黄土高原是我国乃至世界土壤侵蚀最为严重的地区之一。1999年，我国启动退耕还林（草）工程以来，土地利用格局大幅改变，土壤侵蚀也发生了相应变化。土壤侵蚀模型如中国土壤流失方程（Chinese Soil Loss Equation, CSLE）的大力发展为探讨土地利用变化对小流域坡面侵蚀的影响提供了有力的科学支撑。近年来，监测时间长、覆盖范围大的高分辨率遥感影像的出现，则使深入研究较大时空尺度的切沟分布特征与发育速率成为了可能。选择典型的黄土高原小流域，探讨退耕还林（草）前后的土地利用与土壤侵蚀空间格局演变，对于当地的土地利用规划与水土流失防治具有重要的科学与实践意义。

本研究以陕西省神木市六道沟小流域为研究对象，收集1990、1995、2002、2010和2017年土地利用数据，采用土地利用转移矩阵分析小流域1990—2017年近30年来的土地利用变化；利用中国土壤流失方程（CSLE）剖析小流域1990—2017年的土壤坡面侵蚀空间格局演变；对照实时动态差分（Real-time Kinematic，RTK）GPS的测量结果，验证WorldView-2（WV2）影像对切沟形态参数的提取精度，并在此基础上结合1978年黑白航空影像、2012年与2018年的WorldView-2影像估算1978—2018年小流域的切沟发育速率，并分析切沟空间分布特征及其随时间的变化规律。得到主要结论如下：

（1）1990—2017年，小流域旱地持续减少，灌木林地、农村居民点和独立工矿用地持续增加。以2002年为节点，各土地利用的变化速率呈先缓慢后急剧趋势，土地流转主要发生在耕地、林地和草地之间。

（2）1990—2017年，小流域土壤侵蚀总体呈减弱趋势。年均土壤侵蚀模数从1990年的27.8 t/(hm²·a)减少至2017年的7.2 t/(hm²·a)，减幅74.1%。各土地利用类型中，独立工矿用地侵蚀最严重，旱地次之，草地和灌木林地最小。

（3）WV2解译切沟长度、最大表面宽度、周长和面积具有较高的精度，与RTK GPS实测数据相比，各参数相对误差主要集中在5%以内。1978—2012年集水区切沟总长度、总最大表面宽度、总周长和总面积年均变化率分别为-0.1、-0.03、0.05和0.04 %/a，切沟扩张速率与集水区面积呈显著正相关。2012—2018年集水区切沟各参数变化率对应为-0.02、0.01、0.01和0.2 %/a，切沟扩张速率与集水区平均坡长和平均剖面曲率呈显著正相关，与平均坡度呈负相关，且当植被盖度大于45%时，切沟扩张速率较低。

The Loess Plateau was one of the regions suffering from the most severe soil erosion in China, even across the world. Since the implementation of the Grain for Green Project in 1999, the land use in the Loess Plateau has darmatically modified, resulting in great changes in soil erosion. The improvement of soil erosion models such as the Chinese Soil Loss Equation (CSLE) provides a powerful tool to explore the impact of land use change on soil erosion in the small catchment. Recently, the development of the high-resolution remote sensing images that have large spatial and temporal coverages, makes it possible to systematically examine the spatio-temporal distribution patterns and growth rates of gullies. It is therefore critical to examine the spatio-temporal variations of land use and soil erosion as affected by the Grain for Green Project at a typical catchment in the Loess Plateau, which holds important scientific and practical implications for local land use planning and soil conservation.

The current study was performed in the Liudaogou cathchment of the Loess Plateau, which was located in Shenmu, Shaanxi Province. Land use transition matrices were employed to evaluate the spatio-temporal variations of land use based on the land use data in 1990, 1995, 2002, 2010 and 2017. Using CSLE, the spatial pattern of the slope erosion in the Liudaogou catchment and its evolution with time were analyzed for 1990-2017. The accuracy of the WorldView-2 (WV2) image in gully morphology extraction was evaluated via the comparison with the real-time kinematic (RTK) GPS results. On this basis, the grayscale aerial image taken in 1978, WV2 images in 2012 and 2018 were compared to estimate the gully development rate and to acquire the spatio-temporal distribution patterns of gullies. The main conclusions are presented as below:

(1) During 1990-2017, the land use pattern substantially changed, with the area of dryland persistently decreasing, and the areas of shrubland, rural residential land as well as industrial and mining land continuously increasing. Besides, the land use change has apparently accelerated since the local implementation of the Grain for Green Project in 2002. The transition of land use mainly occurred among cropland, woodland and grassland.

(2) Soil erosion generally weakened in the past 30 years during 1990-2017. The annual soil erosion rate decreased continuously, i.e., from 27.8 t/(hm²·a) in 1990 to 7.2 t/(hm²·a) in 2017, with a decrease rate as high as 74.1%. Among all the land use types, the industrial and mining land suffered from the most serious erosion, followed by the dryland, and the grassland and shrubland successively.

(3) Gullies were accurately delineated using WV2 images, with the mean relative errors of gully length, maximum surface width, perimeter and area mostly smaller than 5%, as compared to the corresponding RTK GPS measurements. During1978-2012, the mean change rates of the total gully length, maximum surface width, perimeter and area in the Liudaogou catchment were -0.1、-0.03、0.05 and 0.04 %/a, respectively, whereas the corresponding change rates during 2012-2018 were -0.02、0.01、0.01 and 0.2 %/a, respectively. According to the correlation analyses, the gully development during 1978-2012 was mainly regulated by drainage area. While for 2012-2018, the gully development was positively affected by the mean slope length and profile curvature of the drainage area, but negatively by the mean slope steepness. The latter, in addition, was also related to the vegetation coverage, i.e., relatively lower for the areas with the vegetation coverage greater than 45%.