Integration of Fourth Industrial Revolution technologies and indigenous knowledge in developing a smart and integrated pollution monitoring system: case of Free State province

Abstract

South Africa’s National Ambient Air Quality Standards (NAAQS) emphasize clean air as a fundamental pillar of human health and well-being. Yet, in mining-intensive regions of the country, persistent exposure to harmful emissions continues to compromise both occupational safety and community health. Mine workers remain highly susceptible to respiratory illnesses such as silicosis and pulmonary tuberculosis (PTB), at the same time, surrounding populations experience secondary exposure as particulate matter, sulfur dioxide, nitrogen oxides, and other toxic gases are dispersed into residential areas, particularly during windy seasons. Beyond immediate respiratory complications, this exposure contributes to long-term cardiovascular disease, reduced life expectancy, and escalating healthcare costs. Conventional monitoring systems, while present, are often static, resource-intensive, and geographically limited, providing delayed or fragmented data that fail to support real- time decision-making. Moreover, enforcement of NAAQS remains inconsistent, with communities largely excluded from monitoring processes, leaving them without accessible tools to validate or report their lived experiences of pollution. The disconnect between regulatory frameworks, technological capacity, and community participation perpetuates environmental injustice in mining-affected provinces such as the Free State. Air pollution, dominated by particulate matter and toxic gases, therefore remains one of the most significant environmental and public health threats in South Africa, as in many developing countries, with strong links to increased mortality, morbidity, and climate impacts. Addressing this challenge requires innovative monitoring frameworks that transcend the limitations of traditional systems by integrating advanced technological tools with inclusive, culturally resonant approaches rooted in Indigenous Knowledge (IK). This dissertation explores the integration of Fourth Industrial Revolution (4IR) technologies, specifically Machine Learning (ML) and the Internet of Things (IoT), with an Indigenous Knowledge (IK) approach to design, implement, and evaluate a smart, adaptive, and community-centred air pollution monitoring system in the Free State Province. To achieve this, the study employed a mixed-methods research design. Quantitative techniques included the collection and pre-processing of datasets from the Pelonomi air quality monitoring station in Mangaung, complemented by limited deployments of IoT wireless sensors. Supervised ML algorithms, namely: Random Forest, Gradient Boosting, Support Vector Machine, and Decision Tree Regression, were trained to forecast pollutants such as PM₂.₅, PM₁₀, and SO₂ up to four days in advance. These predictions were benchmarked against NAAQS thresholds. Qualitative techniques, on the other hand, involved structured surveys and interviews with mine workers, residents, and environmental knowledge holders in the Lejweleputswa district. Indigenous indicators, such as dust storms during windy seasons, odour perception, and respiratory discomfort, were systematically documented and modelled using Fuzzy Cognitive Maps (FCMs), ensuring formal representation, validation, and integration with scientific datasets. A three-phase, agile systems development approach guided the framework’s implementation: (1) Data Acquisition and Prediction – real-time and secondary data collection, cleaning, and ML-based forecasting; (2) Indigenous Knowledge Integration – rigorous modelling and validation of IK indicators through FCMs to complement scientific predictions; and (3) Communication and Dissemination – the development of a mobile Android application that provides accessible forecasts, validates machine outputs against IK, and allows communities to log observations directly. This participatory design ensured inclusivity, especially for semi-literate and illiterate populations often excluded from technology-driven monitoring. Evaluation confirmed both technical robustness and societal acceptance. Forecasts showed strong alignment with South African Weather Service (SAWS) datasets, demonstrating predictive reliability. Cross-validation revealed a high degree of complementarity between IK observations and scientific outputs, reinforcing the value of hybrid knowledge systems. A community-based evaluation achieved an 89% user approval rating, highlighting the usability, cultural relevance, and potential for wider societal adoption. The contributions of this research are multi-dimensional. Scientifically, it advances environmental informatics by demonstrating a hybrid framework that systematically integrates qualitative IK indicators with quantitative ML forecasts. Technologically, it delivers one of the first prototypes in South Africa to merge IoT and ML forecasting with IK validation in a mobile platform. Societally, it empowers marginalized communities by transforming them from passive recipients of information to active contributors in environmental monitoring. Policy-wise, it provides empirical evidence for inclusive environmental governance, directly aligning with national frameworks and global agendas, including SDG 3 (Good Health and Well-being), SDG 11 (Sustainable Cities and Communities), and SDG 13 (Climate Action). Limitations included restricted access to mining shafts, which limited large-scale IoT deployment and necessitated reliance on secondary datasets. Nevertheless, the results underscore the feasibility, scalability, and replicability of the system. Future work should expand sensor coverage, enhance interoperability across regions, and extend the hybrid framework to other domains such as water and soil pollution monitoring.