Stochastic poromechanical modeling of anthropogenic land subsidence

A key issue in poromechanical modeling, e.g. for predicting anthropogenic land subsidence due to fluid withdrawal, is the evaluation and use of representative mechanical properties for the deforming porous medium at a regional scale. One such property is the vertical uniaxial rock compressibility cM which can be obtained through either laboratory oedometer tests or in situ measurements, and typically exhibits quite a marked scattering. This paper addresses the influence of the cM uncertainty on the predicted land settlement using a stochastic simulation approach where cM is regarded as a random variable and a large number of equally likely cM realizations are generated and implemented into a poroelastic finite element model. A compressibility law, characterized by a log-normal distribution with depth-dependent mean, constant variance and exponential covariance, is assumed. The Monte Carlo simulation provides a set of responses which can be analyzed statistically. The results from a number of numerical experiments show how the cM variance and covariance affect the reliability of the simulated land subsidence and provide a quantitative evaluation of the intrinsic uncertainty of the model prediction