Representation And Analysis Of The Morphology Of Loessal Erosion Gullies At High Resolution

Being the primary component of the landform pattern of the Loess Plateau,the erosion gully has always been a key research topic in the field of soil erosion and geomorphology,in that it serves both as the product of landform evolution and,in some degree,of soil erosion.There were many limitations of early studies in surveying and mapping technology and data accumulation,which were because a comprehensive and effective expression and analysis of erosion gullies at the watershed scale,cannot be conducted at middle and low resolutions,especially for small and active erosion gullies.Therefore,it was not suitable for the requirements of accurate simulation of soil erosion.In recent years,with the development of remote sensing topographic survey technology at high-resolution,the related studies on erosion gullies have obtained a fresh opportunity.Given this context,by using modern measurement methods such as Global Navigation Satellite System(GNSS),Unmanned Aerial Vehicle(UAV)aerial survey and remote sensing,this thesis takes the erosion gully as the research object,and then analyzes the high-resolution expressions of erosion landform of morphological characteristics,erosion gully extraction(the areas of permanent gully and newly incised valley),and the scale effect among extracted topographic variables.This thesis has further deepened the micro and macro features of erosional landform,thus providing a scientific support for the study of soil erosion at high-resolution.It consists of the following research contents and conclusions:(1)The high-resolution acquisition of erosion gully morphology can be achieved using low altitude UAV photogrammetry: By analyzing the morphological characteristics of existing erosion gully types as well as new topographic surveys using remote sensing techniques,a data acquisition system for erosion gully morphology has been proposed based on the opportunities of low altitude UAV photogrammetry.The technical system has the advantages of speed,flexibility,low cost,and wide area of operation,which can ensure the accuracy of landform expression at a lower-cost performance and meet the requirements of data acquisition for the high-resolution expression of erosion gully morphology.(2)The digital expression of erosion gully morphology can be established by fully considering man-made morphological characteristics and natural morphological characteristics of the evolving land forms: Focusing on the analysis of the man-made morphological characteristics such as terraces and the natural morphological characteristics of evolving land forms including gully edges,this thesis constructs the lines of evolution of the morphological characteristics using the interpolation operation to generate DEM(Digital Elevation Model).It then compares and analyzes the most important elevation characteristics(surface characteristics,area elevation integral,and section line elevation difference),slope characteristics and profile curvature characteristics(statistical distribution and spatial pattern),and establishes a high-resolution digital expression of erosion gully morphology.It thus provides a more accurate basic data for the study of geomorphological characteristics of micro scale erosion.(3)Based on machine learning and object-oriented image analysis,the high-resolution extraction of erosion gully morphology can be realized: By using a multi-resolution data set constructed from the high-precision landform or image data,this thesis selects the object-oriented analysis of the integration of landform data or image data and uses the automatic classification strategy of Random Forests.It takes trenches or gullies and the evolving land forms as the extraction objects,and establishes the effect of scale in the high-resolution series data environment on the erosion gully extraction.The study shows that there are significant advantages of the high-resolution data set(0.2m DOM,Digital Orthophoto Map +1m DEM)in the extraction of erosion gullies,and the classification result is closest to the actual spatial distribution of the objects.However,as the data set resolution decreases(as low as 5m DOM+5m DEM),the overall classification accuracy drops from90.74% to 53.63%.The main extraction differences are found in gully heads with relatively broken and complicated terrain,in the parts with insignificant gully characteristics,and in the small-scale gully areas formed by short-term evolution.(4)The accuracy of the topographic parameters of the high-resolution erosion gully changes regularly with the resolution: In terms of DEM data accuracy,with the decrease of resolution(1m?5m),the standard deviation,mean square error,mean absolute error and terrain description error all increase gradually.In the aspect of the scale effect of slope extraction,with the decrease of resolution,slope attenuation occurs to varying degrees,which is more serious in areas where the slope changes rapidly(morphological characteristic line and valley area),such as the average slope of Wangmaogou watershed sample area decreased from 31.80° to 29.70°,and Erlaohugou watershed sample area decreased from 19.88° to16.73°.For the scale effect of stream line extraction,with the decrease of resolution,the level of extracted stream lines of the two sample areas is reduced from three to two,and the extraction number is reduced from 28 to 5(Wangmaogou watershed sample area)and 25 to 4(Erlaohugou watershed sample area),respectively,and the accuracy and integrity of the whole channel expression are reduced to a certain extent as the resolution decreases.In terms of the scale effect of LS factor extraction,with the decrease of resolution,the LS factor shows an upward trend,and accordingly,and the spatial distribution range of larger values also increases,specifically,the average values increased from 10.97 to 15.22(Wangmaogou watershed sample area)and from 6.03 to 7.34(Erlaohugou watershed sample area).