•A comprehensive framework consists of TDR installation, data acquisition, post-processing, and warning index.•The old colluvium slope at Lushan was characterized by TDR as a buckling fold.•Creep found by TDR contributed to the large surface displacement at Lushan.•Rebounded deformation was captured by TDR’s high temporal and spatial resolution at Jufen.•Simplified index for early warning was substantiated by field implementation. Due to the climate change and hazardous geological environment in Taiwan, real-time landslide monitoring for safety of adjacent tribes or infrastructures is much conspicuous. Time domain reflectometry (TDR) is considered as valuable technology for continuously capturing dynamics and locations of shear deformations with high temporal and spatial coverages. In this study, a comprehensive framework was proposed by integrating up-to-date procedure of installation guide, data acquisition, sliding deformation identification, and early warning processing. The advantages of the proposed framework were then examined and substantiated via field cases. The first case is at a large colluvium slope in Lushan, central Taiwan. The colluvium is still active after heavy rainfall events, but limited evidence of sliding characteristic was collected previously by manual inclinometer logging. Therefore, the TDR framework was deployed and multiple landslide surfaces at great depths were timely observed. Movement of colluvium mass was thereby interpreted as a buckling fold. Additionally, creep occurred among the colluvium layer during the monsoon season because of strength weakening influenced by pore pressure fluctuations. Monitoring results of the second landslide case in Jufan, northern Taiwan, reported the dynamics of a thin sliding plane captured with TDR high temporal resolution. This unique behavior was further applied for devising of a novel early-warning strategy in terms of time-variation of TDR response, validating that the comprehensive framework of TDR landslide can effectively contribute to early warning.