Time Prediction of Sudden Landslides in Open-Pit Mines Based on Velocity Multiplication Inverse Element (VMI) Method

Sudden open-pit mine landslides are characterized by an extremely short acceleration phase and severe noise interference in monitoring data. Conventional single-point velocity multiplicative inverse (VMI) models suffer from low prediction accuracy and frequent early warning lags. To address these issues, this paper proposes an improved VMI-based landslide temporal prediction method integrated with the quadtree algorithm. First, inverse distance weighting (IDW) interpolation is employed to interpolate discrete monitoring data into continuous two-dimensional spatial grid data. Subsequently, the quadtree algorithm is introduced to achieve spatial data compression by setting a deformation difference threshold, which eliminates redundant data in stable regions and identifies the core main sliding block. Next, monitoring points distributed within the main sliding block are extracted, and the arithmetic mean of their monitoring data is calculated to construct a “virtual monitoring point”, thereby eliminating random fluctuations in single-point data. Finally, the smoothed and denoised inverse velocity sequence is input into the VMI model to predict the instability time. The extraordinary landslide that occurred at the Pingzhuang West open-pit mine on 17 April 2013 is taken as an engineering case for method validation. The results demonstrate that compared with the traditional single-point prediction model, the inverse velocity–time curve obtained by the proposed improved method exhibits a smoother linear attenuation trend during the accelerated deformation phase, and the coefficient of determination (R2) of the linear fitting increases from 0.94 to 0.97; the prediction error (ΔT) is reduced from a 0.4 day lag to an approximate 28 min advance, which effectively addresses the defects of poor anti-interference capability and early warning lag of traditional single-point prediction, and conforms to the engineering principle of “prefer early warning rather than delayed warning” in disaster monitoring. Based on data from existing monitoring networks, the improved methodology requires no additional investment, contributes to reducing casualties caused by landslide disasters and additional carbon emissions, and supports the safe and sustainable development of mineral resources.