溯源侵蚀作为沟道发育的重要方式，在中国黄土高原丘陵沟壑区发生频繁，对该区域生态环境和社会经济发展造成了严重影响。已有研究探讨了溯源侵蚀时空分布、影响因素、力学机制、模型模拟等，但仍存在定量研究匮乏、过程认识不清、力学机制不明、模型精度低等问题。随着遥感技术在地学研究中推广，基于三维激光扫描技术的溯源侵蚀研究为上述研究问题的解决提供了新思路。鉴于此，本研究在黄土高原丘陵沟壑区原位坡面小区开展放水冲刷试验，以研究沟头溯源侵蚀过程，采用高速摄像机等采集产流数据，利用三维激光扫描技术获取沟头密集三维点云。在此基础上，提出了基于点云的溯源侵蚀量化算法；研究了溯源侵蚀产流产沙特征及其形态演变过程；分析了溯源侵蚀地形特征、上方来水特征对产流产沙的影响；量化了溯源侵蚀子过程的贡献量，探明了溯源侵蚀子过程的形态演变过程；研究了溯源侵蚀水力机制和沟头重力侵蚀的土力机制，建立并验证了溯源侵蚀模型，实现了溯源侵蚀过程的模拟。主要结论如下：

（1）提出了基于地形变化点云的复杂地形变化空间微分算法（3D-M3C2），与实测数据对比表明，3D-M3C2算法可准确量化沟头回退过程中的侵蚀-沉积-产沙过程（平均相对误差<12%）；基于3D-M3C2算法提出了重力侵蚀-水力侵蚀分离方法，为沟头溯源侵蚀中的水力侵蚀和重力侵蚀独立量化提供了方法支撑；改进了传统切片法，并基于此提出了溯源侵蚀子过程提取算法、溯源侵蚀形态特征提取算法，为溯源侵蚀子过程量化、三维形态特征量化奠定基础。

（2）开展了溯源侵蚀水沙过程和形态演变及其影响因素研究，结果表明，溯源侵蚀由水力侵蚀和重力侵蚀构成，其中水力侵蚀是其产沙量（贡献量>87.22%）和侵蚀量的主要来源（贡献量>94.29%），而重力侵蚀主要影响沟头形态特征。溯源侵蚀在水平尺度上为水力侵蚀主导的沟头回退和重力侵蚀主导的沟头拓宽交替出现的循环过程；在垂直尺度上为水力侵蚀主导的沟壁下方内凹洞扩张和重力侵蚀主导的沟壁上方回退交替出现的循环过程，内凹洞扩张与重力侵蚀为相互促进的耦合。溯源侵蚀产流产沙影响因素中，径流量和径流宽显著影响产沙量，而地形不仅是产沙的主要影响因素，还显著影响产流。

（3）研究了溯源侵蚀子过程及其时空分布特征，结果表明，溯源侵蚀由内凹洞发育区扩张（对产沙量平均贡献67.09%）和崩塌区回退（对产沙量平均贡献33.35%）主导，其中内凹洞发育区扩张和崩塌区回退之间存在相互促进的耦合。沟壁回退过程由汇水区来水分离出的占比约15%的贴壁流驱动，这部分贴壁流同时也是溯源侵蚀泥沙运移的主要载体；而占比约85%的跌水主要影响溯源侵蚀产流特征和协助溯源侵蚀泥沙运移，仅有在内凹洞顶端分离的部分跌水（占汇水区来水约10%）会对沟壁产生轻微的侵蚀。

（4）开展了沟头溯源侵蚀力学机制研究并构建了理论模型，结果表明，溯源侵蚀中沟壁水力侵蚀驱动力为径流对土壤的滑动摩擦力，决定此滑动摩擦力大小的主要因素为局部地形坡度、土壤对径流的吸附力、径流流量。崩塌区重力侵蚀核心驱动力为悬空土体重力对悬空体断面处产生的空间力偶，决定重力侵蚀是否发生的主要因素主要为悬空体的含水量、断裂面处内聚力、悬空体形态特征。内凹洞发育区的扩张通过增大悬空体形心至断裂面的水平距离，减少悬空体高度，使得断裂面顶部受拉力的增大，进而促进重力侵蚀；沟头重力侵蚀通过减少内凹洞的顶部长度减少沟壁径流入渗量，进而促进内凹洞发育区扩张。基于力学机制构建的理论模型由沟头水力侵蚀模拟模块和重力侵蚀运算模块构成，与实测数据的对比表明，此模型可准确模拟溯源侵蚀过程（回退量模拟结果的相对误差小于15%）且具备预测沟头重力侵蚀的潜力。

Gully headcut erosion is an important process driving the development of gullies, which occurs frequently in the hilly and gully Loess Plateau of China, leading to adverse impacts on the ecological environment and socio-economic development of the region. Previous studies on gully headcut erosion have focused on its spatial and temporal distribution, influencing factors, mechanical mechanisms, and model simulation. However, those studies were still subject to limitations such as lack of quantitative studies, insufficient understanding of processes and underlying mechanisms, and low accuracy of models. With the application of remote sensing techniques in geoscientific research, the investigations of gully headcut erosion based on 3D laser scanning technology provides a promising means for addressing the above research problems. In view of this, this study carried out a runoff scouring experiment in a field runoff plot established in the hilly and gully Loess Plateau in order to study the gully headcut erosion processes, using equipment (e.g., high-speed cameras) to collect runoff generation data, and 3D laser scanning technology to obtain dense point cloud data of the gully headcut. On this basis, several algorithms for gully headcut erosion detection was proposed based on 3D point clouds; the characteristics of runoff generation and sediment yield of gully headcut erosion as well as morphological evolution of gully headcuts were clarified; the influence of the topography and upstream runoff on the runoff generation and sediment yield were elucidated; the contribution of the sub-processes to gully headcut erosion was quantified and the morphology of the sub-processes was revealed; the hydraulic mechanism of gully headcut erosion and geotechnical mechanism of mass movement at gully headcut were investigated, and gully headcut erosion model was established and validated to achieve the simulation of gully headcut erosion processes. Main findings are as follows:

(i) A differential algorithm for complex terrain change (3D-M3C2) based on terrain change point cloud was proposed. A comparison with measured data (i.e., runoff sample) showed that the 3D-M3C2 algorithm was able to accurately quantify the mass of erosion, deposition, and sediment yiled during the gully headcut erosion (the average relative error was <12%). Based on 3D-M3C2 algorithm, a mass movement-hydraulic erosion separation method was proposed, which provided a methodological support for the independent quantification of hydraulic erosion and mass movement during gully headcut erosion; the traditional slicing method was adapted, and the algorithm for extraction of the sub-processes and extraction of morphological features were proposed based on the adapted slicing method, which facilitated the quantification of the sub-processes and the quantification of the 3D morphological features.

(ii) The processes of runoff generation and sediment yield, morphology of gully headcut erosion, and its influencing factors were investigated. Results showed that gully headcut erosion consisted of hydraulic erosion and mass movement, of which hydraulic erosion was the main source of sediment yield (contribution > 87.22%) and erosion (contribution > 94.29%), while mass movement mainly influenced the morphological characteristics of the gully headcut. Gully headcut erosion was found to be a cyclic process that alternated between hydraulic erosion-driven gully headcut retreat and mass movement-driven gully headcut widening on the horizontal perspective; on the vertical perspective, it was demonstrated to be a cyclic process that alternated between hydraulic erosion-driven cave expansion below the gully headwall and mass movement-driven gully shoulder line retreat on the upper gully headwall. A coupling relationship was found between cave expansion and gully shoulder line retreat. Among the influencing factors of runoff generation and sediment yield, runoff discharge and runoff width significantly affected sediment yield, while topography was not only the main influencing factor of sediment yield, but also significantly affected runoff generation.

(iii) The sub-processes of gully headcut erosion and their spatial and temporal distribution characteristics were studied. Results showed that gully headcut erosion was dominated by the expansion of the cave areas (contributing 67.09% of sediment yield on average) and the retreat of collapse areas (contributing 33.35% of sediment yield on average), with the expansion of cave areas and the retreat of collapse areas reinforcing one another. The gully headwall retreat process was driven by the on-wall flow from the upstream (about 15% of the upstream runoff), which was also the main carrier of sediment transport for gully headcut erosion; while the jet flow (about 85% of the upstream runoff) mainly influenced the characteristics of the runoff generation and contributed to the transport of sediment. A small part of jet flow separated from the top of the cave areas (about 10% of the upstream runoff) led to slight erosion to the gully headwall.

(iv) Mechanical mechanisms of gully headcut erosion was investigated and a theoretical model was also constructed. Results showed that the driving force of hydraulic erosion of the gully headwall was the sliding friction force of runoff on the soil, and the main factors determining the magnitude of this sliding friction force were the local topographic slope, the adsorptive force of the soil to the runoff, and the runoff discharge. The core driving force of mass movement in the collapse was the spatial force coupling generated by the gravity of the overhanging to the cross section, and the main factors determining whether mass movement occurs or not were the water content of the overhanging, the cohesive force at the fracture surface, and the morphological characteristics of the overhanging. The expansion of the cave area promoteds the mass movement by increasing the horizontal distance from the center of the overhanging to the fracture surface and decreasing the height of the overhanging to increase the tensile force at the top of the fracture surface; and the mass movement promoted the expansion of the cave area by reducing the length of the top of the cave to reduce the runoff infiltration at the gully headwall. The theoretical model constructed based on the mechanical mechanism consisted of a hydraulic erosion simulation module and a mass movement estimation module. A comparison with measured data showed that the model was capable of accurately simulating the gully headcut processes, with a relative error between the simulated retreat distance and the true value of less than 15%, and had the potential to predict the mass movement of the gully headcut erosion.