Rapid Liquefaction Detection Using Remote Sensing Techniques: 2011 Christchurch Earthquake

A sequence of earthquakes causing surface manifestations of liquefaction struck the Canterbury region in New Zealand in 2010–2011. Satellite and aerial imagery is readily available after an earthquake and can be used efficiently and quickly to map liquefaction occurrence with the use of automated classification techniques. Automated techniques applied to satellite multispectral imagery could allow for a minimal compromise between time and accuracy and result in a reliable map of liquefaction surface affects for an area. In this study, multispectral satellite imagery after the February 22, 2011, Christchurch earthquake has been used to identify liquefaction occurrence following the event. The image has multi-spectral bands with approximately 2 m spatial resolution in addition to the panchromatic band with approximately 50 cm spatial resolution. An aerial image taken after the event with 10 cm spatial resolution has been used to establish ground truth. In this study, the effectiveness of supervised and unsupervised classification methods have been evaluated in classifying different types of liquefied materials on the ground level. Liquefied materials have been classified into different categories such as wet and dry sands and a confusion matrix has been generated to evaluate the method accuracy. The K-means method is used for unsupervised classification and the minimum distance method and artificial neural network (ANN) algorithm have been used for supervised classification. ANN outperformed the other classification methods; however, all methods had difficulty in discriminating the “liquefied dry materials” and “liquefied wet materials.”