<p>Accurate parameter calibration is essential for reliable landslide numerical simulation. However, conventional trial-and-error approaches introduce substantial uncertainty, which limits the accuracy of hazard zonation and process reproduction, thereby reducing the effectiveness of disaster assessment and risk mitigation. To address this challenge, this study develops a Smooth Particle Hydrodynamics (SPH)-based landslide model and proposes an automated parameter optimization framework grounded in particle system theory. The framework integrates parameter dimensionality reduction, landslide footprint tracking, and iterative optimization, enabling dynamic adjustment of SPH model parameters in response to evolving hazard zones. The method is applied to the 2017 rainfall-induced long-runout landslide at Xinmo, Sichuan Province, China. Results demonstrate that the proposed approach achieves a simulation accuracy of 85.42% in reproducing the Xinmo landslide, with high reliability in predicting deposit thickness and sliding velocity, and effectively capturing spatiotemporal hazard information throughout the failure process. This study demonstrates the feasibility of the proposed optimization strategy within the SPH framework, and the method shows potential value for application to other particle system-based numerical models of geological hazards. These findings provide valuable technical support for engineers and decision-makers and contribute to the development of more effective measures to reduce landslide risk in vulnerable regions.</p>