Quantifying reliability in geological prediction 15 Mar 2021

Trevor G Carter, TGC-GeoSolutions/Golder and Wayne P Barnett, SRK, Canada
Major improvements over the past decades in computer methods for analysing and synthesizing geological data have coincided with an almost inverse rise in the number of projects where unexpected problems have occurred, not all related to lack of site investigation, more rather due to a lack of fundamental appreciation of how geology may impact construction. Trevor Carter and Wayne Barnett examine some steps to help reverse this worrisome industry-wide trend of projects undertaking significantly increased data synthesis of collected geological parameter information but often ending up with a diminished understanding of actual site geological structural conditions.(1)

Unexpected geology is arguably the one factor most blamed for why failures occur and why projects go wrong, get delayed, or lead to construction claims, cost over-runs or major design implementation changes. Sometimes geological challenges lead to inappropriate invoking of unforeseen ground condition contract clauses as an easy way out of complex contractual disputes. Unexpectedness on some projects, however, stems directly from an inability to differentiate real risk due to geological variability from data inadequacy (Fig 1).(2)

True geological understanding is becoming even further divorced from reality because of the widespread delusionary perception of realism and accuracy created in available modelling software, using little to no real data or hard verification checks (Fig 2).(3) Irrespective of however good the physics of the modelling codes might be, if the correct failure mechanics for the site conditions are not replicated in the constructed model, the results may be totally misleading. This trend towards modelling without adequate understanding needs to be reversed by better verification, calibration and reliability ranking.

Much more attention needs to be paid to properly sub-dividing project sites into realistic geological entities through use of rigorous geological structural domaining techniques, such that much greater understanding is gained of project-specific geological risk. Proper domaining will lead to improved reliability and representativeness in project geomechanical parameter selection for use in design modelling codes and potentially dramatically reducing project risk (Fig 3).

Geological domains

Careful domaining of the region around, and including, the project site is necessary so that unambiguous characterization and engineering scale zoning is achieved of the entire rock volume of significance. Consistency is the key to geologically domaining any site. Each and every domain must reflect this. A domain must encompass only a zone of consistent structural fabric and material properties, and typically should be bounded by definable geological interfaces (Fig 4).(1)

At the most basic level, domaining a project site requires nothing more than mapping and reliably establishing the distribution and variability of distinguishable rock units across the rock volume of interest. This necessitates defining controlling parameters and determining their spatial variability to an appropriate level of detail matched to the scale of the proposed engineering structures.

When 3D geological models are to be developed to help condition engineering analyses for design of major infrastructure components, such as an underground hydropower station, the geological and rock mass domain characterization process needs to be very closely linked with purpose. Geometry and size of defined domain shapes must not only match with the scale of the engineering structures being designed but also must be tailored to the objective of the design.

Fig 4. Workflow for defining and analysing prevailing geological domains<sup>(1)</sup>
Fig 4. Workflow for defining and analysing prevailing geological domains(1)

Domaining detail may need to be different in different parts of the geological model. For example, considerable detail may be needed in parts of the model where significant sensitive excavations are planned, whereas only adequate representativeness may be needed into the far-field. However, stress field boundary condition assumptions must be properly validated by appropriate domain boundary geometry, albeit at much larger domain block size, than may be needed for specific engineering structures. It is thus important to clarify early, and with all stakeholders, what the modelling objective is and what the final product will look like and how it will be used for further analysis.

Categorizing and interpreting data for models

A refocus of effort is needed to ensure that reality is properly replicated in models and that data gaps and uncertainties are highlighted, not hidden, so that filling in such gaps is deemed a critical and necessary task for providing justification for acquisition of additional hard site-specific field data.

Rigorous domaining not only will highlight data gaps and identifiable uncertainties but also should provide focus for proper design. Undertaking feedback checks of the adequacy of available raw mapping and drilling data must be a priority during preparation of the overall engineering geological model of the project site, and also its various linear components. Problems in interpretation can only be addressed with additional data collected during the modelling process (and mapping is data), or can be highlighted to be carried into the next phase of design, as specific, but as yet unquantified, risks (Fig 5).(5)

Fig 5. Typical construction project workflow, showing where structural geology domain input is likely required to inform the design strategy<sup>(5)</sup>
Fig 5. Typical construction project workflow, showing where structural geology domain input is likely required to inform the design strategy(5)

Appropriate categorization of raw data is essential for generating a good representative geological model. The early focus of the model creator should be directed towards defining the attributes of the rock mass in each domain identified across the project site and to identify to what extent the data available will be adequate for subsequent geomechanical analysis. This calls for interaction between the interpreter and the designer.

Objectivity and rational simplification of complexity must prevail in the early geological model interpretation stages. Ideally synthesis should be undertaken while investigations are still ongoing. At this stage geological understanding must guide selection of investigation targets, rather than solely concentrating boreholes around engineering structure details or following routine regular closely spaced grid style data acquisition approaches. Synthesis and simplification through appropriate categorization while maintaining geological credibility must be the focus for streamlining the geo-model creation.

Reliability ranking

Even with thorough knowledge of data distribution, there will be situations where data deficiencies and uncertainties exist that compound the problem of properly assessing project risk. Verifying the reliability of the primary base data itself and also iteratively establishing the degree of credibility of the geological interpretations based on that data is a critical requirement as a project moves from concept through feasibility to implementation of a design (Fig 3). Reliability must be the yardstick for baselining interpretations when preparing geotechnical baseline reports.

Confidence in interpretation, which in turn means confidence in representativeness and in extensiveness of base data coverage, must be seen as the ultimate control on model credibility. Rigorous reliability checking is thus a must for any scale of project, and particularly so for complex underground works designs.

Fault confidence ranking

Faults are a particular challenge for validating. In consequence, use should be made of the fault ranking reliability scale of terminology and colour coding shown in Table 1 so that clarity is transparent of the reliability of specific wireframe solids incorporated into a 3D geological fault model interpretation. This type of ranking helps geotechnical designers understand which faults the interpreter deems to have more or less certainty for inclusion into any detailed geomechanical model.

Application of a familiar red to green credibility-based colour coding scale should become standard for initial stages of fault and lithological contact wireframe construction in 3D models and also for fault portrayals in cross sections, as it quickly informs the reader of the level of geo-fantasy inherent in the information.

Table 1. Scale for ranking uncertainty in structural feature and rock mass characterization for domain definition<sup>(4)</sup>
Table 1. Scale for ranking uncertainty in structural feature and rock mass characterization for domain definition(4)

For 3D geological models to have credibility, they must define where hard data exists and elsewhere make it clear that the section, or the 3D form geometry, is of lesser certainty. It must also be made clear to any user of the information that interpretations can never fully replicate reality and must always be understood to represent simplifications of presumed geological conditions, with more precision to be expected where better data constraint exists.

Reliability matrix approaches

Approaches employed in Europe for deep excavations have advanced significantly the procedures of reliability checking. Perello et al developed and employed a reliability matrix ranking method for quantifying the relative reliability of interpreted geology along the Alp base tunnels as part of the site investigation and tendering process.(6,7) They developed the R-Index as a matrix measure to basically codify all raw geological and geotechnical data sources and compare their reliability with respect to an assessment of the estimated complexity of geological conditions for that stretch of the tunnel, as interpreted from the same base data. The domaining process outlined in Figs 4 and 5 should provide the basic framwwork on which to build such estimates as a basis for generating the reliability matrix.

Calculation of the index on a segment-by-segment or domain-by-domain basis, depending on the scale of interest, typically allows checking of the:

  • input data types (drillholes, mapping and geophysics data) with offset distance from the tunnel or other key infrastucture of concern, and data quality factored in;
  • geological complexity of the domain being traversed via specifically examining the credibility of the lithological and structural interpretation forecasts, the latter being further sub-divided to examine brittle and ductile deformation interpretations; and
  • experience of the interpreter/modeller and also the reviewer of the geological interpretation.

For construction of linear infrastructure, such as deep mountain tunnels, application of such a matrix ranking approach can be relatively straightforward to undertake on a segment-by-segment basis along the alignment. This success has led to its being recommended as a standard procedure for inclusion in geotechnical baseline reports for civil tunnel contracts in particular.(6-8) While somewhat more difficult to apply for a complex block volume, such as a large open pit or underground mine, proper geological domaining is seen as providing the necessary key to this type of reliability ranking for its future successful implementation on a more volumetric basis.

Author references

  1. Barnett, WP and Carter, TG 2020. Structural domaining for engineering projects. In: Proc. 54th US Rock Mechanics/Geomechanics Symposium, Golden, Colorado. ARMA Paper 20-2105 p12
  2. Carter, TG and Miller, RI 1995. Crown pillar risk assessment - Cost effective measures for mine closure remediation planning. Trans Inst Min Metl Vol 104 pA41-A57
  3. Carter, TG 2015. On increasing reliance on numerical modelling and synthetic data in rock engineering. Proceedings 13th ISRM International Congress on Rock Mechanics, Montreal, Canada, Paper 821, p17
  4. Carter, TG 2018. Suggested standards for improving structural geological definition for open pit slope design, Proceedings 2018 Slope Stability. International Symposium on Slope Stability in Open Pit Mining and Civil Engineering, XIV Congreso, Mineria (IERM), Sevilla, Paper #102, p26
  5. Hudson, JA, and Feng, XT 2007. Updated flowcharts for rock mechanics modelling and rock engineering design. International Journal of Rock Mechanics and Mining Sciences, 44(2), p174-195
  6. Perello P, Venturini G, Dematteis A, Bianchi G W, Delle Piane L, Damiano A. Determination of reliability in geological forecasts for linear underground structures: The method of the R-Index. Proceedings 2005 Geoline, Lyon France, p8
  7. Perello, P. Estimate of the reliability in geological forecasts for tunnels: Toward a structured approach. 2011 Rock Mechanics and Rock Engineering, p671
  8. Venturini, G, Bianchi, G W, Diederichs, M. How to quantify the reliability of a geological and geotechnical reference model in underground projects. 2019 RETC Chicago

Add your comment

Thank you for taking the time to share your thoughts and comments. You share in the wider tunnelling community, so please keep your comments smart and civil. Don't attack other readers personally, and keep your language professional.
In case of an error submitting Feedback, copy and send the text to Feedback@TunnelTalk.com
Name :

Date :

Email :

Phone No :

   Security Image Refresh
Enter the security code :
No spaces, case-sensitive