monitoring, land-use planning, and ecological protection. Airborne laser scanning (ALS) provides high-resolution three-dimensional topographic and intensity data, offering a robust foundation for these applications. However, existing classification approaches largely depend on manually defined thresholds or supervised learning, with limited attention to the underlying synergies among multidimensional features, thereby constraining both accuracy and automation. To address these limitations, this study introduces an unsupervised classification framework based on multi-feature fusion, where four fusion indicators are derived from 14 geometric and radiometric features, enabling fully automated water–land classification. Experiments on two ALS datasets from central Dublin demonstrate that the proposed method substantially outperforms SLIER, fuzzy logic, and elevation-threshold approaches, achieving overall accuracies of 98.3% and 97.3%, with Kappa coefficients of 0.903 and 0.914. Beyond improving classification accuracy and reducing computational complexity, the fusion indicators also enhance digital elevation model (DEM) reconstruction by repairing voids in water regions and refining boundary delineation, thereby reinforcing the value of ALS data for environmental monitoring and disaster management.
zhao et al. (Mon,) studied this question.