STABILITY ESTIMATION OF LANDSLIDE DAMS

Landslide dams usually formed by collapse, landslide, debris flow, et al.,which have the characteristics of sudden formation and great dangerous. If the dam breaching, it is easy to flood the downstream area by surges and cause huge loss of life and property. However, the study of rapid stability assessment of landslide dams is greatly restricted due to its various geometrical shapes, inhomogeneous material composition and complex internal structure. In this paper, a database with 1757 landslide dams from all over the world was compiled. The influences of dam characteristic parameters(triggers, hydrology/geometric parameters, materials) and dam stability are analyzed. Rainfall-induced dam and the dam with soil material have the worst stability, mainly because of the material particles of rainfall-induced landslide dam are relatively small water content, while the soil landslide dams usually have poor permeability and low impact velocity. Besides, the dam stability first increase and then decrease with dam height and lake volume increase because of when the dam height is too small, it's easy to overflow and when the dam height too high, the reservoir water potential energy increases. The dam stability increases with the dam width and dam volume increase, the main reason is that with the dam width and volume increase will increase the resistance of the dam to sliding. The dam stability decreases with the average discharge increase, the reason is that the increase of the inflow rate reduces the overflow time. It is also found that landslide dam stability cannot be predicted by one or two influence factors since it is affected by multi factors. Based on seven dimensionless variables, a rapid evaluation model for estimating the stability is developed. The accuracy of the model is 85.1%. The model developed in this paper show better accuracy due to considering dam materials and triggers. Case studies(Jiala landslide dams) are presented to show the application of the model developed in this paper.