Towards Automatic Detection of Quick Clay Using Field Testing

The identification of quick clay is of paramount importance for proper landslide hazard and risk assessments in regions of Scandinavia and North America. Field investigation methods, such as rotary pressure sounding and total sounding, are effective and robust methods used in Norway to identify the presence of quick clay. However, their interpretation usually relies on visual inspection.Theinterpretation of soil layers based onthese methods bear uncertainties and must be supplemented with laboratory tests to be able to predict quick clay presence accurately. This study focuses on developing and employing an algorithm to automate quick clay detection from field tests such as rotary pressure sounding and total sounding and to quantify the uncertainties associated with these predictions.Data from three different sites, Tiller-Flotten, Buvika, and Kvithammer-Åsen in mid Norway, are used to test the algorithm. The algorithm predicts the presence of quick clay by utilizing predefined thresholdson field data and laboratory data. Confusion matrix performance parameters, such as True Positive Rate (TPR), False Positive Rate (FPR), the accuracy, and the precision are computed to evaluate effectiveness of the algorithms. The results indicate that the selection of threshold values is important in the detection of the quick clay presence.

Keywords: Quick clay, Detection, Total sounding, Rotary pressure sounding, Python, Receiver operating characteristics, Optimum operation point.