Multiobjective Optimization Of Soil Water Flow Model

The quantitative study of hydrological cycle in saturated unsaturated zone is an important link to maintain the sustainable development of environment and protect groundwater resources.The mechanism model based on Darcy's law and mass conservation principle is the most widely used method to quantitatively study the hydrologic dynamics of unsaturated zone.Among them,unsaturated hydraulic parameters are necessary to solve the model,which are used to describe the basic information of hydrological processes such as storage,release and transfer of water in soil,so they are the prerequisite and key link to ensure the rationality and effectiveness of the model.The inversion algorithm based on dynamic data provides an effective way for hydraulic parameters calculation and model optimization,and multi-objective optimization can effectively reduce the uncertainty of model inversion and improve the simulation of parameters.In this study,the multi-objective optimization algorithm(AMALGAM)and HYDRUS-1D software are combined to achieve the multi-objective driven model optimization in numerical experiments and field experiments.Based on the meteorological and soil data of Tongzhou District,Beijing,the numerical simulation experiment was carried out with the single objective optimization of soil moisture content as the control group,considering the dynamic data of soil moisture content,matrix potential and boundary flux.The four different data types were combined to optimize the simulation.A total of 48 soil hydraulic parameters were optimized with 6soil hydraulic parameters in each layer of 8 layers,It also considers the multi-objective optimization of different time series and different depths.The performance of the optimized parameters is judged by the water characteristic curve,hydraulic conductivity and the calibration and prediction of water content.Based on the dynamic data of water content measured in Erdos,Inner Mongolia in 2017 and2019,the applicability of the model in the field actual water dynamic was further discussed.In the field experiment simulation,multi-layer water content was taken as the multi-objective optimization,which was divided into homogeneous and four layer soil columns,that is,5 soil hydraulic parameters and 20 soil hydraulic parameters were optimized,The optimized parameters are judged by the calibration of water content and 1:1 line diagram.According to the results of multi-objective optimization,we analyzed the uncertainty of different criteria and Pareto trade-off,and the main conclusions are as follows: 1.Adding different data types will have different effects on the calibration and prediction of water content.The fitting of surface water content is better with the addition of EA;Adding qbot can better fit the water content of the bottom layer;If the added data type can describe the water content change of the whole soil layer,the calibration and prediction fitting of water content will be better.2.The more the objectives are,the worse the optimization effect is;3.Different conditions have different effects on different data types;4.The double objective optimization results of water content and matrix potential are the best combination to describe the local soil hydraulic characteristics;5.Different target sensitivities have different effects on the results.