基于高密度电阻率法的农田土壤表面干缩裂隙成像

Imaging dry shrinkage cracks on farmland soil surface using electrical resistivity technology

为探测土壤表面干缩裂隙的发育情况,该研究以农田土壤表面干缩裂隙作为研究对象,以高密度电阻率法为测量手段,采用耦合层次聚类分析的同步连续线性估计算法(Simultaneous successive linear estimation,Sim SLE)反演土壤表面干缩裂隙的形态,结合试验裂隙图像和反演得到的模拟图像,对比分析裂隙形态的成像结果。研究表明:基于SimSLE算法反演得到的模拟裂隙图像,可以准确捕捉到裂隙的形状和位置,试验与模拟图像的面积、长度及欧拉数密度一致性指标均大于0.9,决定系数R^(2)都大于0.6,均方根误差数值较小。引入了层次聚类分析后,通过不断更新分区均值和边界,能够更准确地得到农田土壤干缩裂隙的形态特征,分区后模拟结果较好。引入聚类分区后Minkowski密度的决定系数R^(2)提高了9.76%~18.5%,一致性指标提高了0.93%~1.47%,均方根误差降低了12.1%~21.1%,有效提高了模拟精度。研究可为探究非侵入性探测裂隙的形态特征提供算法参考。

This study aims to detect the development of dry-shrinkage cracks on the farmland soil surface using Electrical Resistivity Technology(ERT).The Simultaneous Successive Linear Estimation(Sim SLE)coupled with the hierarchical clustering was also used to invert the morphology of the dry-shrinkage cracks.The experimental and simulated images were then compared to analyze the crack morphology.It was found that the resistance of the soil around the cracks increased sharply,where there were some cracks in the soil,while the resistance of the soil without cracks still maintained a slowly increasing trend.This phenomenon was set as a reference for the formation of cracks.After that,the estimated resistivity was obtained by inversion using the SimSLE algorithm,further to represent the development state of soil shrinkage cracks.The evaluation of fracture images between the natural cracks and the inversion was made using the geometrical properties of Minkowski numbers Mk for the quantitative analysis,where Mk was used to describe the structure of fissures.Three parameters were utilized for the Minkowski number of fissures on the soil surface:the area A,the length L,and the Euler number E of the fissure,representing the fissure area,fissure length,and connectivity in the two-dimensional fissure structure,respectively.The Minkowski density of the fracture images and the simulated binary networks were quantified by the open-source code realized by MATLAB software.Three indicators were selected to evaluate,including the coefficient of determination(R^(2)),the consistency index(IA),and the Root Mean Square Error(RMSE).The results show that the simulated images presented accurate to capture the shape and position of the fracture.The area,length,and Euler number density IA of the test and simulation images were all greater than 0.9,and the R^(2) values were greater than 0.6,but the RMSE values were small.The morphological characteristics of dry-shrinkage cracks in farmland soil were more accurately to continuously update the partition mean and boundary after coupling hierarchical clustering.The simulation images were better after clustering and partitioning.The partition of hierarchical clustering from the resistivity was estimated to obtain via the inversion and calculation of the uniform resistivity as the initial input values in the traditional SimSLE algorithm.The Euclidean distance criterion was used to obtain the optimal number of clusters for the dataset of estimated values.Then,the mean of the estimates for each region was calculated to adjust the covariance matrix.Finally,the updated mean and covariance were used as the prior information for the next round of iterative calculation.As such,the computational accuracy was improved to identify the small-scale spatial variability within soil layers.After the introduction of clustering partition,the R^(2) and IA of Minkowski density increased by 9.76%-18.5%,and 0.93%-1.47%,respectively,whereas,the RMSE was reduced by 12.1%-21.1%,indicating an improved accuracy of the prediction.Therefore,effective detection of soil fissures can greatly contribute to preventing the occurrence of cracks from the damaged integrity of the soil structure,even various engineering geological disasters.The finding can provide an ideal way to characterize the fractures morphologies using non-invasive detection.