红壤坡面水蚀微地貌发育机理及动态模拟研究

Microrelief is an important factor affecting the slope erosion process. Exploring the relationship between microrelief development and water erosion process is one of the key issues for increasing slope soil erosion model precision. Field investigations, rainfall simulations, system modeling and numerical simulations are conducted study the slope surface microrelief in hilly red soil region in south China. This project will explore the characteristics of microrelief development, runoff and sediment yield on slope, clarify the coupling mechanism between microrelief development and water erosion process, and build the microrelief development model and its dynamic simulation system. The above research contents are of great significance for revealing mechanism and building model of red soil slope erosion. Three-dimensional laser scanner and close-range photogrammetry are utilized to obtain the developing microrelief in water erosion process. The methods of fractal, stochastic process and geostatistics are used to describe the spatial distribution and dynamic evolution features of microrelief, and to study the relationship of microrelief traits with slope degree and tillage pattern. The microrelief development model of red soil slope is built based on cellular automata. The dynamic simulation system is established on OpenCL, which can predict runoff and sediment yield, simulate and visualize the microrelief development. The project has important scientific significance. Its research results can provide theoretical support for soil and water loss management, side slope engineering protection, etc.

坡面微地貌是影响坡面水蚀过程的重要因素，深入理解微地貌发育与水蚀过程的相互作用，是提高坡面水蚀模型精度的关键问题之一。本项目以南方红壤丘陵区坡面微地貌为研究对象，通过野外调查、降雨试验、系统建模和数值模拟，探讨微地貌发育特征与坡面产流产沙规律，阐明微地貌发育与水蚀过程的耦合关系，构建微地貌发育模型及动态模拟系统，对于揭示红壤坡面水蚀机理和建立红壤坡面水蚀模型具有重要意义。利用三维激光扫描和近景摄影测量获取水蚀过程中的微地貌发育形态，结合分形、随机过程和地统计学等方法定量描述微地貌的空间分布特征和动态发育特征，探索微地貌空间分布特征与坡度和耕作措施等因素间的关系；基于元胞自动机思想构建红壤坡面微地貌发育模型，建立基于开放计算架构的动态模拟系统，集产流产沙预测、微地貌演化和数据可视化于一体。项目成果将为南方红壤丘陵区的水土流失治理、边坡工程防护等提供理论依据，具有重要的科学意义。

为了探究坡面微地貌与坡面水蚀过程的耦合机制，本项目开展了不同坡长坡度、耕作措施、植被覆盖和砾石覆盖条件下的红壤坡面降雨侵蚀试验。针对不同的现场环境，分别利用三维激光扫描和倾斜摄影测量方法获取坡面微地貌数据，两种方式获取的点距精度均达到毫米级，基于无人机的倾斜摄影方式解决了三维激光扫描可能出现的孔洞问题。针对室外坡面点云存在杂草、灌木等目标而难以人工分割的问题，设计了包括空间变换层、特征抽象层和特征分类层的深度卷积神经网络，具有旋转和平移不变性，实现了端到端的坡面点云自动分割，融合反射强度信息后的平均分类精度达到94.3%。从结构化与自相似角度，采用半变异函数与多重分形谱结合的方式表征坡面微地貌空间的分异性，设计了分形生成与分析系统。为了降低人工采集降雨试验数据的成本，设计了壤中流自动测量装置，通过调节水位开关位置可以适应不同的时间采样精度。针对人工测量坡面流速主观性性大的问题，提出了基于坡面径流波纹特征的稠密光流法测量坡面流速，平均正确率达到90.4%。通过人工降雨试验，获得了不同耕作措施、地表覆盖等条件下的红壤坡面壤中流产流、径流产流、产沙规律以及微地貌形态分布特征，提出了基于生长曲线的壤中流产流特征拟合模型。从复杂系统和并行计算角度，设计了基于元胞自动机的红壤坡面降雨侵蚀模型，并设计了基于异构计算的坡面侵蚀过程可视化模拟系统。本项目研究为坡面降雨侵蚀试验提供了新思路，可为红壤坡面治理提供参考。