On December 18, 2023, a destructive M S . 6.2 earthquake occurred in Jishishan County, Gansu Province, China. The earthquake triggered a loess earthflow in Zhongchuan Township, resulting in 13 fatalities. The catastrophic earthflow moved about 3 km without previous rainfall and its mobility index (H/L) is as low as 0.026. We combined multi-source data, including remote sensing images, drone images, and field investigation data, to analyze the mechanism of this disaster. It is found that highly saturated loess liquefied under intense seismic vibration, eventually forming an earthflow and moving rapidly along the northern gully. Moreover, a depth- integrated continuum model that comprehensively considers soil and water lateral pressure is adopted to reveal the above-mentioned mechanism. A MUSCL-HLLC finite volume method with well wet-dry boundary treatment is proposed to solve the full dynamic process of the earthflow. The simulated results reproduce the movement process of this disaster and are in good agreement with the measured data. Compared with other models, the coupling model of soil pressure and water pressure with the partitioned pore water pressure ratio can improve the simulation accuracy. This comprehensive analysis offers valuable insights and guidance for the risk management and prevention of such disasters.