A machine learning–based approach to regional‐scale mapping of sensitive glaciomarine clay combining airborne electromagnetics and geotechnical data

ABSTRACT Sensitive glaciomarine clays, often referred to as ‘quick clay’, commonly occur in many countries at high, northerly latitudes, causing frequent and occasionally devastating landslides. The salt content of quick clay is strongly correlated both its shear strength electrical resistivity. Hence, it can be mapped using electromagnetic methods more efficiently than traditional intrusive methods, the latter which slow costly. However, resistivity signature non‐unique, leading ambiguous, imprecise interpretations geophysical models. In this study, we present an improved method for predicting probability airborne electromagnetics. Using machine learning algorithms, combine models with geotechnical data address issue their non‐unique signature. Beyond values, algorithms use spatial derivatives attributes. We evaluate performance collected from a road construction project central Norway. Results show that able make plausible accurate predictions occurrence few 10 boreholes across area 14.8 km 2 , outperforms simple interpretation based on intervals alone. addition ‘best guess’ categorical classification, these output estimates, demonstrate they are reliable indication uncertainty. accuracy also tends increase included training data, helping compensate limited resolution electromagnetics data. Given petrophysics clays test site consistent observations other regions, expect has potential hazard mapping efficient by offering valuable early‐phase insights, large time cost savings infrastructure planning regional mapping.