Lava flow mapping using Sentinel-1 SAR time series data: a case study of the Fagradalsfjall eruptions

<p>Lava flows can threaten populated areas, cause casualties and considerable economic damage. Therefore, understanding lava flows and their evolution is important because they can be linked to lava transport systems and eruption parameters. However, timely and accurate lava flow mapping in the field can be time-consuming and dangerous. Earth observation (EO) data plays an important role in improving lava flow mapping and monitoring. Synthetic Aperture Radar (SAR) data provide a unique opportunity to study lava flows, especially in areas with high cloud coverage during the year. Moreover, smoke and ash clouds can be partially penetrated by SAR. The freely available Sentinel-1 SAR data (C-band), with its high temporal and spatial resolution, opens new opportunities for studying lava flow evolution and lava morphology. However, Sentinel-1 data have mainly been used to study surface deformation using Differential Interferometric SAR (DInSAR) techniques, and the utilisation of SAR backscatter information for lava flow characterisation has not been thoroughly exploited.</p> <p>The Fagradalsfjall volcanic system is located on the Reykjanes Peninsula in southwest Iceland. The eruption began on the 19^th of March and lasted until the 18^th of September 2021. The resulting lava flows cover an area of 4.8 km² (Pedersen et al., 2022). Another eruption occurred in August 2022. We used time series of dual-polarisation, including VH (antenna sends vertical pulses and receives horizontal backscatter) and VV (antenna sends vertical pulses and receives horizontal backscatter), Sentinel-1 data to study the changes in lava flow extent and morphology during the 2021 and 2022 Fagradalsfjall eruption phases. The pre-processing of Sentinel-1 data included orbit state vector correction, radiometric calibration to reduce the radiometric biases caused by topographic variations, co-registration, and range doppler terrain correction. In addition to backscatter polarisations, we calculated the image texture using the grey-level co-occurrence matrix (GLCM) algorithm, including several measures such as contrast, homogeneity, and entropy. We used object-based segmentation and classification algorithms to delineate the lava extent and evaluated the applicability of different polarisations. To validate the mapping results, we used reference layers derived from high-resolution optical images available from Pedersen et al. (2022). The results showed that cross-polarisation was the most suitable for mapping the extent of lava. Additionally, the integration of texture information allowed us to distinguish lava types to some extent.</p> <p>The results demonstrate the potential and challenges of utilising SAR backscatter information from Sentinel-1 data for studying the spatio-temporal lava flow evolution and mapping lava flow morphology, especially when the applicability of optical EO data is limited.&#160;</p> <p>Pedersen, G. B. M., Belart, J. M. C., &#211;skarsson, B. V., Gudmundsson, M. T., Gies, N., H&#246;gnad&#243;ttir, T., et al. (2022). Volume, Effusion Rate, and Lava Transport During the 2021 Fagradalsfjall Eruption: Results From Near Real-Time Photogrammetric Monitoring. <em>Geophysical Research Letters</em>, 49, 13, e2021GL097125. https://doi.org/10.1029/2021GL097125</p>