Research On Image-guided Hydrogeological Parameter Inversion Method

Groundwater,which accounts for 99%of the total liquid freshwater on Earth,is crucial for the development of human society and ecosystems.However,groundwater is currently facing serious problems such as pollution and overexploitation.Groundwater numerical simulation is an important means to solve groundwater problems and can be used for groundwater resource assessment,pollutant prediction and control,identification of potential overexploitation problems,providing scientific basis and decision support for the resolution of ecological and environmental geological issues.Accurately describing the heterogeneity of aquifers is the key to achieving simulation of real and reliable groundwater behavior.Therefore,researching methods for accurately characterizing hydrogeological parameters of aquifers has important research value and practical significance.The classical Tikhonov regularization-based least squares inversion method is a common approach for interpreting hydraulic head observation data to obtain hydrogeological parameters of aquifers.To avoid overfitting of data and overcome the instability of the inversion problem,this method imposes smooth constraints on the parameter estimation process,which leads to the problem of fuzzy characterization of the structural boundaries and inaccurate description of aquifer heterogeneity.In this paper,we fully utilized prior information and improved the smoothing effect of the Tikhonov regularization by extracting structural information from guided images,making the parameter distribution at the structural position tend towards the actual aquifer structure.By establishing numerical models and interpreting observation data from laboratory experiments,we verified that the least squares inversion method based on image-guided constraints has a significant improvement in parameter estimation,thus solving the problem of ambiguous structural boundary identification in hydraulic tomography imaging.The specific research work and results are presented below.(1)Based on the traditional Tikhonov regularization-based least squares inversion method.this study introduces a guiding image that describes the hydrogeological structure information as an inversion constraint,which optimizes the smoothing effect of the regularization term on the parameters and achieves accurate identification of hydrogeological parameters related to the boundary positions of the aquifer structure.The structural constraint of the guiding image is successfully integrated into the COMSOL-MATLAB inversion platform,achieving the purpose of synchronously inverting the permeability coefficient and the storage rate under twodimensional unsteady flow conditions.(2)An improvement to the regularization term is proposed by considering two additional smoothing constraints along two structural feature directions,in addition to the traditional Tikhonov regularization that only smooths in the two coordinate directions.Different smoothing weights are set for each smoothing direction to control the distribution of parameters at the boundary positions.A full range of structural constraints is then applied to the elements at the boundary positions based on the boundary type identified in the guiding image to ensure that the smoothing constraint of the guiding image on the parameter inversion is effective.(3)During the numerical simulation stage,three numerical models were created,each with distinct geological structural features.The box-shaped anomaly model and the tilted anomaly model were used to estimate two types of hydrogeological parameters using two different inversion methods before and after the improved regularization term was applied,respectively.By comparing the parameter distributions and numerical ranges,the effective identification of different boundary structures was verified using the image-guided least squares inversion method.The step-like anomaly model was used for further analysis of parameter inversion under image-guided conditions.The results showed that the constrained inversion effect of the low-resolution numerical model was better than that of the high-resolution model,and the prior distribution of a known parameter could improve the estimation of another parameter.The image-guided inversion method had better interpretation of noisy data than the inversion method without image guidance,and the higher the noise level,the more significant the advantage of the image-guided inversion method.(4)Using the image-guided inversion method to interpret experimental data of rock fractures,the high permeability and low specific storage in fracture locations were successfully captured.Compared with the inversion results without image guidance,the image-guided inversion results can characterize the complete fracture structure and improve the estimation of the matrix area,verifying the feasibility and superiority of this method at the laboratory scale.The effects of parameter initial values and image guidance settings on parameter estimation were further investigated,and the results showed that the initial parameter values have a significant impact on the inversion effect of rock fracture models with large differences in media properties.Partially correct image guidance can improve the inversion effect in locations with structural information,but incorrect structural information can have a significant negative impact on the inversion results.Therefore,when using this method to invert hydrogeological parameters,the accuracy of the image guidance should be carefully considered.