Geophysical prospecting and machine learning for estimating landfill leakage risk

The main methods for monitoring the damage of the anti-seepage layer of leachate are to collect water and soil samples from the landfill site, to determine the content of pollutants through indoor chemical analysis, and ground geophysical prospecting.On the basis of these traditional environmental geological exploration methods, this paper uses GIS (geographic information system) technology combined with machine learning method and impact factor prediction method, the leakage of landfill leachate is effectively predicted. Firstly, the landfill area is meshed with a resolution of 5m×5m, and the corresponding position coordinate information is assigned to each meshing unit, BP neural network model was established by using machine learning method to predict the leakage point of landfill site.Based on an investigation of a landfill site in northern China, 880 test samples and 220 verification samples were selected. The groundwater depth, resistivity, polarizability, whether or not the cophase axis of the radar reflection waveform is abnormal, the relationship between the cophase axis and the position of the nearby monitoring hole, and the half-life time information are selected as the input factors. The output is the leakage probability of each grid cell. The prediction results of two machine learning methods (BP neural network and RBF Neural Network) are compared.The results show that the BP neural network model has better prediction effect, the influence factors of groundwater table depth, resistivity and the relationship between resistivity and location of monitoring hole have the greatest influence on the prediction of seepage probability of impermeable membrane.