Variations in resilient modulus (Mr) values of unbound materials as determined in accordance with AASHTO T307 are studied. Three different granular materials and an elastic polyurethane (control) sample were tested using commercially available laboratory test equipment. A field test bed was also constructed to measure vehicle (Class 3) induced loading conditions at the bottom of an unbound granular layer underlying new hot-mix asphalt pavement to determine stress-pulse duration as a function of vehicle speed. Various interpretation issues were identified within the framework of the testing methods and equipment including: (1) insufficient laboratory sensor sampling rate (per standard); (2) the laboratory specified 0.1 second load-pulse duration and haversine shape are not matching the field stress-pulse duration and shape; (3) the need for careful tuning of the load system gain settings; (4) the number of LVDTs used in the vertical strain calculation; (5) limiting quality control and quality assurance to deformation ratio values in the preconditioning sequence; (6) limiting the load step calculations to the last 5 of 100 load cycles; and (7) the k3 coefficient used in the MEPDG suggested generalized universal model function generally not being statistically significant.

Detailed geostatistical analysis procedures are presented in this study to provide a guide for pavement engineers to study spatial variability of pavement foundation properties with consideration of choosing the best fit semivariogram model and characterization of anisotropy. Measurements from two densely gridded pavement reconstruction sites are presented in studying the geostatistical modelling parameters that characterize spatial variability of stiffness and compaction properties. Preliminary study on anisotropy in spatial variability of pavement foundation properties is performed, but different major and minor anisotropic directions were identified in the two small square study sections. Comparisons of three theoretical semivariogram models (i.e., spherical, exponential, and Matà ©rn with k=1) in studying different pavement foundation properties shows that there is no single best fit model. The isotropic semivariogram model works as well as the anisotropic semivariogram model in estimating the data at unsampled locations across the studied small square area. The range that indicates the spatial correlation length is less than 5 m in all studied properties of both test sections, without considering anisotropy behavior. When anisotropy behavior is considered, longer spatial correlation length, up to about 11 m, can be expected in the major direction.

The second objective of this study is to investigate variability of pavement foundation properties (e.g., ELWD-Z3, γd, and w) that are determined from four major in-situ tests (i.e., FWD, LWD, NG, and DCP) over 18 test sections of 6 project sites. Change in variation of in-situ measured properties has been studied in relationship to the number of compaction passes. Univariate statistics of pavement foundation properties is documented for providing references to pavement engineers and researchers to know the range of variability that in-situ measured properties can have. In addition to univariate statistics, spatial analysis was performed on selected sites that contain relatively large data sets for spatial analysis. The difference in spatial variation can be expected in longitudinal and transverse direction. The correlation length of about 2 m to 3 m in the minor or less uniform direction was quantified for spatial variability of dense gridded data on the base layer. The spatial variability of in-situ measured properties along the longitudinal direction could be expected to be 15 m to 23 m in the CTB layer. This study on spatial variability shows that the correlation length can be different in different pavement foundation layers and materials.