Applications and analytical methods of ground penetrating radar for soil characterization in a silvopastoral system

Authors: Smith, Harrison; Owens, Phillip; Ashworth, Amanda

Submitted to: Journal of Environmental and Engineering Geophysics
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 8/1/2022
Publication Date: 12/28/2023
Citation: Smith, H.W., Owens, P.R., Ashworth, A.J. 2023. Applications and analytical methods of ground penetrating radar for soil characterization in a silvopastoral system. Journal of Environmental and Engineering Geophysics. 27:4. https://doi.org/10.32389/JEEG22-001.
DOI: https://doi.org/10.32389/JEEG22-001

Interpretive Summary: Ground penetrating radar (GPR) is a technology that can be used to collect information on soil properties and underground features without the need for digging or excavations. This technology has a wide range of potential applications in soil science and agriculture, but relatively few researchers have explored practical methods for using GPR in applied agricultural settings. In this study, we explore applications of some common qualitative and quantitative methods for GPR analysis and characterization of soil conditions in a silvopasture system. We first analyzed GPR results using visual interpretation measure depth to bedrock, clay layers, and other important soil characterizations. Additionally, we used machine learning methods to accurately predict coarse fragments and percent clay content at various depths in soil profiles. Our results demonstrate that information from GPR images was a very good predictor of soil properties like coarse fragments and percent clay content. We conclude that GPR can provide valuable information on soil properties in silvopastoral systems. Data generated using these methods can improve our understanding of soil characteristics and help land managers optimize of soil management in silvopastoral systems.

Technical Abstract: The use of ground penetrating radar (GPR) has grown rapidly in recent years due to substantial increases in computer processing power and advances in GPR methodologies and geophysical image analysis. With this growth, interest in GPR applications for analysis in agricultural settings has also increased. In this study, we explore applications of some common qualitative and quantitative methods for analysis and characterization of subsurface conditions in a silvopasture system. We first analyze GPR results using traditional visual interpretation methods to delineate depth to bedrock, clay layers, and other important soil characterizations. Estimates of depth to bedrock correlated well with values measured in the field (r_s=0.61,p<0.01), and estimates of depth to clay layers were marginally correlated with observed values (r_s=047,p=0.09). Additionally, we extracted features such as amplitude values, instantaneous attributes, and image texture features generated from a Grey-Level Co-occurrence Matrix to train a random forest regression model to predict coarse fragment percentage and percent clay content. The GPR features extracted were found to be excellent predictors of soil coarse fragments, with R2 values approximately 0.81 and root mean square error (RMSE) of 18.82 for test data. Our results demonstrate GPR can provide valuable information on subsurface features in silvopastoral systems. These results also suggest a strong potential for machine learning algorithms in GPR data analytics. Data generated using these methods could be integrated with or used to validate existing digital soil mapping methods and contribute to better understanding of subsurface characteristics for optimized soil management in silvopastoral systems.