The K‐factor of the universal soil loss equation is a core component in many erosion models, as a measure of soil erodibility. It can be estimated by a nomograph, where the summed fractions of silt and very fine sand (VFS) are basic inputs. Frequently, only the three broad particle‐size classes of sand, silt, and clay are measured in laboratories; thus, the VFS fraction must be estimated. Three models are currently available for this estimation, namely, (a) the Revised Universal Soil Loss Equation formula, (b) the European Soil Data Centre method, and (c) the Shirazi–Boersma theory, all three use just the sand fraction as explanatory variable. Nevertheless, their accuracy has never been assessed, and this is the main purpose of this study. The data used to test the VFS estimation methods were drawn from the National Cooperative Soil Survey Soil Characterization Database, incorporating data from more than 300,000 soil horizon samples. The test results show a poor performance of the models, all of which were found to be unsuitable for 31.1% of the textural triangle, accounting for 32.3% of the soil samples. Moreover, it is demonstrated that any conceivable model based solely on the broad particle‐size classes would suffer from a high degree of uncertainty. Consequently, the number of explanatory variables should be increased in order to improve the performance of models. An alternative prediction chart is provided for the first approximation of K‐factor, based on the textural triangle.

中文翻译：

---

估算非常细的砂分数以计算土壤可蚀性K因子

该ķ通用土壤流失方程的因子是许多侵蚀模型的核心组成部分，作为土壤易蚀性的度量。可以通过诺模图来估计，其中淤泥和超细砂（VFS）的总和是基本输入。通常，在实验室中只能测量三种粒径范围很广的沙子，淤泥和粘土。因此，必须估计VFS分数。目前有3种模型可用于此估算，即（a）修订后的通用土壤流失方程式，（b）欧洲土壤数据中心方法，以及（c）Shirazi-Boersma理论，所有这三种模型都仅使用砂分作为解释变量。然而，它们的准确性尚未得到评估，这是本研究的主要目的。用于检验VFS估算方法的数据来自国家土壤合作调查土壤特征数据库，其中包含了300,000多个土壤层样本的数据。测试结果表明，模型的性能较差，所有模型均不适用于结构三角形的31.1％，占土壤样品的32.3％。此外，事实证明，任何仅基于宽粒度类别的可想模型都会遭受高度不确定性的困扰。因此，应增加解释变量的数量，以改善模型的性能。提供了一个替代的预测图，用于的第一近似 发现所有这些都不适合纹理三角的31.1％，占土壤样品的32.3％。此外，事实证明，任何仅基于宽粒度类别的可想模型都会遭受高度不确定性的困扰。因此，应增加解释变量的数量，以改善模型的性能。提供了一个替代的预测图，用于的第一近似 发现所有这些都不适合纹理三角的31.1％，占土壤样品的32.3％。此外，事实证明，任何仅基于宽粒度类别的可想模型都会遭受高度不确定性的困扰。因此，应增加解释变量的数量，以改善模型的性能。提供了一个替代的预测图，用于的第一近似K因子，基于纹理三角形。