Abstract: Aiming at the problem that traditional outlier identification methods for dam deformation monitoring data are difficult to characterize the spatiotemporal correlation of outliers at each measuring point, this paper proposes a novel outlier identification method by fusing spatiotemporal clustering and the improved local outlier factor (ILOF).Firstly, the maximum-minimum distance method and dynamic time warping (DTW) algorithm are combined to construct a spatiotemporal comprehensive distance measurement model.Secondly, the K-Means algorithm is used to realize the spatiotemporal clustering partition of measuring points.Then, representative measuring points are selected in each sub-area, and ILOF algorithm based on weighted Mahalanobis distance is adopted to identify outliers.Finally, the effectiveness of this method is verified by analyzing the variation of environmental correlation of the measuring points.The application of this method in the deformation monitoring of Jinping Ⅰ dam shows that under different degrees of outlier interference, the recall rate is 93.02%, the precision rate is 88.54%, and the accuracy rate is 96.54%.Compared with the traditional LOF, K-Means-LOF and DTW-based identification methods, the recognition accuracy is significantly improved.By deeply mining the spatiotemporal correlation characteristics of monitoring data, this method remarkably improves the accuracy of outlier identification, provides effective technical support for the quality control of dam safety monitoring data, and has important engineering application values.