华北水利水电大学学报(自然科学版)

【目的】对比基于归一化差值植被指数(NDVI)与结构化植被指数两种覆盖管理因子构建方法对修正通用土壤流失方程(RUSLE)土壤侵蚀估算的影响，揭示三门峡市土壤侵蚀时空格局差异及其驱动机制，为复杂地形区侵蚀评估与水土保持管理提供参考。【方法】基于2000—2023年Landsat卫星影像、降雨、土壤、数字高程模型(DEM)与土地利用等多源数据，分别构建基于NDVI与结构化植被指数的两类植被覆盖与管理因子(C因子)(C_NDVI和C_S)并驱动RUSLE估算侵蚀模数；从时间序列、空间分布与坡度分级对结果进行对比，并采用地理探测器识别主控因子及交互作用。【结果】(1)两种方法均表明研究期内土壤侵蚀总体显著下降，但基于C_NDVI的结果年际波动更大；(2)空间上高值区主要位于黄河沿岸及沟谷切割强烈的地区，台塬与河谷平原以低值为主；(3)侵蚀强度随坡度增大向高等级集聚，≥25°陡坡剧烈侵蚀占比在C_NDVI法下为29.36%～35.28%，在C_S法下为35.30%～47.97%；(4)地理探测器显示坡度为主控因子，植被单因子解释力较弱，但与坡度、降雨交互呈显著增强，且C_S法在交互项上表现出更高的q值。【结论】两类C因子对总体格局与趋势判断一致，但对RUSLE结果的差异主要由陡坡(≥25°)高侵蚀等级结构与年际波动刻画引起；其中C_S法强调植被结构信息、C_NDVI法更强调绿度变化，分别适用于复杂地形的精细分级评估与大范围长时序快速对比分析。

【Objective】 This study reveals the spatiotemporal variations and driving mechanisms of soil erosion in Sanmenxia City by comparing the effects of two cover management factor(C) construction approaches, one based on NDVI and the other on a structured vegetation index on RUSLE soil erosion estimation. It serves as a reference for erosion assessment and soil and water conservation management in complex terrain regions. 【Methods】 Two types of cover-management factors(C) of NDVI(C_NDVI) and structured vegetation index-based(C_S) were developed to drive RUSLE for estimating soil erosion modulus based on multi-source data from 2000 to 2023, including Landsat remote sensing, rainfall, soil, DEM, and land-use datasets. The GeoDetector approach was used to determine the primary regulating elements and their interactions after the findings were compared in terms of time series, geographical distribution, and slope categorization. 【Results】(1) Both methods indicate a significant overall decline in soil erosion during the study period, the results based on C_NDVI exhibit larger interannual fluctuations, whereas those based on C_S are more stable.(2) Spatially, high erosion values are mainly concentrated along the Yellow River and in areas with intense gully incision, while tablelands and river valley plains are dominated by low erosion values.(3) Erosion intensity increasingly aggregates into higher classes with increasing slope. For steep slopes ≥25°, the proportion of severe erosion accounts for 29.36% to 35.28% under the C_NDVI results, compared with 35.30% to 47.97% under the C_S results.(4) According to the GeoDetector data, the main regulating element is slope. While vegetation alone has comparatively little explanatory power, its interactions with precipitation and slope reveal a strong enhancing impact. Furthermore, larger q-values for the interaction terms are shown in the C_S data. 【Conclusion】Disparities in the RUSLE results are mostly caused by the two C-factor formulations' differing descriptions of the high-erosion class structure and interannual variability on steep slopes(≥25°), but they both produce consistent interpretations of the overall geographical pattern and temporal trends. C_NDVI emphasizes changes in greenness, while Cs gives more weight to vegetation structure information. As a result, whereas the latter is more suited for quick, extensive comparisons across lengthy time series, the former is more appropriate for fine-grained classification and evaluation in difficult terrain