PROBABILISTIC CHARACTERIZATION OF MACAU MARINE DEPOSITS BASED ON CONE PENETRATION TEST (CPT) by Ip Tak Long Thesis Supervisor: Dr. Ng, Iok Tong Master of Science in Civil Engineering In geotechnical engineering practice, it is common to determine the soil properties of a specific site based on in situ tests from a number of soundings, as well as laboratory tests performed on a limited number of field samples obtained from a number of boreholes among different in-situ tests, cone penetration test (CPT) is most commonly used for geotechnical soil characterization because it is simple, economical and gives continuous readings with depth showing obvious changes in different soil layers. In geotechnical engineering, there are numbers of uncertainties such as inherent soil variability, measurement errors and transformation uncertainty. It is well-known that these uncertainties have significant effect on the design of the geotechnical structures and needed to be identified and quantified. This study concentrates on the study of soil variability issues in which the random field parameters and the statistically homogeneity of soil layers are obtained and identified in order to achieve this purpose, a numbers of CPT results of Macau marine deposits from Taipa, Macau are collected for statistical analysis in order to produce vertical random field model parameters as well as identifying statistically homogeneity of soil layers of the marine deposits based on the corrected cone tip resistance. The vertical random field model parameters and statistically homogeneity soil layers are determined based on random field theory and modified Bartlett’s test statistic respectively. An approach by combining Robertson chart and Bartlett statistics is proposed for better identifying the statistically homogeneous soil layers in a more reasonable manner. The results of this study indicate that linear trend and the cosine exponential model are most suitable for random field modeling of Macau marine deposits. It is found that more than 80% of results pass the modified Bartlett test for clay layers, with scale of flucutation and coefficient of variation ranged from 98.4 to 2251.8 mm and 30.76% to 79.85% respectively. Meanwhile, it is also found that the silty clay to clay, silty clay to clayey silt, sandy silt to silt sand layers exhibits greater inherent spatial variability than clay layers, with vertical scale fluctuation ranged from 14.3 to 2602.9 mm and the coefficient of variation ranged from 1.77% to 94.25% respectively. In addition, the modified Bartlett test produces results of stationary and non-stationary layers. However, the Bartlett statistic is sensitive to many parameters, such as window length, scale of fluctuation. Furthermore, it is found that there are limitations for the identification of statistically homogeneity of soil layers by using modified Bartlett test such as layer thickness and unknown horizontal variability. In the case that the cone penetration test with sampling interval of 0.02m, the statistically homogeneity of soil layer with thickness less than 0.4m cannot be identified by using modified Bartlett test. It is suggested that further study on the horizontal spatial variability is valuable for obtaining more comprehensive probabilistic characterization of the site.