Iorhemen, Oliver

Clean drinking water access is essential for public health and regarded as a scarce resource for Indigenous communities in rural and remote areas. In this research, a new iron and manganese prediction method based on Data Augmentation and Machine Learning Algorithms to be applied to drinking water in BC’s First Nation communities is reported. GAN based modelling and NIBS-NI based modelling were developed to investigate the effects of different data augmentation methods and predictors for iron and manganese prediction results. Reliable synthetic data was obtained through both data augmentation methods, allowing 4 machine learning algorithms to predict iron and manganese utilizing 3 and 5 physical properties respectively. Compared with RF, XGB, and DT machine learning models, the GBR model showed the strongest fitting ability and accurate predictions for both NI-BS-NI based modelling and GAN based modelling in predicting iron and manganese, with the Train R2 and Test R2 of two models nearing 1, and all the RMSE scores are below 0.06. The decision-making tool developed using GAN technology is considered to have greater application potential due to its ability to provide accurate predictions while requiring only 3 input physical parameters.

Manganese is one of the most occurring heavy metals in the source and drinking water of rural and remote communities in Canada. The removal of manganese from groundwater was explored by designing a biochar-assisted electrochemical technology as a small-scale and chemical-free method. Electrocatalytic oxidation of manganese was acquired while utilizing wood residue biochar as a coating layer on the surface of activated carbon felt anode. The biochar-coated anode showed enhanced conductivity and surface area, consequently facilitating the electron exchange and manganese removal efficiency. Manganese initial concentration, current intensity, pH, and time were considered the main variables and their effect on the removal efficiency of manganese from groundwater was investigated. Current had the most influence on removal efficiency, as evidenced by the fact that no manganese removal occurred at zero current and the removal efficiency was increased by enhancing current from 25 to 75 mA. With the Mn initial concentration of 2 mg/L, current of 75 mA, pH 9, 97.5 % manganese removal efficiency was acquired, which helped to reduce the contaminant concentration under the maximum acceptable concentration. The results demonstrated a well-fitted pseudo-first-order model with a rate constant of 0.0411 min-1 for electrocatalytic oxidation. Real groundwater of Lheidli T'enneh community in Northern BC was employed to evaluate the impacts of co-existing ions on the manganese removal performance. The results confirmed that not only did the use of real groundwater have no detrimental impact on the system’s efficiency, but also the system was able to decrease the high hardness of 268.2 mg CaCO3/L in the groundwater to 72.2 mg CaCO3/L suitable for drinking purposes. Investigation of the manganese removal mechanisms indicated that various species, such as hydroxide ions, sulfate, and hydroxide radicals can play key roles in transferring or removing manganese from groundwater by oxidation and precipitation pathways. The formation of a black powder precipitated on the anode surface and in the cell after treatment proved that oxidation of manganese takes place in the system. Furthermore, the evolution of pH after 90 minutes of reaction further confirmed the presence of hydroxide ions in the system. Overall, the designed system achieved a significant performance in removing manganese and hardness from groundwater while utilizing biochar as a waste and cost-effective material.