Increasingly, tunnels are being planned in weak rock beneath urban areas in Australia and worldwide. Designers of these projects are required to predict the tunnel induced settlement so that an assessment of the possible effects on buildings and services above can be made. However, the existing methods for making such predictions require either a database of similar projects, or a knowledge of suitable parameters for use in numerical models. Numerical models of tunnelling are sensitive to model inputs. Some of these model inputs are irrelevant to the actual physical problem. The thesis discusses two key issues in solving elastoplastic tunnel problems, one being the inertial effects in finite difference methods, and the other being the mesh discretisation error. The causes and possible remedies for these is discussed, and an a priori predictive technique for discretisation error is developed. The method of choice for selecting parameters in numerical models is to use the backanalysis of a suitable pilot activity. However, traditional backanalysis methods suffer from non-uniqueness. The results of a backanalysis is not a single combination of parameters, but a zone in parameter space within which the fit to the observations is acceptable. The shape of this zone in parameter space is affected by the degree of correlation of the parameters, and by the non-linear nature of the problem. The technique of predictive analysis solves the problem of non-uniqueness in parameter estimation by backanalysis by choosing the point in parameter space which gives the most adverse prediction for the future construction, yet which maintains a reasonable fit to the pilot activity. The thesis uses three reference projects in Australia to show the techniques of backanalysis and predictive analysis. Cross sections of the Brisbane Rail Tunnels are backanalysed, and the resulting range in parameter space is discussed. For prediction of settlement in the New Southern Railway tunnel in Sydney, and the South East Transit project in Brisbane, the weak rock pressuremeter is proposed as a suitable and inexpensive pilot activity for the future tunnel construction. The results of the predictive analysis show that the method works. Therefore an improved method of predicting tunnel induced settlement in shallow tunnels is proposed, using the combination of numerical modelling, the weak rock pressuremeter and predictive analysis.