The Normalized Difference Vegetation Index (NDVI), has been increasingly used to capture spatiotemporal variations in cover factor (C) determination for erosion prediction on a larger landscape scale. However, NDVI-based C factor (Cₙdᵥᵢ) estimation per se is sensitive to various biophysical variables, such as soil condition, topographic features, and vegetation phenology. As a result, Cₙdᵥᵢ often results in incorrect values that affect the quality of soil erosion prediction. The aim of this study is to multi-temporally estimate Cₙdᵥᵢ values and compare the values with those of literature values (Cₗᵢₜ) in order to quantify discrepancies between C values obtained via NDVI and empirical-based methods. A further aim is to quantify the effect of biophysical variables such as slope shape, erodibility, and crop growth stage variation on Cₙdᵥᵢ and soil erosion prediction on an agricultural landscape scale. Multi-temporal Landsat 7, Landsat 8, and Sentinel 2 data, from 2013 to 2016, were used in combination with high resolution agricultural land use data of the Integrated Administrative and Control System, from the Uckermark district of north-eastern Germany. Correlations between Cₙdᵥᵢ and Cₗᵢₜ improved in data from spring and summer seasons (up to r = 0.93); nonetheless, the Cₙdᵥᵢ values were generally higher compared with Cₗᵢₜ values. Consequently, modelling erosion using Cₙdᵥᵢ resulted in two times higher rates than modelling with Cₗᵢₜ. The Cₙdᵥᵢ values were found to be sensitive to soil erodibility condition and slope shape of the landscape. Higher erodibility condition was associated with higher Cₙdᵥᵢ values. Spring and summer taken images showed significant sensitivity to heterogeneous soil condition. The Cₙdᵥᵢ estimation also showed varying sensitivity to slope shape variation; values on convex-shaped slopes were higher compared with flat slopes. Quantifying the sensitivity of Cₙdᵥᵢ values to biophysical variables may help improve capturing spatiotemporal variability of C factor values in similar landscapes and conditions.

中文翻译：

---

利用星载地球观测数据量化基于NDVI的C因子估算的敏感性和潜在的土壤侵蚀预测

归一化植被指数（NDVI）已越来越多地用于捕获覆盖因子（C）确定中的时空变化，以便在较大的景观尺度上进行侵蚀预测。但是，基于NDVI的C因子（Cₙdᵥᵢ）估计本身对各种生物物理变量敏感，例如土壤条件，地形特征和植被物候学。结果，Cₙdᵥᵢ通常导致不正确的值，从而影响土壤侵蚀预测的质量。这项研究的目的是对Cestimatedᵥᵢ值进行多时估计，并将其与文献值（Cₗᵢₜ）进行比较，以便量化通过NDVI和基于经验的方法获得的C值之间的差异。另一个目标是量化生物物理变量（例如斜坡形状，易蚀性，Cₙdᵥᵢ的作物生长阶段变化和农业景观尺度的水土流失预测。将2013年至2016年的多时态Landsat 7，Landsat 8和Sentinel 2数据与来自德国东北部Uckermark地区的综合管理和控制系统的高分辨率农业用地数据结合使用。春季和夏季的数据中Cₙdᵥᵢ和Cₗᵢₜ之间的相关性得到改善（最高r = 0.93）；但是，Cₙdᵥᵢ值通常比Cₗᵢₜd值高。因此，使用Cₙdᵥᵢ进行侵蚀建模的速率比使用Cₗᵢₜdₙ进行建模的速率高两倍。发现Cₙdᵥᵢ值对土壤易蚀性条件和景观的坡度形状敏感。较高的腐蚀条件与较高的Cₙdᵥᵢ值相关。春季和夏季拍摄的图像显示出对异质土壤状况的显着敏感性。Cₙdᵥᵢ估计值也显示出对边坡形状变化的敏感性变化；凸形坡度的值比平坡度高。量化Cₙdᵥᵢ值对生物物理变量的敏感性可能有助于改善捕获相似景观和条件下C因子值的时空变异性。