Abstract: The upper reaches of the Lancang River represent a critical geomorphic transition zone from the Qinghai-Tibetan Plateau to the Yunnan-Guizhou Plateau. The distinctive alternation of high mountain gorges and wide valleys in this region leads to frequent landslides, severely constraining sustainable socio-economic development. However, landslide monitoring studies in this area remain limited. Moreover, in existing hazard genesis attribution analyses, precipitation, a key triggering factor, is often oversimplified by correlation-based approaches, which fail to capture its non-linear spatiotemporal driving effects and lagged responses, thereby hindering quantitative interpretation of the underlying physical mechanisms. This study proposed an integrated research framework combining multi-source remote sensing-based collaborative monitoring with causal discovery. First, ascending and descending Sentinel-1 SAR data together with high-resolution optical imagery were used to perform regional early identification and high-precision time-series deformation monitoring of active landslides using the Small Baseline Subset InSAR (SBAS-InSAR) technique, followed by analysis of their spatial distribution characteristics. Subsequently, long time-series deformation data from two representative landslides with distinct geomorphic and deformation characteristics were selected, and Sequential Variational Mode Decomposition (SVMD) method was applied to separate periodic and trend components. Finally, to overcome the limitations of correlation analysis in identifying nonlinear lagged causality, the Convergent Cross Mapping (CCM) method was introduced to quantitatively diagnose the causal driving strength between precipitation and periodic landslide deformation, as well as the optimal lag time. The results indicate that: (1) Joint ascending-descending observations detected maximum line-of-sight deformation rates of −83 mm/year and −96 mm/year, respectively, and identified a total of 3,009 active landslides. (2) Active landslides exhibited significant spatial clustering, predominantly concentrated on steep slopes, within fault-influenced zones, and near river systems and transportation corridors. (3) The CCM method successfully captured the optimal lag effect between precipitation and periodic landslide deformation, with an optimal lag of 36 days for the Xiangda Village landslide. The CCM method effectively reveals the causal relationship between precipitation and deformation and its optimal lag, providing a better representation of the non-linear physical mechanisms governing evolution of complex landslide systems. This research not only expands the technical means for analyzing landslide deformation mechanisms in complex terrains but, more importantly, validates the applicability and strong potential of the proposed integrated framework combining multi-source remote sensing collaborative monitoring and causal analysis for deciphering the driving mechanisms of key factors in geological hazards. The findings provide a data-driven scientific support for elucidating the physical driving mechanisms and performing dynamic risk assessments of landslides in complex geomorphic regions.