The Survey of Lava Tube Distribution in Jeju Island by Multi-Source Data Fusion

Abstract

Lava tubes, a major geomorphic element over volcanic terrain, have recently been highlighted as testbeds of the habitable environments and natural threats to unpredictable collapse. In our case study, we detected and monitored the risk of lava tube collapse on Jeju, an island off the Korean peninsula’s southern tip with more than 200 lava tubes, by conducting Interferometric Synthetic Aperture Radar (InSAR) time series analysis and a synthesized analysis of its outputs fused with spatial clues. We identified deformations up to 10 mm/year over InSAR Persistent Scatterers (PSs) obtained with Sentinel-1 time series processing in 3-year periods along with a specific geological unit. Using machine learning algorithms trained on time series deformations of samples along with clues from the spatial background, we classified candidates of potential lava tube networks primarily over coastal lava flows. What we detected in our analyses was validated via comparison with geophysical and ground surveys. Given that cavities in the lava tubes could pose serious risks, a detailed physical exploration and threat assessment of potential cave groups are required before the planned intensive construction of infrastructure on Jeju Island. We also recommend using the approach established in our study to detect undiscovered potential risks of collapse in the cavities, especially over lava tube networks, and to explore lava tubes on planetary surfaces using proposed terrestrial and planetary InSAR sensors.

Keywords

• lava tube
• InSAR
• machine learning
• detection
• Jeju Island