Abstract:
The Waterview Connection Project involved the construction of 4.8 km of two 3-lane tunnels (one in each direction) beneath Auckland’s western suburbs, linking two existing state highways to complete a motorway ring route around the city. At each end of the tunnel, approach trenches were also excavated. Prior to the development of the Waterview tunnel, the majority of large road tunnels were constructed as undrained structures to reduce the risk of consolidation settlement. However, due to the long construction period of two years, it was identified that the Southern Approach Trench (SAT) could be fully drained and the Northern Approach Trench (NAT) partially drained. Dewatering approaches were used to lower the groundwater table at the Nat and SAT, and moderate settlement was predicated to ensue at each site. However, from surface monitoring since excavation and completion, actual groundwater drawdown-induced settlement has been only a fraction of that predicted by pre-excavation modelling. Hence, this apparent disparity between predicated and actual groundwater drawdown-induced settlement at both the SAT and NAT is of interest to future tunnel development in the Auckland region, and is the focus of this project. The SAT and NAT sites were conservatively calculated to experience a settlement of 50-100 millimetres upon drainage after excavation. However, despite more than three years of drawdown, actual settlement stabilised at c. 10 mm. One issue could be that a possible bias exists from in situ soil sampling and deriving strengths from laboratory tests in ‘undisturbed’ soils - this may be responsible for producing calculated settlement rates of up to five times what is actually observed in drained soils. Here, Rocscience’s Settle 3D software was used to apply soil parameters from laboratory and in situ testing to undertake a back-analysis of settlement in drained soil at both sites. The aim was to develop a more complete understanding of groundwater flow, subsurface drainage, and settlement than previous 1D and 2D analyses have allowed. Results show that the conservative settlement estimates stem from design parameters falling in the lower bound of sample geotechnical data. In particular, the stiffness parameter Young’s Modulus had a significant effect on predicted settlement, and here exerts a strong control on modelled settlement. Sensitivity analyses show that selecting mean stiffness values led to modelled settlement reflecting actual observed settlement. Nevertheless, other factors such as the highly variable shallow geology and discontinuities across both sites, limited monitoring data, seasonality of ground surface movement, and expansive soils are important. Such factors need to be accurately delineated if predicted and observed settlement patterns are to be reconciled. This will be of particular importance in Auckland, where ongoing and planned underground transport infrastructure excavation is occurring within the context of a highly complex geological history.