Fractal Features Of Landslides In The Area Southern Gansu Province In The Bailongjiang River Basin

Considering that it is one of the more severely affected areas, this paper selects a research area of the Gansu section of the Bailongjiang River Basin. Based on the latest available survey data of geological disasters, we analyzed the distribution and the fractal features of the landslide hazards in detail using the fractal theory with GIS software. The main conclusions are as follows:(1) The spatial distribution of landslide hazards are more concentrated which mostly located in the middle reaches of the Bailongjiang River Basin, especially in Wudu area. A vast majority of landslide hazards are distributed in the slope of20～40°, S slope and SSW, SSE slope, within the range of1200-2100m elevation and phyllite, slate and thin limestone rock group. In addition, the landslide hazards are also densely distributed along the river and road and near the fault zone.(2) The spatial distribution of landslides in the study area is in second-order cumulative total fractal distribution with the elevation, fault, rivers, roads, gradient, lithology and landuse whose fractal dimension is larger, and is in one-order cumulative total fractal distribution with the aspect and rainfall having smaller fractal dimension. The value of the fractal dimension reflects the degree of influence of each factor on the spatial distribution of landslides to some extent. The more times the cumulative transformation and the larger the fractal dimension, the more complex the variable dimension fractal features between the landslide and the above factors.(3) The cumulative or non-cumulative frequency-area of landslides meets the negative power law distribution generally. As reflected in the corresponding graph, it is the heavy tail part of the curve. Conversely, the smaller landslides show a rollover with the gentler slope.(4) The scale parameter of landslide hazards has power-law dependence. We put forward the empirical relationship between its area and volume which provides relevant evidence for calculating landslides volume by the area. On this basis, the relationship between the landslide area, length and width is also obtained.(5) We analyzed the probability distribution of the "5.12" earthquake-induced landslides based on a three parameters inverse gamma probability distribution model, and predicted the total number of the landslides. The further analysis of landslide probability density curve for power-law section (tail of the curve) indicates that landslides with area in the region of794～1995m2play a more important role in the evolution of landforms, landslide risk and other aspects.