9The strength of rocks in the subsurface is critically important across the geosciences, with 10 implications for fluid flow, mineralization, seismicity, and the deep biosphere. Most studies 11of porous rock strength consider the scalar quantity of porosity, in which strength shows a 12 broadly inverse relationship with total porosity, but pore shape is not explicitly defined. Here

Glacigenic and fluvial deposits of variable lithological composition underlie many major cities in Europe and North America. Traditional geological mapping and 3D modelling techniques rarely capture this complexity as they use lithostratigraphic designations which are commonly based on genesis and age rather than lithological compositions.In urban areas, thousands of boreholes have been, and continue to be, drilled to facilitate the planning, design and construction of buildings and infrastructure. While these data may provide the basis for geological maps and 3D models based on lithological interpretation, they are too numerous for manual correlation to be undertaken efficiently. In this paper we explore the application of largely automated stochastic modelling techniques to develop predictive lithology models for glacial and fluvial deposits in the city of Glasgow, UK. These techniques are commonly used to assess facies variation in oilfield models and are applied here in an urban setting using over 4000 borehole records.Predictions derived from these methods have been evaluated by removing control data and rerunning the simulations. We demonstrate a moderate improvement in the prediction of lithology when using a lithologically-derived stochastic model compared with a conventionally interpolated lithostratigraphic model. It is possible to report uncertainty within the resulting models, either with probability maps or through a suite of plausible simulations of the lithologies across the study region. IntroductionThe growth and decay of high-and mid-latitude Pleistocene ice sheets has left 8% of the Earth's land surface, including one third of Europe and a quarter of North America, covered by glacigenic and fluvial deposits Gibbard, 2004a, 2004b). These deposits underlie many major cities and much of their associated infrastructure networks, and exert a significant influence on the groundwater system. Increasing urban development, and its demands (e.g. suitable foundation conditions, the need for waste storage, contaminant migration, drainage re-routing) requires that information about subsurface glacial deposits, which are often highly lithologically variable across short distances, is available for those involved in planning and construction (Campbell et al., 2010). A key challenge for the three-dimensional (3D) geological modelling community is therefore to represent these subsurface deposits in appropriate ways across large, city-wide areas (Culshaw, 2005;MacCormack et al., 2005;Kessler et al., 2009).In Glasgow, west central Scotland (Figure 1), the British Geological Survey (BGS), in partnership with Glasgow City Council and other local authorities, have used extensive borehole datasets to develop and successfully apply a suite of 3D Quaternary lithostratigraphic models (Merritt et al., 2007;Campbell et al., 2010) (Figure 2). A key strength of lithostratigraphic modelling is that it brings together the expertise of geologists and known geological relationships, enabling a geologically realistic r...

11Estimating the permeability of superficial deposits is fundamental to many aspects of catchment 12 science, but can be problematic where insufficient in situ measurements are available from pumping 13 tests in piezometers. Consequently, common practice is to estimate permeability from the material 14 description or, where available, particle size distribution using a formula such as Hazen. In this 15 study, we examine the relationships between particle size, relative density and hydraulic 16 conductivity in superficial deposits in Morayshire, Northern Scotland: a heterogeneous environment 17 typical of many catchments subject to previous glaciations. The superficial deposits comprise 18 glaciofluvial sands and gravels, glacial tills and moraines, raised marine sediments, and blown sands. 19Thirty-eight sites were investigated: hydraulic conductivity measurements were made using 20 repeated Guelph Permeameter measurements, cone resistance was measured in situ with a Panda 21 dynamic cone penetrometer; material descriptions were made in accordance with BS5930:1999; and 22 disturbed samples were taken for particle size analysis. Overall hydraulic conductivity ( independent predictors of log K and together gave a relationship with an R 2 of 0.80. Material 28 description using the largest fraction (e.g. sand or gravel) had little predictive power. Therefore, in 29 heterogeneous catchments, the permeability of superficial deposits is most strongly related to the 30 finest fraction (d10) and relative density of the material.In situ Guelph permeameter 31 measurements at outcrops with good geological characterisation provide an easy and reliable 32 method of determining the permeability of particular units of superficial deposits. 33

A new lithostratigraphical framework for Singapore is proposed, based on the analysis of c. 20,000 m of core recovered from 121 c. 205 m deep boreholes and augmented with 218 field localities from across Singapore. The new framework describes a succession dating from the Carboniferous to the Quaternary. New U-Pb detrital zircon dates and fossil analysis were used to constrain the ages of key sedimentary units. The oldest known sedimentary rocks in Singapore are found to be the deformed Carboniferous (Mississippian) Sajahat Formation. These are succeeded by the newly erected, Middle and Upper Triassic, marine to continental Jurong Group and Sentosa Group successions that accumulated in the southern part of the Semantan Basin. The Jurong Group comprises four formations: the Tuas Formation, the Pulau Ayer Chawan Formation, the Pandan Formation and the Boon Lay Formation. The Sentosa Group contains two formations: the Tanjong Rimau Formation and the Fort Siloso Formation. In Singapore, the depositional record during this time is related to late Permian to Triassic arc magmatism in the southern part of the forearc basin to the Sukhothai Arc. The Jurong and Sentosa groups were deformed and weakly metamorphosed during the final stages of the Late Triassic to Early Jurassic orogenic event, deformation that led to the formation of the syn-orogenic conglomerates of the Buona Vista Formation. Following this, two distinct Lower Cretaceous sedimentary successions overstepped the Jurong and Sentosa group strata, including the Kusu Formation and the Bukit Batok Formation, both deposited in the southern part of the Tembeling Basin. A series of Neogene to Quaternary formations overly the Mesozoic and Palaeozoic stratigraphy, including the Fort Canning Formation, Bedok Formation and the Kallang Group.

Characterising the 3D distribution of hydraulic conductivity and its variability in the shallow subsurface is fundamental to understanding groundwater behaviour and to developing conceptual and numerical groundwater models to manage the subsurface. However, directly measuring in situ hydraulic conductivity can be difficult and expensive and is rarely carried out with sufficient density in urban environments.In this study we model hydraulic conductivity for 603 sites in the unconsolidated Quaternary deposits underlying Glasgow using Particle Size Distribution (PSD) and density description widely available from geotechnical investigations. Six different models were applied and the MacDonald formula found to be most applicable in this heterogeneous environment, comparing well to the few available in situ hydraulic conductivity data. The range of the calculated hydraulic conductivity values between the 5 th and 95 th percentile was 1.56 x 10 -2 -4.38 m/day with a median of 2.26 x 10 -1 m/day. These modelled hydraulic conductivity data were used to develop a suite of stochastic 3D simulations conditioned to existing 3D representations of lithology. Ten percent of the input data were excluded from the modelling process for use in a split-sample validation test, which demonstrated the effectiveness of this approach compared to non-spatial or lithologically unconstrained models. Our spatial model reduces the Mean Squared Error between the estimated and observed values at the excluded data locations over those predicted using a simple homogenous model by 73%.The resulting 3D hydraulic conductivity model is of a much higher resolution than would have been possible from using only direct measurements, and will improve understanding of groundwater flow in Glasgow and reduce the spatial uncertainty of hydraulic parameters in groundwater process models. The methodology employed could be replicated in other regions where significant volumes of suitable geotechnical and site investigation data are available to predict ground conditions in areas with complex superficial deposits.