Mitigating Groundwater Risks with Advanced Aquifer Testing and Machine Learning

Estimated Reading Time: 4 minutes Listen to this

Manuel Díaz-Cofré Member Name


Groundwater systems exist within complex geological terrains. Complexity gives rise to uncertainty, and where there’s uncertainty, there’s risk.

In the mining, infrastructure and coal seam gas sectors, these risks apply throughout a project’s lifespan. Protecting the availability and quality of groundwater is essential, and we also need to understand the risks that groundwater can have on projects, for example inflow into mines, tunnels and other subsurface excavations during construction and operations. If unmitigated or poorly understood, these risks may be expensive, delaying, and dangerous, and exploring them early in the project can help avoid adverse effects and unexpected costs.

The primary goal of groundwater investigations is to identify the significant geological features, hydraulic properties, boundaries, and quantities flowing within the groundwater system early in the project – as these will affect a project’s operations as well as potentially affecting the natural resources sustained by the groundwater system.

Advanced aquifer test analyses and machine learning are revolutionizing our understanding of complex groundwater systems and of how to mitigate groundwater-related risks. One of the secrets of success is to extract as much information as possible from your datasets by embracing new digital techniques and technologies.

Maximizing the value of the data may lead to opportunities to optimize operational expenditure while reducing potential environmental impacts.

Advanced aquifer testing and machine learning accelerate understanding

Assessing the potential impacts of development on groundwater and groundwater-dependent ecosystems often relies on hydrogeologic conceptual models (HCM) and subsequent numerical groundwater modelling. These tools are also used to predict groundwater inflow to mines, tunnels and other subsurface excavations during construction and operations.

Typically, HCMs are developed based on very limited data, and the resulting uncertainty poses risks during permitting, construction and operation of projects. For example, if key conditions or features are overlooked, forecasts of project impacts and water management requirements are not accurate. To mitigate these risks, it is critically important to get the most out of the site investigation datasets when developing the HCM.

Advanced aquifer testing and exploratory data science techniques such as machine learning provide opportunities to accelerate our understanding of the roles of the key features in the HCM that may materially influence groundwater management at a given site. This can lead to smoother, faster regulatory approval processes and greater operational efficiency.

Advanced aquifer testing goes beyond estimating hydraulic parameters using standard analytical flow equations. It uses diagnostic flow analysis to explore the influences that affect how an aquifer responds to pumping stresses. These influences can include ‘no flow’ barriers, which impede the lateral and/or vertical flow of groundwater, and sources of recharge to the system like the aquifer’s connection to rivers, to other aquifers (through leakage), or to permeable conduits such as faults or karstic terrain. For example, the dewatering requirements for a mine will be much greater when the mine is developed in an aquifer that is in direct hydraulic connection to a major river system. Understanding these hydraulic influences is key to achieving reasonable estimates of groundwater inflow during the feasibility stage of a project when risks and costs are being heavily scrutinized.

If we combine advanced aquifer testing with machine learning, we’re able to extract more information out of our limited datasets. The combination of techniques helps to create more meaningful relationships and interpretations to better characterize the site. Machine learning can identify patterns and trends in the data that would not be obvious through standard hydrogeological interpretation. Examples of these types of exploratory data analysis include characterizing and hypothesis testing of which specific fault zones or geologic conditions have potential to alter groundwater flow dynamics, and identifying subtle hydrochemical trends before, during and after pumping tests that provide insights on aquifer connectivity to other formations or water bodies.

Combining advanced aquifer testing and machine learning allows hydrogeological conceptual uncertainty to be highlighted at an early stage of the project. When the potential influence of boundaries in the groundwater system is better understood and acknowledged, the influences can be better evaluated, either through development of alternative conceptual models, or through further targeted site investigations.

For example, if there is evidence of a potential connection between a river and an aquifer targeted for dewatering, numerical models can test alternative hydrogeologic scenarios with and without the connection and run uncertainty analysis on both scenarios. Furthermore, once a boundary has been identified, field investigations can be designed to better characterize the physical extent and parameters of the boundary, which would otherwise go undetected.

Putting it in practice in mining

In complex mining environments, advanced aquifer testing techniques and machine learning can extract more value from site investigation data to better understand the uncertainty in the HCM. These advanced techniques can help to identify the locations and roles of hydraulic boundaries in complex mining terrain, and it is often possible to determine distances of these features from the pumping bore. This enables correlation with the geological model and HCM features and leads to better understanding of the hydraulic behavior of individual hydraulic boundaries, such as faults and their damage zones.

Applying these techniques to historical data, like older pumping tests, is yielding new insights into groundwater flow patterns and HCMs. The results are dramatic: boundaries and linear features that influence flow can be better understood, leading to revised HCMs, better prediction of drawdown impacts, and narrower estimates of potential inflows to mines. For a detailed example, check out our recent IMWA paper.

Improvements like these enable a much more accurate estimation of open pit and/or underground mine inflow during the feasibility stage of a project, which affects everything from predicting water quality to optimizing the sizing and design of the dewatering and water treatment system and has significant implications for optimizing capital expense (CAPEX) and operating expense (OPEX) budgets.



Manuel Díaz-Cofré Member Name


discover more

Golder uses cookies to ensure that we give you the best experience on our website. By continuing to use this website, we assume that you consent to receive all cookies on our website.

OK Learn More