Study On Fractal Characteristics Of Soil Erosion Influencing Factors In Dianchi Basin Based On GIS

Focused on Dianchi Basin, this study calculated the fractal dimension values of key soil erosion influencing factors, involving contour, river network and land use types based on fractal theory. Through analyzing the physical meaning of these factors'fractal dimension values and fitting the none linear relationship between them and soil erosion fractal dimension values, this thesis discussed the inherent mechanism of soil erosion and also did a beneficial exploration for applying fractal theory into soil erosion prediction. The studied area had been divided into mountainous and plain areas at the first level and 107 sub watersheds at second level after extracting the topographic information of contour, slope and river network in this area. After completing the calculation of fractal dimension values, through the changes of basin general fractal dimension values under different extracting distance and distinct river threshold values, the comparisons of mountainous and plain general fractal dimensions, the calculation and analysis of terrain, river network and land use type fractal dimension values in sub watersheds, we acquired the physical meanings of kinds of fractal dimension values and the relation among them. Finally, taking advantage of factor analysis and back propagation neural network, we got the prediction model among soil erosion fractal dimension values and terrain, river network and land use type fractal dimension values. This paper got the conclusions as followings:With the increase of extracting distance from 5 to 100 meters, the basin general contour fractal dimension values decrease from 0.9983 to 0.9756. The general contour fractal dimension values, the average and peak values of sub watersheds in mountainous area are bigger than those in plain area. With the increase of river threshold values from 100 to 1500 meters, the basin general river network fractal dimension values decrease from 1.2665 to 1.0702. At the same time, the relation factor of logarithm fitting between threshold values and their river network fractal dimension values reaches to 0.9984, which offers new dimension to the determination of threshold values during the river network extracting process. The general river network fractal dimension values and average values of sub watersheds in plain area are bigger than those in mountainous area. Through comparing the fractal dimension values and stability indexes of three land use types in mountainous and plain area, it indicates that three types of general land use fractal dimension values in mountainous area all bigger than those in plain area. Forest land is the most stable in both areas, grass land is most unstable in mountainous area, and arable land is most unstable in plain area. After testing, the predicting model with other influencing factors' fractal dimension values as input values and soil erosion fractal dimension values as output values is stable and reliable. However, the value of R is not high enough, so we still need some other factors in the model to complement the remaining information.