Abstract: Due to spatial auto-correlation and inter-correlation among geotechnical observed data, the spatial geotechnical characteristics distribution at unknown location has to be extrapolated or interpolated by some special methods. However, classical statistical methods could not rationally resolve the problems which include selection of sample points, and comparison of spatial estimating weights between bilateral data. The distance-weighted ordinary Kriging (OK), universal Kriging (UK), and co-Kriging (CK) prediction methods for scattered data are introduced at first, which are known as the Kriging family in global geostatistics. Moreover, multi-fractal theory combining with co-Kriging (MK) is presented to depict the local singularity which should be ignored by Kriging of sliding weighted average algorithm. The performance is compared in different typical geotechnical parameters: cohesion coefficient c, friction angle and compression modulus . This study is based on the main axis (525 m long, 80 m wide) of Expo 2010 area of Shanghai, geotechnical test samples come from 42 boreholes. The performance of the different model fitness used in this study is MK, CK, OK and UK from the best to poor for parameters c and , and for parameter simulation, the sequence is MK, CK, UK and OK. The results prove that, in the most geotechnical occasions, auxiliary information would improve the prediction accuracy, and MK theory is useful tool to measure local singularity.

Key words: ordinary Kriging(OK), universal Kriging(UK), co-Kriging(CK), multi-fractal(MK), random field