Document Type : Original Research Paper
Authors
1
Department of Chemistry, Faculty of Science, Abubakar Tafawa Balewa University, Bauchi
2
Department of Environmental Management Technolgy
3
Interdisciplinary Research Centre for Membranes and Water Security, King Fahd University of Petroleum and Minerals, Dhahran, 31261, Saudi Arabia.
4
Chemistry Department, Near East University
5
Department of Civil Engineering, Prince Mohammad Bin Fahd University, Al Khobar, Saudi Arabia
10.22059/poll.2025.400497.3073
Abstract
This study used advanced machine learning models to examine how soil chemistry, specifically pH, organic carbon (OC), organic matter (OM), and cation exchange capacity (CEC), influences the movement and spread of the 232Th radionuclides in abandoned mine soils. Soil samples were collected across different seasons. The models employed were Gaussian Process Regression (GPR), Long Short-Term Memory (LSTM) networks, Adaptive Neuro-Fuzzy Inference System (ANFIS), and Random Forest (RF). Feature selection was used to identify optimal model combinations, labeled as C1, C2, and C3. The GPR-C1 model demonstrated the highest predictive accuracy, with its performance improving significantly during the verification phase. It achieved the best results during both training (RMSE = 7.0851, DC = 0.6482) and testing (RMSE = 4.5808, DC = 0.5848), highlighting its ability to capture the complex, non-linear relationships between soil properties and 232Th mobility. Conversely, RF models performed poorly, likely due to their inability to handle intricate geochemical interactions. A key finding was the 20-30% improvement in prediction accuracy during the testing phase compared to the training phase, suggesting better generalization. This aligns with field data showing increased 232Th mobility in wet seasons due to leaching and runoff. The GPR-C1 model’s significantly lower testing RMSE compared to its training RMSE reinforced the importance of seasonal dynamics. Therefore, this model demonstrates significant potential for accurately predicting 232Th behaviour and distribution, crucial for environmental risk assessments. Hence, accurate predictions of 232Th distribution can guide targeted remediation efforts and inform land management practices, mitigating risks associated with 232Th exposure.
Keywords
Main Subjects
)