- Organisation type
- Public administration at regional level
- Use case area
- Environment
- Geographic scope
- Spain
- Domain
- Economy
- Environment
Challenge:
In the mining and metals sector, environmental risks are often evaluated using traditional, qualitative field-based methods that can be time-consuming and subject to human bias. Mining operations require thorough evaluations to mitigate the impacts on biodiversity, water resources, and land degradation. The challenge is to improve the accuracy and efficiency of Environmental Impact Assessments (EIAs), especially in ecologically protected areas within the Natura 2000 Network.
Approach:
This use case introduces an innovative AI-based approach to conducting Environmental Impact Assessments (EIAs) using AutoML (Automated Machine Learning) and Bayesian modelling. These techniques analyse large datasets to identify patterns between environmental variables, reducing human bias. The approach is demonstrated at a slate mine in Galicia, Spain, within protected areas under the Natura 2000 Network. With the rising demand for materials like lithium and cobalt, significant environmental challenges are anticipated. This solution aims to create accurate EIA indexes by employing data discretisation, unsupervised Bayesian learning, and computing risk and uncertainty. Automating the EIA process allows for dynamic updates and improved decision-making. Bayesian networks are utilised for their prediction accuracy and ability to represent probabilistic relationships. As a result, key environmental factors and their interrelationships are identified, providing a robust framework for assessing and mitigating environmental impacts in mining.
Outcomes:
The application of AI models in Environmental Impact Assessments (EIAs) has led to significant improvements in the accuracy of evaluating environmental risks. These AI-driven assessments provided a more precise evaluation of the moderate environmental impact of mining operations, particularly identifying critical factors such as soil erosion and species protection that required immediate attention. The data-driven nature of these AI models allowed for more reliable, evidence-based recommendations, with key environmental variables being weighted 6 to 12 times higher than less impactful factors.
Data Sources:
The data used for the environmental impact assessment was collected from various attributes of the study area in Vilarbacú, Galicia, including 10 environmental factors (e.g., landscape quality, soil degradation, surface waters, fauna, flora). These were analysed using AI tools that processed the relationships among these factors, resulting in a Bayesian network model that could simulate and predict environmental impacts.
Additional Information:
Point of Contact:
- informacion
uvigo [dot] es (informacion[at]uvigo[dot]es)