Soil hydraulic properties are important for studying Earth science. The pore-solid fractal (PSF) model, combined with a critical path analysis from percolation theory, seems to be more promising in the modelling of soil hydraulic properties. The accuracy of the PSF model depends on the accurate acquisition of fractal dimensions, which requires the combination of micro-CT scanning and image analysis technology. In addition, there is a changepoint in soil water movement due to the coexistence of soil micro- and macromorphology. Determining the changepoint and using different fractal dimensions to predict hydraulic properties on different sides of the changepoint can further improve the accuracy of the PSF model. Therefore, in this study, we tested the changepoint in soil water movement and adopted an improved PSF model to predict hydraulic parameters in saline soil based on image analysis. The results showed that the two-sample t-test could identify the changepoint accurately. There was only one changepoint in coastal saline soil when predicting hydraulic properties. Micro-CT scanning and image analysis can obtain fractal dimensions more accurately and quickly. The coefficients of determination of all treatments were above 0.9. The improved PSF model was more accurate than the previous model in predicting soil hydraulic properties. A comparison of goodness-of-fit criteria showed that it is better to adopt the geometrical mean error ratio ($\mathit{\text{GMER}}$) and geometrical standard deviation error ratio ($\mathit{\text{GSDER}}$) as the judgement standard. Due to the anisotropy of soil, the improved PSF model demonstrated a higher accuracy in predicting water content than hydraulic conductivity. The hydraulic conductivity prediction accuracy was negatively correlated with the degree of anisotropy ($\mathit{DA}$) parameter, and the improved model was more suitable for soils with weak anisotropy. Our research can provide a simple and accurate method for parameter calculation of the PSF model to predict soil hydraulic properties more accurately.

中文翻译：

---

通过图像分析使用改进的孔隙固体分形 (PSF) 模型对土壤水力特性进行建模

土壤水力特性对于研究地球科学很重要。孔隙-固体分形 (PSF) 模型与来自渗流理论的关键路径分析相结合，似乎在模拟土壤水力特性方面更有前景。PSF模型的准确性取决于分形维数的准确获取，这需要结合显微CT扫描和图像分析技术。此外，由于土壤微观形态和宏观形态并存，土壤水分运动存在变化点。确定变化点并使用不同的分形维数来预测变化点不同侧的水力特性可以进一步提高PSF模型的准确性。因此，在本研究中，我们测试了土壤水分运动的变化点，并采用改进的 PSF 模型基于图像分析预测盐渍土的水力参数。结果表明，两样本t -test 可以准确地识别变化点。在预测水力特性时，沿海盐渍土只有一个变化点。Micro-CT扫描和图像分析可以更准确、更快速地获得分形维数。所有处理的决定系数均在0.9以上。改进的 PSF 模型在预测土壤水力特性方面比以前的模型更准确。拟合优度标准的比较表明，最好采用几何平均误差比 ($\mathit{\text{GMER}}$) 和几何标准偏差误差比 ($\mathit{\text{GSDER}}$) 作为判断标准。由于土壤的各向异性，改进的 PSF 模型在预测含水量方面表现出比水力传导率更高的准确性。导水率预测精度与各向异性程度呈负相关（$\mathit{达}$) 参数，改进后的模型更适用于各向异性较弱的土壤。我们的研究可以为PSF模型的参数计算提供一种简单而准确的方法，从而更准确地预测土壤水力特性。