Ground-based visible-near infrared hyperspectral imaging for monitoring cliff weathering of a volcanic island in Dokdo, South Korea

Coastal cliffs undergo erosion and weathering more rapidly under the influence of strong waves and sea winds, leading to stability and environmental conservation issues. Ground-based hyperspectral imaging is useful for the identification and geological interpretation of minerals or rocks in vertical outcrops that are difficult to confirm from an aerial view or through in situ investigation for safety reasons. High spatial and spectral resolutions of visible-near infrared (VNIR) sensors can be advantageous for detecting weathering in cliffs made of volcanic rocks; however, their potential is not well known. In this study, two classification techniques, mixture-tuned matched filtering (MTMF) and support vector machine (SVM), were applied to VNIR hyperspectral data of the cliff face of a volcanic island in Dokdo, South Korea, and the classification results were compared. Results show that SVM is superior to MTMF for the classification of volcanic rocks and weathering minerals. The distinction between volcanic rocks with similar compositions and textures deteriorated using both methods. The shading of the surface owing due to unevenness and stratification also affected the accuracy of classification. This study shows that ground-based VNIR hyperspectral image analysis is a powerful and an effective approach to predict possible geomorphological changes and safety on volcanic islands, as it can explore the weathering of sea cliffs and highlight potentially vulnerable locations.

Automated summary

Coastal cliffs are prone to erosion and weathering due to strong waves and sea winds, leading to stability and environmental conservation issues. This study used ground-based hyperspectral imaging techniques to analyze volcanic island cliffs and found that support vector machine (SVM) outperformed mixture-tuned matched filtering (MTMF) in classifying volcanic rocks and weathering minerals. However, distinguishing volcanic rocks with similar compositions and textures proved to be challenging using both methods. This research highlights the effectiveness of ground-based hyperspectral imaging analysis in predicting geomorphological changes and ensuring safety on volcanic islands.