Bayesian updating for improving long-term performance prediction of composite liners in landfills with temporally variable leachate head

Abstract

Predicting the long-term performance of landfill composite liners remains challenging due to temporally variable leachate heads combined with various sources of uncertainties. This study proposes a Bayesian updating framework to enhance the prediction of composite liner performance by systematically integrating field observations with prior knowledge. The framework addresses uncertainties in key parameters—including wrinkle length, hole frequency, hydraulic conductivity, and interface transmissivity—while accounting for model bias and measurement errors. An illustrative example of a geomembrane/geosynthetic clay liner/attenuation layer (GM/GCL/AL) system under high leachate head conditions demonstrates the method’s effectiveness. Results show that iterative Bayesian updating significantly improves leakage rate predictions across different landfill operational stages and refines estimates of contaminant breakthrough time (BTT), reducing prediction errors to within − 10 % to + 5 %, outperforming approaches that partially address uncertainties. Parametric studies further reveal that higher prior coefficients of variation (COVs) for model parameters and model bias lead to more stable predictions. This approach provides a dynamic and adaptable tool for optimizing landfill liner design and maintenance, offering improved accuracy in long-term environmental risk assessment.