Inversion Of Non-gaussian Aquifer Parameters Based On Multi-source Observation Data

The model parameter plays an important role for model prediction.However,the complex natural environment conditions and the limited number of observation bring great challenges for parameter estimation.As an important hydrogeology parameter,hydraulic conductivity usually shows a non-Gaussian pattern due to lithofacies heterogeneity in alluvial aquifers,which might cause great difficulties for traditional inverse methods.Multiple-point geostatistics(MPG)methods have been widely used to estimate non-Gaussian conductivity fields,but cannot integrate dynamic data.The data assimilation can integrate dynamic data,but only works for Gaussian fields.Therefore,we developed a new data assimilation framework,ESMDA-DS,by coupling the Direct Sampling method(DS,one of multiple-point geostatistics method)and ESMDA method(one of data assimilation method).Synthetic cases are conducted to explore the influence of different types of datasets.Besides,the influence of the number of ensemble and pilot points are discussed.The content is divided into four parts:(1)The ESMDA-DS method can not only preserve the non-Gaussian pattern,but also integrate the dynamic data to identify the non-Gaussian aquifers accurately.(2)Compared with integrating single observation data,integrating piezometric head,concentration,temperature and hydraulic conductivity can effectively improve accuracy of the estimation,and reduce the uncertainty.Also,data value need to be considered while assimilating,such as the data variation and observation error.Integrating data of low assimilation value may lower the accuracy.(3)Results from the discussed cases show assimilating temperature can achieve a higher estimation accuracy than concentration.Since heat migrates faster than solute,so the temperature can carry more useful information than the concentration when at the same time step.However,this conclusion might be case-by-case.For different combinations of parameters describing mass and heat transfer,it is possible that the solute migrates faster than the temperature.Therefore,in reality,it is important to choose appropriate parameter combination for the estimation.(4)The number of ensemble and pilot points have a greatly influence on the estimation.For the number of ensembles,within a certain range,the realization increasing can improve the estimation accuracy.Beyond this range,it may reduce the accuracy.As for the number of pilot points,with the increase of the number of pilot points,the structure of channels can be better captured,and more additional information can be excavated,obtaining a better estimation results.However,too many pilot points may amplify the error carried by the data,and may lose the nonGaussian pattern,therefore cannot obtain an ideal result.