Discontinuity data of a 400 m tunnel located close to the proposed shiplock area were investigated to characterize the discontinuity geometry of the rock mass around the tunnel. Traces of 39 major discontinuities (faults and dykes) were found on the tunnel exposures. Over 2000 minor discontinuity (joint) trace data available showed that the rock mass can be separated into about 5 statistically homogeneous regions. Three to four joint sets were found to exist in each of these regions. Available theoretical probability distributions were found to be insufficient to represent the statistical distribution of orientation of joint clusters. For more than 94% of the joint sets, gamma distribution was found to be the best probability distribution for representing a joint size distribution. Exponential and gamma distributions turned out to be the best probability distributions for representing joint spacing distributions for the joint data studied here. Too many censored joint trace data showed a very significant effect of the trace length biases on the estimated mean trace length. This indicated that the 2 m wide exposures which were used for collecting the joint trace data are not sufficient to produce reliable estimates for joint size parameters for the Shiplock area To obtain better estimates, it is necessary to collect joint data for several exposures which are at least 4-5 m wide and have different orientations. When such data become available for the Three Gorges region, validation studies should be performed to check the applicability of the built-up models, in addition to performing the modelling. Rock Mechanics, Daemen & Schultz (eds).