CHARACTERIZING SOIL TEXTURE USING GEOSTATISTICAL METHODS WITH GEOPHYSICAL DATA

Many agricultural, environmental, and engineering activities require accurate characterization of soil texture over large areas. While point measurements of soil texture are relatively easy to obtain, natural environments often have heterogeneous soil textures, so acquiring enough point measurements to accurately characterize large areas is usually prohibitively expensive. If soil texture is inadequately characterized, inefficient land management or engineering practices may result. Geophysical data have potential to improve soil texture characterization, and these data can be collected very quickly and with high resolution over large areas. However, no geophysical technique is currently capable of independently and non-ambiguously identifying different soil textures. This project investigates the potential of geostatistical analysis of large geophysical data sets and sparse point measurements of soil texture to improve soil texture estimation. For this investigation, ground penetrating radar groundwaves were used to generate high-resolution grids of near-surface soil water content and groundwave amplitude over a heterogeneous field site. Point measurements of soil texture were also collected across the site. Variograms and cross-variograms were calculated for the soil texture, water content, and amplitude measurements, and maps of soil texture were generated using kriging of the soil texture measurements and co-kriging of soil texture and geophysical data. Cross-validation was performed for each estimated soil texture distribution to determine the error in the kriged estimates, and geophysical data were found to significantly improve the accuracy of soil texture estimation. Then, the relationships between the statistical moments of the geophysical data and the soil texture measurements were investigated to determine whether geophysical data could reasonably indicate soil texture without any point measurements. This investigation showed that point measurements of soil texture are necessary for very accurate soil texture estimation, but statistical analysis of geophysical data can indicate the soil texture distribution even without soil texture measurements.