Spatial Variability Models And Prediction Analysis Of Soil Properties Using Geostatistics

Abstract

Soils are composed of solid, water and air phases whose characteristics are highly variable. The interactions of these phases in the soil matrix can lead to different types of topographical formations and characteristics. Due to the uncertainty and complex interactions among these phases, studies on soils have always been a challenging problem for engineers. These variations and uncertainties make it necessary for engineers to adopt new techniques and methods to analyze soil properties in order to determine or interpret their generalized behaviors and patterns. Existing research in variability analysis tends to focus on the distribution of the soil properties, reliability-based design, and simulation of random fields. Despite an increase in the probabilistic and statistical analysis, many challenges remain in incorporating the spatial variability present in the soil properties into prediction analysis. In this research study, a framework was developed using univariate statistics and randomized random variable theory for analyzing the spatially-varied soil properties. The spatial variability present in the soil properties was modeled using the geostatistical tool, Variograms. The variability models were utilized to interpret the soil properties in three different studies in geotechnical engineering, encompassing natural soils, man-made soils, and natural soils with rich minerals. This research highlights the adaptability of the framework for analyzing the soil properties varying from low-to-high variability.