根系与土壤水分的关系被认为是地表生态系统的重要组成部分，研究二者关系对深入理解根系、土壤和地表水循环之间的相互作用至关重要。然而，现有研究受到野外采样方法的限制，存在研究尺度小、分析维度单一、结果可代表性差等问题。因此，本研究试图利用无损的探测方法，建立估计地下根系生物量模型，基于林分尺度，从水平和垂直两个维度探究根系生物量与土壤含水量的关系。

        探地雷达（Ground-Penetrating Radar，GPR）作为一种无损探测技术，被成功用于地下根系的研究。本文首先利用该方法结合剖面挖掘法构建了用于估计小叶锦鸡儿（Caragana microphylla）地下根系生物量的模型，建模数据包含野外采集的32个剖面（5m 长、1m 深）的GPR根点数量和实际剖面根系生物量。基于以上数据，分析模型趋势以确定可能的模型表达式，之后根据最优模型判据（AIC）评价各个模型，最后对选定的模型进行检验。模型检验采用两种方法：留一法交叉验证（LOO-CV）和与现有研究对比，检验结果均证明了模型的有效性。基于构建的模型，本研究分析了同一实验区域两个样方（30 × 30 m²）的根系生物量与土壤含水量的关系。首先，每个样方被划分为5个深度层，每个深度层在水平方向又被划分为3个区域：富根–树冠覆盖区、富根–无树冠覆盖区和少根–无树冠覆盖区。最后，利用广义加性模型（GAMs）分析了垂直和水平方向上的根系生物量与土壤含水量的关系。结果显示，根系生物量与土壤含水量的关系在不同方向存在各向异性。垂直方向上为正相关关系，即根系生物量大的区域土壤含水量相对较大，根系生物量小的区域土壤含水量较小。水平方向上的关系相对复杂。对于样方1的富根–无树冠覆盖区，根系生物量与土壤含水量在中间三个深度层（20–40、40–60和60–80 cm）呈显著的负相关关系。对于样方2的富根–无树冠覆盖区，根系生物量与土壤含水量在中间三个深度层（20–40、40–60和60–80 cm）呈显著的非单调关系，即随着根系生物量的增加，土壤含水量先减小再增大。此外，我们从根系吸水、根系空间分布和优势流补偿作用等角度讨论了根系与土壤水分相互作用关系，分析了这些关系与林分尺度下根水关系空间各向异性的关系。

      本研究构建了利用GPR探测的根点数量估计地下根系生物量的模型，并将该模型成功应用于根系与土壤水分关系研究中，从而揭示了GPR相比于传统方法在野外生态水文学研究中的优势。研究结果补充了干旱–半干旱环境下根系与土壤水分关系的现有知识，为根区复杂的水文过程模拟提供了新思路。

       The relationship between the root system and soil water is considered to be a vital part of the surface ecosystem. Studying the relationship between the root system and soil water is very important to understand the interaction among root system, soil, and the surface water cycle. However, the existing research is limited by the field sampling method, and there are some problems such as small research scale, single analysis dimension and poor representativeness of results, etc. Therefore, this study attempts to use non-destructive detection methods to establish a model for estimating underground root biomass, and study the relationship between the root biomass and soil water content from horizontal and vertical dimensions based on stand scale.

      As a non-destructive detection technology, ground-penetrating radar (GPR) has been successfully used in the study of the underground root system. In this paper, a model for estimating the underground root biomass of Caragana microphylla is constructed by using GPR combined with the profile sampling method. The modeling data include the number of root points detected by GPR and the actual root biomass of 32 profiles (5 m long and 1 m deep) collected in the field. Based on the collected data, we first analyze the trend of the model to determine the possible model expression, then use the optimal model selection criterion (Akaike information criterion, AIC) to evaluate different models, and finally test the selected model. The model is validated by two methods: Leave–One–Out Cross–Validation (LOO–CV) and the comparison with the existing research, and the results of two methods all prove the effectiveness of the selected method. Based on such model, the relationship between root biomass and soil water content of two plots (30 × 30 m²) in the same experimental area was explored. Firstly, each plot was divided into five depth intervals, and each depth interval was divided into three areas in the horizontal direction: root–rich canopy–covered area, root–rich canopy–free area, root–poor canopy–free area. Finally, the generalized additive models (GAMs) were used to model the relationship between root biomass and soil water content in the vertical and horizontal directions. The results showed that the relationship between root biomass and soil water content was anisotropic in different directions. In the vertical direction, there is a positive correlation, that is, the soil water content in the area with large root biomass is relatively large, and the soil water content in the area with small root biomass is relatively small. The relationship in the horizontal direction is relatively complex. For Plot 1, root biomass and soil water content were negatively correlated in the middle three depth intervals (20–40, 40–60, and 60–80 cm). For Plot 2, there was a significant non-monotonic relationship between root biomass and soil water content in the middle three depth intervals (20–40, 40–60, and 60–80 cm), that is, with the increase of root biomass, soil water first decreased and then increased. In addition, we discussed the interaction mechanism between root and soil water from the perspectives of root water uptake, root spatial distribution, and preferential flow compensation, and analyzed the relationship between these mechanisms and the spatial anisotropy of root water relationship at stand scale.

       In this study, we proposed a model to estimate the underground root biomass using the number of root points detected by GPR and successfully applied the model to study the relationship between the root system and soil water to reveal the advantages of GPR in the field of ecological hydrology research compared with traditional methods. The results complement existing knowledge of the relationship between the root system and soil water in an arid and semi-arid environment and provide new sights for the simulation of complex hydrological processes in the root zone.