Investigation of Recent Volcanic Activity on Reykjanes Peninsula, Iceland, as an Analog for Venus

Kurzfassung

Despite their key similarities, Venus differs from Earth in fundamental ways. However, how its evolution diverged from Earth’s and whether it was much different in the past than it is today remains unknown. Although Venus has been a center of attraction since the beginning of the era of planetary missions, due to its dense and opaque atmosphere, orbiters could only make use of radar waves to observe the surface. The last direct observation of the surface of Venus was made by the Magellan spacecraft in 1994 with its synthetic aperture radar instrument. Since then, research on the atmosphere of Venus has revealed that it enables the transmission of electromagnetic waves emitted from its hot surface. This transmission occurs through 6 spectral bands that fall within 5 atmospheric windows, all around the 1-micron wavelength. Subsequently, recent advancements in developing near-infrared imagers and laboratory measurements of the emissivity of rocks paved the way for a new era of Venus exploration. In June 2021, a fleet of Venus missions was announced that includes two orbiters aiming to leverage these 6 bands for globally mapping the surface. This study aims to inquire into the potential of near-infrared remote sensing observations in investigating recent or active volcanic activity by employing machine learning methods for the classification of surface units and change detection in preparation for future Venus missions. Iceland serves as a primary analog for Venus in addressing the research questions of this study. This is because Iceland has a young, mostly vegetation-free surface that is formed by ongoing volcanic activity. In the past three years, there have been four fissure eruptions in the Reykjanes Peninsula. Two of these eruptions, known as the 2021 and 2022 Fagradalsfjall eruptions, resulted in the formation of a 5.01 km2 flow field called Fagradalshraun. In this study, the Reykjanes Peninsula was investigated, focusing on Fagradalshraun, using the near-infrared bands of Sentinel-2A datasets acquired before and after the eruptions. Supervised classification with the Support Vector Machines method and unsupervised classification with the K-means clustering method were performed to distinguish between geological units based on their age differences. In addition, the contributions of the Texture Analysis and the Principal Components Analysis methods for improving the classification results were tested. Then, the Normalized Difference of Time Series method was used for change detection. The results indicated that even by using only the near-infrared bands, both classification methods could differentiate between fresh basalt, weathered basalt, and aeolian cover. The accuracy of the classification methods improved with the inclusion of texture measures. Additionally, it was observed that the highest accuracy in unsupervised classification was achieved through the utilization of Principal Component Analysis. Finally, it was seen that the Normalized Difference of Time Series method detected the Fagradalshraun with an accuracy of 95%.