| Summary: | Quantitative research on and the prediction of a landslide deformation area is an important point to accurately and comprehensively understand the failure mechanism of landslides and the degree of slope failure. This study uses image processing techniques to quantitatively identify the area and volume of deformation regions during rainfall-type landslide destabilization under multifactor conditions. The findings revealed that (1) an increase in rainfall intensity and slope angle, as well as the existence of slope crest, will accelerate the process of slope instability. In our study, when the rainfall intensity was 140 mm/h and the landslide volume reached 35.68%, the landslide failure was the most serious. (2) Slopes with high compaction of subsoil as well as those without perimeter pressure are relatively more damaged. (3) The higher the density of vegetation cover, the stronger the protection ability of the slope, and the higher the wind speed, the greater the failure to the slope. Furthermore, an improved S-growth curve model was proposed to predict landslide volumes in 16 sets of experiments. In detail, the proposed S-growth curve model predicted landslide volumes with an average absolute percentage error of 4.34–16.77%. Compared with the time series analysis moving-average method (average absolute percentage error of 6.39–68.89%), the S-growth curve model not only has higher prediction accuracy but also can describe the three stages of deformation region development from a physical perspective and can be applied to the volume during landslide change prediction.
|
|---|