An early geostatistical approach combining geotechnical and geophysical data

Introduction

Airborne electromagnetic (AEM) survey data was used to supplement geotechnical investigations for a highway construction project in Norway. Heterogeneous geology throughout the survey and consequent variable bedrock threshold resistivity hindered efforts to directly track depth to bedrock, motivating us to develop an automated algorithm to extract depth to bedrock by combining both boreholes and AEM data. We developed two variations of this algorithm: one using simple Gaussian or inverse distance weighting interpolators, and another using ordinary kriging and combined probability distribution functions of input parameters.

Evaluation shows that for preliminary surveys, significant savings in boreholes required can be made without sacrificing bedrock model accuracy. In the case study presented, we estimate data collection savings of 1000 to 10,000 NOK/km (c. $160 to $1600 USD/km) would have been possible for early phases of the investigation. However, issues with anthropogenic noise, low signal, and uncertainties in the inversion model likely reduced the comparative advantage that including AEM provided.

AEM cannot supersede direct sampling where the model accuracy required exceed the resolution possible with the geophysical measurements. Nevertheless, with the algorithm we can identify high probability zones for shallow bedrock, identify steep or anomalous bedrock topography, and estimate the spatial variability of depth at earlier phases of investigation. Thus, we assert that our method is still useful where detailed mapping is the goal because it allows for more efficient planning of secondary phases of drilling.

Conclusion

We have successfully created a depth to bedrock tracking algorithm which combines AEM and borehole data and can account for variable bedrock resistivity. We can quantify prediction error given sufficient borehole information, yet these are underestimates. Finally, while some user input is required to find reasonable bounds on interpolation parameters, the algorithm is far more efficient than the cognitive modeling approach previously used in this project.

Evaluation of the algorithm developed shows that the cost of site investigations can be significantly reduced by using this method. Based on our cross-validation of the algorithm, combining AEM and borehole data in this way can reduce costs for this type of site investigation by 1000 to 10,000 NOK/km depending on the desired accuracy of the depth to bedrock model. The tool is only applicable to early phases of site investigation, however, due to the precision limits of the geophysical measurements themselves.

Despite the measurement resolution limitations, by using AEM this algorithm in an early phase of site investigation, we can (a) identify zones where shallow bedrock is likely to be; (b) identify areas of steep or locally anomalous of bedrock topography; and (c) estimate the spatial variability of depth, giving a more informed choice of borehole spacing. Acquiring a detailed depth to bedrock model should typically be more cost efficient by using AEM because secondary phases of drilling can be planned to target high-priority areas.

We suspect that with some modifications, our method may be applied to cases where geological conditions are not ideal for mapping the bedrock-sediment interface, but further testing is required.

Acknowledgement

We thank the Norwegian Public Road Authority (SVV), in particular road design manager Arvid Sagbakken, for financing this study and giving permission to publish.

In the presented case study NGI acted as a geotechnical advisor to COWI and we are grateful to Frode G Bjørvik for supporting our concept to use a combined geophysics and geotechnical approach. Numerous colleagues at NGI have contributed to our results presented here, Steinar Herman, Kristoffer Kåsin, Kristine Eksethand Magnus Rømoen to name a few. Thank you especially to Asgeir Kydland Lysdahl for preparing location and sedimentary maps for us (Figs. 1 and 3 in the article).

We also thank Dr. Alexander Braun for supervising a portion of this work undertaken at Queen's University, Kingston, ON, Canada.

Neither SkyTEM Surveys nor Århus University have directly contributed to this work but we would not have gotten to these results without SkyTEM's focus on innovative systems and excellent data acquisition and a general collaboration between NGI and the hydrogeophysics group at Århus University.

Reference

The full paper can be requested through the download link above or found directly at www.sciencedirect.com.

Christensen, C. W., Pfaffhuber, A. A., Anschütz, H., & Smaavik, T. F. (2015). Combining airborne electromagnetic and geotechnical data for automated depth to bedrock tracking. Journal of Applied Geophysics, 119, 178-191.