The purpose of this study is to improve site investigation in geotechnical engineering via the evaluation and development of statistical approaches for characterizing the spatial variability of soil properties and the development of site investigation simulation software for educational use. This study consists of four components: statistical characteristics, data measurement, simulation, and educational training. Statistical measures of spatial variability of soil properties were examined for three different geographical areas where soil formation processes differ to assess the influence on the spatial variability of soils. Statistical measures of spatial variability were also calculated for a case history where blasting was used as a method of soil improvement to evaluate the effects of man-made changes to soil structure. The concept of spatial aliasing was employed to estimate the maximum allowable sampling interval for field data as a function of the spatial correlation properties. Once a maximum statistically allowable sampling interval is determined for a specific soil property, the minimum statistically required number of soundings / borings is calculated to perform an economical site investigation at a specific site. A simple and efficient simulation technique was proposed to generate correlated, multi-dimensional simulations of soil properties. Based on limited data, the proposed simulation technique generated accurate and correlated simulations of soil properties that are consistent with the observed or proposed correlation structures of soil properties. Lastly, a geotechnical site investigation simulation program with a wide variety of in situ and laboratory tests was developed to allow students to plan and perform a comprehensive site investigation program. The simulation generates an input file based partly on the statistical characteristics of the spatial variability of soil properties analyzed in this study and partly on traditional values. Spatial variability in soil properties is modeled via correlated random fields, interpolation, and a decomposition method to yield realistic geotechnical data. Via the simulation, students are able to obtain experience and judgment in an essential component of geotechnical engineering practice. The four components of this research (statistical characteristics, data measurement, simulation, and educational training) focus on the improvement of site investigation performance in geotechnical engineering, thereby improving reliability analysis in geotechnical practice.