Abstract: Conventional outlier detection methods for dam deformation monitoring data often fail to adequately capture spatio-temporal correlations among outliers across different measurement points. To overcome this limitation, this study introduces a novel approach for detecting outliers in concrete dam deformation by integrating spatio-temporal clustering with an improved Local Outlier Factor (ILOF) algorithm. The proposed methodology involves four main steps: First, a composite spatio-temporal distance metric is developed by combining the maximum-minimum distance method with the Dynamic Time Warping (DTW) algorithm. Second, the K-Means algorithm is applied to partition monitoring points into clusters based on the proposed distance metric. Third, representative points within each cluster are selected and analyzed using an ILOF algorithm enhanced with weighted Mahalanobis distance. Finally, the effectiveness of the method is evaluated by examining variations in environmental correlations across monitoring points. Experimental results demonstrate that the proposed method achieves a recall (R) of 93.02%, precision (P) of 88.54%, and accuracy (A) of 96.54% under varying levels of outlier interference. Comparative analysis shows that the proposed method outperforms conventional approaches including standalone LOF, K-Means-LOF, and DTW-based methods in terms of detection accuracy. By effectively leveraging spatio-temporal correlations within monitoring data, the method significantly improves detection precision and offers a robust framework for quality control of dam safety monitoring data, demonstrating substantial potential for practical engineering applications.