The long-term environmental management of closed landfills presents significant challenges due to persistent uncertainty regarding residual contamination and pollutant release processes. While conventional aftercare practices, such as those mandated by the European Landfill Directive, focus on emission monitoring and maintenance of engineered barriers, they often overlook the complex subsurface dynamics of pollutant mobility within landfill waste bodies. Accurately quantifying the internal releasable pollutant content, referred to as the emission potential, is essential for developing realistic and scientifically-grounded aftercare strategies.
In this dissertation, I present an integrated framework to estimate and predict landfill emission potentials by combining stochastic modeling, Bayesian uncertainty quantification, data assimilation, and hydrogeophysical measurements. The research introduces a stochastic Lagrangian-based travel time modeling approach to simulate the heterogeneous water flow and solute transport within landfill bodies. This method, unlike traditional grid-based models, captures preferential flow phenomena and accommodates the spatial variability inherent in landfill waste structures.
The model calibration is performed using Bayesian inference, employing long-term observational data of leachate production and quality from the Braambergen landfill in the Netherlands. This probabilistic calibration explicitly quantifies uncertainties in model parameters and outputs, providing more credible risk assessments and long-term predictions of leachate emissions.
Recognizing the risk of error accumulation in history-matching methods, I further implement data assimilation techniques, notably the Weakly Coupled Particle Filter (WCPF) and a hybrid Particle Filter–Markov Chain Monte Carlo (PF-MCMC) method. These approaches enable sequential updating of model parameters and system states as new data become available, improving the predictive performance and reducing uncertainty over time. The PF-MCMC method, in particular, can estimate parameters and hidden processes, which is very helpful for understanding the dynamics in the landfill.
To further enhance the accuracy of emission potential estimations, the framework integrates hydrogeophysical data obtained through Electrical Resistivity Tomography (ERT). Using a Bayesian evidential learning approach, resistivity measurements are directly mapped into probabilistic water storage estimates within landfill waste bodies. This additional constraint strengthens the characterization of subsurface hydrological conditions, distinguishing between leachable and isolated water fractions.
The dissertation is structured across six chapters, beginning with an overview of the landfill aftercare problem, followed by the development of the stochastic modeling framework, the application of particle filtering and PF-MCMC, the incorporation of ERT data through Bayesian evidential learning, and concluding with a synthesis of findings and recommendations for future research.
Overall, this work advances the scientific understanding of landfill emission dynamics by offering a unified methodological framework that integrates stochastic modeling, data assimilation, and hydrogeophysical surveying. The contributions herein support the development of more robust, data-driven, and cost-effective strategies for landfill aftercare, ensuring long-term environmental protection and sustainability.

Accurate estimation of water storage in municipal solid waste landfills is critical for assessing leachate-generation risk yet remains challenging due to pronounced heterogeneity. Here we apply Bayesian Evidential Learning (BEL) to directly relate Electrical Resistivity Tomography (ERT) data to total water storage (TWS), bypassing explicit inversion. A semi-parametric forward model generates 100,000 synthetic TWS–ERT pairs spanning stochastic saturation fields and petrophysical uncertainty. A Bayesian neural network captures data-dependent predictive uncertainty, while stratified resampling and adaptive weighting mitigate class imbalance across the TWS range. The framework yields well-calibrated posterior estimates and consistent agreement with independent water-balance benchmarks from four field transects. The BEL–ERT workflow provides a rapid, open-source alternative for landfill monitoring and highlights the potential of uncertainty-aware learning from synthetic ensembles to quantify water storage in heterogeneous near-surface systems.

Microbially induced carbonate precipitation (MICP) has emerged as a promising ground improvement technique, with MICP-treated soils exhibiting substantial enhancements in strength. However, experimental results revealed significant variability in strength outcomes of MICP-treated soils, even under identical treatment conditions and soil properties. This uncertainty in strength is challenging to capture using traditional predictive approaches such as conventional constitutive models. The present study leverages artificial intelligence to address the challenge by developing a Bayesian neural network (BNN) model for predicting the strength of bio-cemented soils while considering uncertainty. A dataset comprising 480 experimental samples was used to develop the model. The results indicate that carbonate content and confining pressure emerge as the most influential factors governing the strength of bio-cemented soils. The BNN model exhibits lower uncertainty when predicting bio-cemented soils with relatively low strength, while demonstrating higher uncertainty for soils with strength exceeding 2 MPa. Moreover, micromechanical investigations using the discrete element method (DEM) reveal that multiscale factors, including crystal distribution patterns, fabric and spatial heterogeneity of precipitates, contribute significantly to the strength uncertainty of bio-cemented soils. The developed BNN model provides an alternative tool for predicting bio-cemented soil strength with quantified reliability, facilitating the design of MICP treatment and its application in geotechnical engineering.