Uncertainty analysis of numerical simulation of groundwater inflow into Safarood Kerman water transfer tunnel

Groundwater inflow is one of the most important problems in Constructing a conveyance tunnel. Increasing pressure on the tunnel wall and reducing its stability, the related issues of drainage and pumping, destructive impacts on the mechanical and geological condition of the tunnel surrounding environment, loss of life, increased costs, and advance delays are among the most important challenges that can be existed during excavation. Therefore, it is crucial to evaluate the amount of water inflow and predict the required measures previously. Conventional techniques for estimating the water inflow are analytical-experimental techniques whose efficiency in complex heterogeneous and anisotropic aquifers is always tainted. Accordingly, this study intends to investigate the effectiveness of the Meshfree (Mfree) numerical method for simulating the groundwater level in the environment surrounding the Safarood water transfer tunnel in Kerman. Also, considering the uncertainty analysis, uncertainty of parameters (hydraulic conductivity), input data, and structure of numerical modeling were addressed using DiffeRential Evolution Adaptive Metropolis (DREAM) algorithm. Hence, an open-source framework based on a Mfree numerical method and DREAM algorithm was proposed for the simulation-optimization process of groundwater level prediction in the environment surrounding the tunnel, and finally, the water inflow discharge was estimated. The results of uncertainty analysis indicated that hydraulic conductivity parameters may be ranged between 0.0002 to 0.2 m/day in different homogeneous zones. Also, the study of thin sections samples collected from field observation shows that hydrothermal conditions have influenced directly the alteration of rocks and minerals in some zones and likely it is the main factor in increasing permeability in these areas. The results showed that the recorded input data has a four percent underestimation. The uncertainty of the parameters involved with the structure of numerical modeling also proved that to obtain an adequate accuracy, the size of the local domain must be about 0.85, and the support domain should be considered at least three nodes to estimate the weight function. The simulation results of groundwater level fluctuations using the derived true values of parameters showed that there is a good accuracy between the observed and simulated values (the RMSE index was estimated to be about 2.531 meters). In addition, the assessment of numerical simulation of groundwater inflow into tunnel indicated that inflow rate in the north and south parts is respectively 72.43 and 09.45 l/s.