We found a recent paper by Curtis Jensen and Michael Jefferies to be particularly interesting. Here’s a quick summary on this novel work and how we see its potential practical implications for geotechnical engineering applications.
In an era when geotechnical practice increasingly defaults to computational sophistication, it is worth pausing to appreciate what can be learned from careful, long-term observation of the ground. A recent paper by Jensen and Jefferies (2023) does exactly that — and the lessons are substantial.
The paper revisits the Harbor Bay Isle (HBI) residential development in Alameda, California, reclaimed from the margin of San Francisco Bay beginning in 1967. The soft estuarine clay underlying the site — San Francisco Bay Mud — is among the most compressible soils in the engineering literature, with pre-reclamation void ratios exceeding 2.0, compression indices (Cc) in the range 0.8–1.1, and liquid limits of 85–88%. Around 7 m of hydraulically placed sandfill generated settlements of up to 2.5 m, recorded by settlement plates established at ground level before filling began. Monitoring continued for 43 years, spanning primary consolidation and well into the secondary creep phase — an unusually complete record by any standard.
What makes this study worth reading is not simply its longevity, but what the authors choose to do with it.
Empirical data, seriously used
The analysis is anchored throughout in measured data: in situ void ratios from original boreholes, constant-rate-of-strain oedometer tests, groundwater records from observation wells installed during filling, and vane shear profiles from offshore investigations where Bay Mud had never been desiccated. Rather than treating laboratory data as definitive inputs, the authors use them as constraints on what is physically plausible, iterating numerical predictions against the full settlement record to identify the most probable ground conditions.
Where laboratory-derived pre-consolidation pressures proved inconsistent with observed settlements, the authors interrogated why. Their conclusion is instructive: Bay Mud, despite being classified as a high-plasticity clay, behaves in CPT soundings as a sensitive silt, and is prone to densification during sample extrusion and handling. The result is a systematic overestimation of yield stress in oedometer tests, such that settlements computed from laboratory pc values reached only around three-quarters of those actually measured. The authors argue that pre-consolidation pressure profiles are better estimated from back-analysis of well-documented case histories than from laboratory testing alone — a pointed observation for practitioners working with similar materials.
Equally significant is the finding that pre-consolidation pressure in Bay Mud is not simply a product of desiccation or ageing in the Bjerrum (1967) sense. Analysis of CPT data and vane shear profiles — including from Bay Mud that had never been exposed to the atmosphere — points to a depositional fabric that contributes a residual component of pre-consolidation pressure independent of stress history. The authors express this as pc = 1.1σ’v0 + 50 kPa, with the constant 50 kPa term representing this fabric effect. It is a finding that challenges the standard desiccated-crust narrative that had previously dominated interpretation of the HBI record.
An old case history, revisited with better tools
The HBI settlement record was the subject of a Terzaghi Lecture by Duncan (1993), which identified the limitations of Terzaghi consolidation theory and attributed the marked differences in behaviour between monitoring points largely to the presence or absence of a desiccated surface crust. Jensen and Jefferies (2023) now show, with the benefit of the complete 43-year record, that the crust hypothesis does not hold. Modelling scenarios that incorporate a desiccated crust over nominally normally consolidated Bay Mud consistently over-predict late-time settlements and fail to match observed void ratio profiles. What actually controls the differing settlement histories between monitoring points is the relative proportion of two stratigraphic subunits within the Bay Mud — a softer, looser upper unit and a denser, stiffer lower unit — each with distinct Cc and cv values. The boundary between these subunits appears at a broadly consistent elevation across the site, so that differences in Bay Mud thickness above and below this boundary largely explain the spatial variation in settlement behaviour.
A numerical model that earns its keep
The consolidation modelling is implemented in Excel VBA using large-strain finite difference theory with convected (Lagrangian) coordinates. This matters because vertical strains reached 23% at some locations — well beyond the range where small-strain Terzaghi theory is reliable. The large-strain formulation captures the increased hydraulic gradients that develop as the soil compresses, and the authors note that cv values back-figured from settlement histories using large-strain theory are systematically lower than those inferred from the same oedometer data using small-strain reduction methods, consistent with findings by Jia et al. (2013).
Boundary conditions are honoured in full throughout: time-varying fill thickness including mid-history additions and removals, an exponentially decaying groundwater head within the fill driven by horizontal drainage, and a constant underdrain head at the base of Bay Mud from the underlying Merritt Sand aquifer. The result is close matches to all three settlement histories across 43 years, including sharp inflections in settlement rate caused by modest regrading during residential development — inflections that only a model with accurate loading history could reproduce.
The broader lesson
The takeaway is not that the modelling tool is clever — it is that the tool is honest. Detailed numerical analysis only repays its cost when constrained by equally detailed empirical observation. Regional CPT datasets, carefully preserved borehole logs, and decades of settlement monitoring are not archival curiosities; they are the raw material from which reliable geotechnical judgement is built. This paper is a compelling reminder that case histories, properly maintained and seriously interrogated, remain among the most powerful things in the geotechnical engineer’s toolkit — and that revisiting them with better methods can overturn conclusions that once seemed settled.
References
Bjerrum, L. (1967) Engineering geology of Norwegian normally-consolidated marine clays as related to the settlement of buildings. Géotechnique, 17(2), pp. 83–118.
Duncan, J.M. (1993) Limitations of conventional analysis of consolidation settlement. Journal of Geotechnical Engineering, 119(9), pp. 1333–1359.
Jensen, C. and Jefferies, M. (2023) Retrospective evaluation of large settlements of a very soft clay: Alameda, California. Geotechnical Research, 10(4), pp. 175–197.
Jia, R., Chai, J. and Hino, T. (2013) Interpretation of coefficient of consolidation from CRS test results. Geomechanics & Engineering, 5(1), pp. 57–70.
