Abstract The hydrological community often turns to widely available spatial datasets such as the NRCS Soil Survey Geographic database (SSURGO) to characterize the spatial variability of soil properties. When used to spatially characterize and parameterize watershed models, this has served as a reasonable first approximation when lacking localized or incomplete soil data. Within agriculture, soil data has been left relatively coarse when compared to numerous other data sources measured. This is because localized soil sampling is both expensive and time intense, thus a need exists in better connecting spatial datasets with ground observations. Given that hydrogeophysics is data-dense, rapid, non-invasive, and relatively easy to adopt, it is a promising technique to help dovetail localized soil sampling with spatially exhaustive datasets. In this work, we utilize two common near surface geophysical methods, cosmic-ray neutron probe and electromagnetic induction, to identify temporally stable spatial patterns of measured geophysical properties in three 65 ha agricultural fields in western Nebraska. This is achieved by repeat geophysical observations of the same study area across a range of wet to dry field conditions in order to evaluate with an empirical orthogonal function. Shallow cores were then extracted within each identified zone and water retention functions were generated in the laboratory. Using EOF patterns as a covariate, we quantify the predictive skill of estimating soil hydraulic properties in areas without measurement using a bootstrap validation analysis. Results indicate that sampling locations informed via repeat hydrogeophysical surveys, required only five cores to reduce the cross-validation root mean squared error by an average of 64% as compared to soil parameters predicted by a commonly used benchmark, SSURGO and ROSETTA. The reduction to five strategically located samples within the 65 ha fields reduces sampling efforts by up to ∼90% as compared to the common practice of soil grid sampling every 1 ha.

中文翻译：

---

使用水文地球物理学和经验正交函数对近地表土壤保水函数进行空间预测

摘要 水文界经常求助于广泛可用的空间数据集，如 NRCS 土壤调查地理数据库 (SSURGO) 来表征土壤特性的空间变异性。当用于空间表征和参数化流域模型时，当缺乏局部或不完整的土壤数据时，这已作为合理的第一近似值。在农业领域，与测量的众多其他数据源相比，土壤数据相对粗糙。这是因为局部土壤采样既昂贵又费时，因此需要更好地将空间数据集与地面观测连接起来。鉴于水文地球物理学是数据密集、快速、非侵入性且相对易于采用的，它是一种很有前途的技术，可以帮助将局部土壤采样与空间详尽的数据集相结合。在这项工作中，我们利用两种常见的近地表地球物理方法，宇宙射线中子探测器和电磁感应，来识别内布拉斯加州西部三个 65 公顷农田的测量地球物理特性的时间稳定空间模式。这是通过在一系列湿地到干地条件下对同一研究区域进行重复地球物理观测来实现的，以便使用经验正交函数进行评估。然后在每个确定的区域内提取浅层岩心，并在实验室中生成保水功能。使用 EOF 模式作为协变量，我们使用引导验证分析量化了在没有测量的情况下估计区域土壤水力特性的预测技能。结果表明，通过重复水文地球物理调查获知的采样位置，与常用基准 SSURGO 和 ROSETTA 预测的土壤参数相比，仅需要五个核心即可将交叉验证的均方根误差平均降低 64%。与每 1 公顷土壤网格采样的常见做法相比，在 65 公顷土地内减少到五个战略位置的样本可减少高达 90% 的采样工作。