Abdul Wahab Khan
Panaji
A scientific study has used artificial intelligence to monitor Goa’s mining belt, predict environmental risks and support early intervention by regulators.
Published in the Ain Shams Engineering Journal, the study Leveraging Artificial Intelligence for Minimizing Environmental Footprints in the Mining Industry was conducted by researchers from institutions in Nerul (Goa), and Pune, Mira Bhayander and Ambi in Maharashtra.
The study focused on mining areas in Bicholim, Sattari, Sanguem and Quepem.
“Existing AI-based environmental studies largely examine single factors such as water quality or air pollution and are often based in arid regions, making them less applicable to monsoon-driven landscapes like Goa,” said researchers.
The present study
has, however, developed a combined framework to assess multiple environmental risks.
The system uses satellite imagery, drone surveys and IoT-based sensors, processed through machine learning and deep learning models and autoencoder algorithms. These models assess factors such as water contamination, dust levels, vegetation loss and structural stability of tailings dams.
The outputs were integrated into GIS-based risk maps identifying high-risk areas in Bicholim, Sankhali, Sirigao and Sanguem.
The AI model was highly accurate in estimating water pollution levels based on the data it analysed. It also forecast PM10 dust concentrations along haul roads reached about 89 per cent accuracy. Land cover analysis showed vegetation loss ranging from 21 to 38 per cent in active mining zones between 2000 and 2023, with erosion observed in Pissurlem, Codli and Sirigao.
The study also examined tailings dams by combining satellite-based deformation data with rainfall inputs
and time-series models. It detected ground movement of 2 to 8 mm per month and generated early warnings
24 to 48 hours before possible instability during the monsoon.
“AI-enabled environmental intelligence enables accurate, predictive, and continuous monitoring, offering a robust pathway toward proactive risk mitigation,” the study stated.
All datasets were split into training, validation and testing sets, and the models were validated using five-fold cross-validation, the study said.
The researchers said the framework allows a shift from inspection-based monitoring to predictive environmental management and can be adapted for other mining regions in India with similar climatic conditions.
