A probabilistic approach to soil layer and bedrock-level modelling for risk assessment of groundwater drawdown induced land subsidence

Artikel i vetenskaplig tidskrift, 2016

Sub-surface construction in urban areas generally involves drainage of groundwater, which can induce subsidence in soil deposits. Knowledge of where compressible sediments are located and how thick these are is essential for estimating subsidence risk. A probabilistic method for coupled bedrock-level and soil-layer modeling to detect compressible sediments is presented. The method is applied in an area in central Stockholm, where clay is the compressible sediment layer. First, a bedrock-level model was constructed from three sources of information: (a) geotechnical drillings reaching the bedrock; (b) drillings not reaching the bedrock; and (c) mapped bedrock outcrops. Input data for the probabilistic bedrock-level model was generated by a stepwise Kriging procedure. Second, a three layer soil model was constructed, including the following materials: (a) coarse grained post glacial and filling material below the ground surface; (b) glacial and post-glacial clays; and (c) coarse grained glaciofluvial and glacial till deposits above the bedrock. Layer thicknesses were transformed to proportions of the total soil thickness. Since Kriging requires data to be normally distributed, the proportions were transformed from proportions (P) to standard normal quantiles (z). In each iteration of a Monte-Carlo simulation, a spatial distribution of the bedrock level was simulated together with the transformed values for the soil-layer proportions. From the iterations, the probability density of the clay thickness (compressible sediments) at each grid cell was calculated. The results of the case study map the expected value (mean) and the 95th percentile of the probability of compressible sediments at specific locations. The resulting model is geologically realistic and validated through a cross-validation procedure in order to be in good agreement with a reference dataset. The case study showed that the method can efficiently handle large amounts of data and requires little manual adjustment. Moreover, the mapped results can provide useful decision support when planning risk-reducing measures and when communicating with stakeholders. Although this novel method is developed for risk assessment of groundwater drawdown induced subsidence, it is useful for other applications involving spatial soil strata modeling.