Conventional outputs of physics-based landslide forecasting models are presented as deterministic warnings by calculating the safety factor ( F _s ) of potentially dangerous slopes. However, these models are highly dependent on variables such as cohesion force and internal friction angle which are affected by a high degree of uncertainty especially at a regional scale, resulting in unacceptable uncertainties of F _s . Under such circumstances, the outputs of physical models are more suitable if presented in the form of landslide probability values. In order to develop such models, a method to link the uncertainty of soil parameter values with landslide probability is devised. This paper proposes the use of Monte Carlo methods to quantitatively express uncertainty by assigning random values to physical variables inside a defined interval. The inequality F _s <1 is tested for each pixel in n simulations which are integrated in a unique parameter. This parameter links the landslide probability to the uncertainties of soil mechanical parameters and is used to create a physics-based probabilistic forecasting model for rainfall-induced shallow landslides. The prediction ability of this model was tested in a case study, in which simulated forecasting of landslide disasters associated with heavy rainfalls on 9 July 2013 in the Wenchuan earthquake region of Sichuan province, China, was performed. The proposed model successfully forecasted landslides in 159 of the 176 disaster points registered by the geo-environmental monitoring station of Sichuan province. Such testing results indicate that the new model can be operated in a highly efficient way and show more reliable results, attributable to its high prediction accuracy. Accordingly, the new model can be potentially packaged into a forecasting system for shallow landslides providing technological support for the mitigation of these disasters at regional scale.

A slope unit is commonly used as calculation unit for regional landslide analysis. However, the capacity of the slope unit to reflect the geomorphological features of actual landslides still needs to be verified. This is because such accurate representation is critical to ensure the physical meaning of results from subsequent landslide stability analysis. This paper presents work conducted on landslides and slope extraction in two areas in China: The Jiangjia Gully area (Yunnan Province) and Fengjie County (Chongqing Municipality). Ground-based light detection and ranging (LiDAR) data are combined with field landslide terrace measurements to allow for the comparison of slope unit extraction methods (conventional vs. MIA-HSU) in terms of their ability to reflect the geomorphological features of shallow and deep-seated landslides. The results indicate that slope unit boundaries extracted by the conventional method do not match the geomorphological variations of actual landslides, and the method is therefore deficient in meaningfully extracting slope units for further landslide analysis. By contrast, slope units obtained using the MIA-HSU method accurately reflects the geomorphological features of both shallow and deep-seated landslides, and thus provides clearer geomorphological meaning and more reasonable calculation units for regional landslide assessment and prediction.