Many geotechnical failures are not caused by the absence of analysis. They arise when decisions are made on incomplete characterization, misinterpreted evidence, or confidence that exceeds what the data and models can support.

 

Modern geotechnical engineering is not short of methods. Mining projects routinely use empirical, observational, analytical, numerical, deterministic, and probabilistic approaches. Yet failures still occur because geotechnical design is not only a calculation task; it is a decision task under uncertainty. As Dunn argues, uncertainty in mining geomechanics is shaped by natural variability, lack of data, and lack of knowledge, and reducing that uncertainty is central to robust design.

 

The first weakness is usually the ground model, not the toolset. The design is only as good as the geological, structural, hydrogeological, and geotechnical models beneath it. Simplification is necessary, but it can also hide the features that control real behavior. If uncertainty in those models is weakly characterized, the design built on them is conditional too.

 

The second weakness is overconfidence in calculation models. Here, the issue is not the geological or geotechnical ground model, but the analytical and numerical calculations built from it. A calibrated calculation model may reproduce past behavior and still be wrong about the mechanism that governs the next decision. Back-analysis can improve fit to history, but it does not necessarily establish predictive adequacy for the future.

 

The third weakness is decision architecture. Geotechnical evidence does not automatically become action; it must pass through a decision structure. In open-pit mining, documented slope-risk decision processes compare options not only on stability, but also on risk, value, and operational consequence, with the final choice made by a risk owner. Once production is underway, the practical flexibility to change slope geometry, cutback sequence, haul access, or ore release may already be limited. The issue is therefore not only whether uncertainty is recognized, but whether the person who owns the risk still has enough operational room to act on it. The same logic applies underground, where committed access development, sequencing, ventilation, support standards, and production commitments can make late geotechnical changes difficult to implement.

 

The fourth weakness is contingency readiness. Even when evidence of developing failure is recognized, it may not be linked to a predefined response system. Monitoring alone is not a safeguard. It reduces risk only when it sits inside a disciplined loop of observation, interpretation, updating, and action — with predefined thresholds, escalation logic, and contingency measures when assumptions are breached.

 

This also points toward the solution. Geotechnical failures become less likely when uncertainty is carried honestly through the full chain from data, to model, to interpretation, to decision, to action. That means building ground models that make uncertainty visible, treating analytical outputs as conditional rather than definitive, preserving enough design and operational flexibility to respond when assumptions weaken, and linking monitoring to threshold-based actions that a risk owner can still implement in time. In mining, better geotechnical outcomes depend not only on better analysis, but on better decision structures around that analysis.

 

* Here, geotechnical failure is used broadly to include not only collapse, but also instability, excessive deformation, support-system breakdown, or loss of safe and intended excavation function in open pit and underground mining.

​