Document Type : Complete scientific research article
Authors
1
PhD student of soil science department Faculty of Agriculture Shahi Chamran University of Ahvaz
2
Department of soil sciences, faculty of Agriculture, Shahid Chamran University, Ahvaz, Iran.
3
Department of Soil Science Aarhus University of Denmark
10.22069/jwsc.2026.24464.3846
Abstract
Background and objectives: Soil structure, particularly the three-dimensional pore network, plays a fundamental role in hydro-physical processes, gas exchange, water infiltration, and root growth, and directly influences the sustainable functioning of agricultural soils. Soil compaction induced by agricultural machinery traffic can alter pore size, shape, and connectivity, thereby degrading soil functionality through reduced permeability, increased bulk density, and restricted root development. Despite the importance of this phenomenon, conventional physical methods for measuring soil compaction and porosity are unable to accurately capture the morphological and three-dimensional characteristics of the pore network. X-ray computed tomography (X-ray CT) is a non-destructive technique that enables visualization and quantitative analysis of the internal soil structure at the micrometer scale; however, the accuracy of extracted structural indices strongly depends on image quality, preprocessing procedures, and the selection of an appropriate thresholding method. The objective of this study was to compare different thresholding approaches for the processing and analysis of soil CT images and to extract quantitative information on soil structure under compaction induced by agricultural machinery traffic using the ImageJ/Fiji software.
Materials and Methods
This study was conducted in a sugarcane field in Khuzestan Province, Iran, covering approximately 25 ha with a clay loam soil. Soil sampling was performed during the third ratoon crop (R3) at two stages: before and after harvesting. Three soil conditions were investigated: before compaction (BH), compaction induced by machinery traffic at 75% loading capacity (AH-75%), and compaction induced by machinery traffic at 100% loading capacity (AH-100%). Undisturbed soil samples were collected from the 0–25 cm depth using 3-inch polyethylene tubes from nine soil profiles in each treatment and prepared for imaging. Three-dimensional imaging was carried out using an X-ray Micro-CT scanner, and reconstructed images were obtained at a spatial resolution of 60 μm. Image processing was performed using ImageJ/Fiji and included region of interest (ROI) selection, conversion of 16-bit images to 8-bit grayscale, image preprocessing, and noise reduction. To separate pore space from the soil matrix, nine thresholding methods—Bernsen, Contrast, Mean, Median, MidGrey, Niblack, Otsu, Phansalkar, and Sauvola—were applied. Quantitative pore indices, including the number of pores, mean pore size, and resolvable porosity, were subsequently extracted. The most appropriate thresholding method was selected based on a combination of quantitative indices, visual evaluation, structural consistency, and Principal Component Analysis (PCA) to identify the method that best represented the actual soil pore structure.
Results
To evaluate the effects of different thresholding methods on the quantitative analysis of soil X-ray computed tomography (CT) images, datasets from three soil compaction treatments (BH, AH-75%, and AH-100%) were processed using nine commonly applied thresholding algorithms across the entire three-dimensional image volume. All analyses were based on 1,000 consecutive slices per sample to ensure the spatial stability of results at the three-dimensional scale. The results demonstrated substantial differences in the performance of thresholding algorithms in separating the solid phase from pore space, indicating that inappropriate method selection can lead to either overestimation or underestimation of soil pore structure. Comparisons of quantitative indices, including pore number, mean pore size, and total porosity, revealed that methods such as Bernsen, Contrast, MidGrey, and particularly Niblack markedly overestimated soil porosity by identifying an excessively high number of fine pores. This behavior was largely attributed to the misinterpretation of image noise and local grayscale intensity fluctuations as real pores. In contrast, algorithms such as Sauvola and Median exhibited overly conservative behavior by extensively removing fine pores and artificially merging adjacent pores, resulting in a severe underestimation of soil pore structure. These contrasting responses highlight the limitations of certain thresholding algorithms in handling grayscale heterogeneity in soil CT images, especially under compacted soil conditions. Among the evaluated methods, Phansalkar thresholding demonstrated a balanced and stable performance, effectively maintaining an appropriate trade-off between accurate pore detection, suppression of artificial pore generation, and preservation of natural phase boundaries. The porosity values and pore size distributions obtained using this method were consistent with the physical ranges reported for clay loam soils, and the spatial stability of its results across the full three-dimensional image volume was confirmed. Principal component analysis (PCA), employed as a descriptive multivariate tool, further supported this balance by positioning the Phansalkar method in an intermediate and well-balanced region of the multivariate space. Overall, the findings indicate that the selection of an appropriate thresholding method should be based on the combined evaluation of multiple structural indices, spatial stability, and physical realism. Based on these criteria, the Phansalkar method is identified as the most suitable thresholding approach for quantitative analysis of soil CT images.
Conclusion
In this study, the effects of different thresholding methods on the quantitative analysis of soil X-ray computed tomography (CT) images were evaluated. The results demonstrated that the choice of thresholding algorithm plays a decisive role in image segmentation quality and the accuracy of extracted structural indices, and may lead to either overestimation or underestimation of soil pore structure. Simultaneous comparison of key indices, including pore number, mean pore size, and total porosity, revealed that the algorithms respond differently to grayscale heterogeneity and image noise; some methods misinterpret noise as real pores, whereas others conservatively estimate pore structure by removing fine pores. Among the evaluated methods, the Phansalkar algorithm exhibited a balanced and stable performance, effectively combining accurate detection of real pores with noise suppression and preservation of natural phase boundaries. Accordingly, this method is identified as the most suitable option for quantitative analysis of soil micro-CT images. This study emphasizes that thresholding method selection should be based on the combined evaluation of multiple structural indices and the assessment of result stability across the entire three-dimensional image volume.
Keywords
Subjects